You are here: indexbrainstorming

Les partages de midi


Qu'est-ce que c'est ?

C'est l'occasion pour les personnes de l'équipe de se retrouver autour d'une question scientifique (45 minutes), de discuter de la stratégie de l'équipe (15 minutes) tout en partageant un repas ensemble. La première partie doit nous retrouver tous. La deuxième partie est ouverte à tous, sans obligation de présence.

Comment ça se passe ?

Un membre de l'équipe prépare une présentation scientifique de 30 minutes : travaux en cours, bilan sur une question, retours sur une conférence… Les interventions peuvent se faire en anglais. Les autres membres de l'équipe participent de manière bienveillante et constructive.

  • A la fin de la présentation, un temps de partage (15 minutes) est prévu autour d'un thé ou d'un café.
  • Les 15 dernières minutes sont réservées à la stratégie/actualité de l'équipe : projet en soumission, orientation scientifique, contacts…

Quelques règles de vie :

  • Les réunions commencent et finissent à l'heure par respect pour les autres.
  • L'équipe étant inter-disciplinaire, la présentation se doit d'être didactique.
  • Tous les membres de l'équipe sans exception sont invités à participer aux partages de midi.
  • Afin que nous soyons le plus nombreux possible aux partages de midi, merci d'informer le/la coordinat-[eur/rice] de vos absences.



Mardi 1er Décembre 2020 13:00-14:00

"A semantic database for integrated management of image and dosimetric data in low radiation dose research in medical imaging"

Présentation : Bernard Gibaud

“Medical ionizing radiation procedures and especially medical imaging are a non negligible source of exposure to patients. Whereas the biological effects of high absorbed doses are relatively well known, the effects of low absorbed doses are still debated. This work presents the development of a computer platform called Image and Radiation Dose BioBank (IRDBB) to manage research data produced in the context of the MEDIRAD project, a European project focusing on research on low doses in the context of medical procedures. More precisely, the paper describes a semantic database linking dosimetric data (such as absorbed doses to organs) to the images corresponding to X-rays exposure (such as CT images) or scintigraphic images (such as SPECT or PET images) that allow measuring the distribution of a radiopharmaceutical.

The main contributions of this work are: 1) the implementation of the semantic database of the IRDBB system and 2) an ontology called OntoMEDIRAD covering the domain of discourse involved in MEDIRAD research data, especially many concepts from the DICOM standard modelled according to a realist approach.”

Mardi 10 Novembre 2020 13:00-14:00

"Flexible recurrent neural network for Audio-Visual sequence analysis"

Présenté par : Tsiry Mayet

“Deep learning models are now brought closer to the user rather than running in a distant cloud, helping to reduce latency, network congestion, and improving data security and privacy. However, smartphones and user devices impose additional constraints such as limited computation or energy. Handling these constraints has become an active research topic. User devices can also host multiple processes running at the same time and starting or stopping abruptly, modifying the constraints affecting the processes. Few works have considered models that can be modified at run time to adapt to an evolving computational limit.”

Mardi 20 Octobre 2020 13:00-14:00

"Data driven methods to support decision making in Deep Brain Stimulation for Parkinson’s Disease "

Présenté par : Maxime Peralta (Répétition de soutenance de thèse)

Mardi 13 Octobre 2020 13:00-14:00

"MISAW - Part of the Endoscopic Vision Challenge"

Présenté par : Arnaud Huaulmé

“Automatic and online recognition of surgical workflows is mandatory to bring computer-assisted surgery (CAS) applications inside the operating room. According to the type of surgery, different modalities could be used for workflow recognition. In the case where the addition of multiple sensors is not possible, the information available for manual surgery is generally restricted to video-only. In the case of robotic-assisted surgery, kinematic information is also available. It is expected that multimodal data would make easier automatic recognition methods.

The “MIcro-Surgical Anastomose Workflow recognition” (MISAW) sub-challenge provides a unique dataset for online automatic recognition of surgical workflow by using both kinematic and stereoscopic video information on a micro-anastomosis training task. Participants are challenged to recognize online surgical workflow at different granularity levels (phases, steps, and activities) by taking advantage of both modalities available. Participants can submit results for the recognition of one or several granularity levels. In the case of several granularities, participants are encouraged (but not required) to submit the result of a multi-granularity workflow recognition, i.e. recognize different granularity levels thanks to a unique model.”

Mardi 22 Septembre 2020 13:00-14:00

"Characterisation of three clinical-neurological subtypes of the bvFTD based on inhibition deficit symptoms assessed through an ecological approach"

Présenté par : Delphine Tanguy

Mardi 15 Septembre 2020 13:00-14:00

"Review of my work"

Présenté par : Kevin Le Mut

“During 1 year of contract I carried out various works at MEDICIS. My main objective was the annotation of hysterectomy videos but I also participated in the annotations of the MISAW project and started to develop an application. In addition to this, various tasks were added to complete my work such as annotation protocol build on SurgePlan. The year 2020 was full of hazards that slowed down the progress of the work.”

Mercredi 09 Septembre 2020 16:00-17:00

"Shared meeting with the VASST team"

Présenté par : Thibaut Dognon

Mercredi 08 Juillet 2020 16:00-17:00

"Shared meeting with the VASST team"

Présenté par : Maxime Peralta

Mardi 30 Juin 2020 15:00-16:00

"Paper presentation : Advancing Our Understandings of Healthcare Team Dynamics From the Simulation Room to the Operating Room: A Neurodynamic Perspective"

Présenté par : Thibaut Dognon

“The initial models of team and team member dynamics using biometric data in healthcare will likely come from simulations. But how confident are we that the simulation-derived high-resolution dynamics will reflect those of teams working with live patients? We have developed neurodynamic models of a neurosurgery team while they performed a peroneal nerve decompression surgery on a patient to approach this question. The models were constructed from EEG-derived measures that provided second-by-second estimates of the neurodynamic responses of the team and team members to task uncertainty. The anesthesiologist and two neurosurgeons developed peaks, often coordinated, of elevated neurodynamic organization during the patient preparation and surgery which were similar to those seen during simulation training, and which occurred near important episodes of the patient preparation and surgery. As the analyses moved down the neurodynamic hierarchy, and the simulation and live patient neurodynamics occurring during the intubation procedure were compared at progressively smaller time scales, differences emerged across scalp locations and EEG frequencies. The most significant was the pronounced suppression of gamma rhythms detected by the frontal scalp sensors during the live patient intubation which was absent in simulation trials of the intubation procedure. These results indicate that while profiles of the second-by-second neurodynamics of teams were similar in both the simulation and live patient environments, a deeper analysis revealed differences in the EEG frequencies and scalp locations of the signals responsible for those team dynamics. As measures of individual and team performance become more micro-scale and dynamic, and simulations become extended into virtual environments, these results argue for the need for parallel studies in live environments to validate the dynamics of cognition being observed.”

Lundi 22 Juin 2020 15:00-16:00

"Paper presentation : Optimization of targeting specific areas by optimization of medical image processing methods based on deep learning"

Présenté par : Quoc-Anh Bui

“Transcranial Magnetic Stimulation treatment is a therapeutic technique that targets a specific area of the brain, in a non-invasive way. Individual positioning of the instruments for each patient improves the quality and effectiveness of the treatment. Therefore, improving the existing registration algorithm is necessary to improve the robustness, accuracy, reproducibility and preprocessing time of imaging data.”

Jeudi 18 Juin 2020 15:00-16:00

"Paper presentation : Person-Based Versus Generalized Impulsivity Disinhibition in Frontotemporal Dementia and Alzheimer Disease"

Présenté par : Delphine Tanguy

“Background: While much disinhibition in dementia results from generalized impulsivity, in behavioral variant frontotemporal dementia (bvFTD) disinhibition may also result from impaired social cognition. Objective: To deconstruct disinhibition and its neural correlates in bvFTD vs. early-onset Alzheimer’s disease (eAD). Methods: Caregivers of 16 bvFTD and 21 matched-eAD patients completed the Frontal Systems Behavior Scale disinhibition items. The disinhibition items were further categorized into (1) ‘‘person-based’’ subscale which predominantly associated with violating social propriety and personal boundary and (2) ‘‘generalized impulsivity’’ subscale which included nonspecific impulsive acts. Subscale scores were correlated with grey matter volumes from tensor-based morphometry on magnetic resonance images. Results: In comparison to the eAD patients, the bvFTD patients developed greater person-based disinhibition (P < 0.001) but comparable generalized impulsivity. Severity of person-based disinhibition significantly correlated with the left anterior superior temporal sulcus (STS), and generalized-impulsivity correlated with the right orbitofrontal cortex (OFC) and the left anterior temporal lobe (aTL). Conclusions: Person-based disinhibition was predominant in bvFTD and correlated with the left STS. In both dementia, violations of social propriety and personal boundaries involved fronto-parieto-temporal network of Theory of Mind, whereas nonspecific disinhibition involved the OFC and aTL.”

Mardi 2 Juin 2020 15:00-16:00

"Assessing scrub nurses' non-technical skills in a virtual and collaborative environement"

Présenté par : Marie-Stéphanie Bracq

“This research is part of a multidisciplinary project to develop a virtual reality (VR) training environment in non-technical skills (NTS) for scrub nurses. The objective of this thesis was to provide the partners with the choice of NTS, to define training scenarios and to provide a first proof of concept on a targeted population. According to the literature and interviews with expert scrub nurses, the NTS selected for the scenarios are situation awareness (SA) and speaking up. Study 1, carried out with 13 scrub nurses and 16 non-experts, aims to assess the acceptability of the environment with the “instrumentation table” scenario (under VR helmet). Study 2, involving 18 scrub nursing students and 8 scrub nurses during an in-service course, aims to develop situation awareness, using the “virtual operating room of errors” scenario (under VR helmet). This scenario enables SA to be assessed and seems a suitable training tool for this NTS. Finally, Study 3, involving 33 scrub nursing students, aims to develop speaking up with a tablet scenario. Our results partially confirm the effect of status on the probability of speaking up and show an effect of status on the discomfort of speaking up. Each study involved numerous exchanges with the project's VR engineers and scrub nursing teachers in order to allow the construction of the environment and its future integration into the scrub nursing school pedagogy. Finally, for each study, specific assessment tools were used and simulation sessions were developed according to recommendations with briefing, simulation, debriefing (HAS, 2012).”

Mardi 26 Mai 2020 15:00-16:00

"Paper presentation : Advancing Our Understandings of Healthcare Team Dynamics From the Simulation Room to the Operating Room: A Neurodynamic Perspective"

Présenté par : Thibaut Dognon

“The initial models of team and team member dynamics using biometric data in healthcare will likely come from simulations. But how confident are we that the simulation-derived high-resolution dynamics will reflect those of teams working with live patients? We have developed neurodynamic models of a neurosurgery team while they performed a peroneal nerve decompression surgery on a patient to approach this question. The models were constructed from EEG-derived measures that provided second-by-second estimates of the neurodynamic responses of the team and team members to task uncertainty. The anesthesiologist and two neurosurgeons developed peaks, often coordinated, of elevated neurodynamic organization during the patient preparation and surgery which were similar to those seen during simulation training, and which occurred near important episodes of the patient preparation and surgery. As the analyses moved down the neurodynamic hierarchy, and the simulation and live patient neurodynamics occurring during the intubation procedure were compared at progressively smaller time scales, differences emerged across scalp locations and EEG frequencies. The most significant was the pronounced suppression of gamma rhythms detected by the frontal scalp sensors during the live patient intubation which was absent in simulation trials of the intubation procedure. These results indicate that while profiles of the second-by-second neurodynamics of teams were similar in both the simulation and live patient environments, a deeper analysis revealed differences in the EEG frequencies and scalp locations of the signals responsible for those team dynamics. As measures of individual and team performance become more micro-scale and dynamic, and simulations become extended into virtual environments, these results argue for the need for parallel studies in live environments to validate the dynamics of cognition being observed.”

Mardi 26 Mai 2020 15:00-16:00

"Paper presentation : Connectivity predicts deep brain stimulation outcome in Parkinson Disease"

Présenté par : Alfonso Estudillo

“Objective: The benefit of deep brain stimulation (DBS) for Parkinson disease (PD) may depend on connectivity between the stimulation site and other brain regions, but which regions and whether connectivity can predict outcome in patients remain unknown. Here, we identify the structural and functional connectivity profile of effective DBS to the subthalamic nucleus (STN) and test its ability to predict outcome in an independent cohort.

Methods: A training dataset of 51 PD patients with STN DBS was combined with publicly available human connectome data (diffusion tractography and resting state functional connectivity) to identify connections reliably associated with clinical improvement (motor score of the Unified Parkinson Disease Rating Scale [UPDRS]). This connectivity profile was then used to predict outcome in an independent cohort of 44 patients from a different center.

Results: In the training dataset, connectivity between the DBS electrode and a distributed network of brain regions correlated with clinical response including structural connectivity to supplementary motor area and functional anticorrelation to primary motor cortex (p < 0.001). This same connectivity profile predicted response in an independent patient cohort (p < 0.01). Structural and functional connectivity were independent predictors of clinical improvement (p < 0.001) and estimated response in individual patients with an average error of 15% UPDRS improvement. Results were similar using connectome data from normal subjects or a connectome age, sex, and disease matched to our DBS patients.

Interpretation: Effective STN DBS for PD is associated with a specific connectivity profile that can predict clinical outcome across independent cohorts. This prediction does not require specialized imaging in PD patients themselves.”

Mardi 11 Mai 2020 15:00-16:00

"Paper presentation : Intraoperative Localization of STN During DBS Surgery Using a Data-Driven Model"

Présenté par : Thibault Martin

“A new approach is presented for localizing the Subthalamic Nucleus (STN) during Deep 
Brain Stimulation (DBS) surgery based on microelectrode recordings (MERs). DBS is an accepted treatment for individuals living with Parkinson’s Disease (PD). This surgery involves implantation of a permanent electrode inside the STN to deliver electrical current. Since the STN is a very small region inside the brain, accurate placement of an electrode is a challenging task for the surgical team. Prior to placement of the permanent electrode, microelectrode recordings of brain activity are used intraoperatively to localize the STN. The placement of the electrode and the success of the therapy depend on this location. In this paper, an objective approach is implemented to help the surgical team in localizing the STN. This is achieved by processing the MER signals and extracting features during the surgery to be used in a Machine Learning (ML) algorithm for defining the neurophysiological borders of the STN. For this purpose, a new classification approach is proposed with the goal of detecting both the dorsal and the ventral borders of the STN during the surgical procedure. Results collected from 100 PD patients in this study, show that by calculating and extracting wavelet transformation features from MER signals and using a data-driven computational deep neural network model, it is possible to detect the borders of the STN with an accuracy of 92%. The proposed method can be implemented in real-time during the surgery to model the neurophysiological nonlinearity along the path of the electrode trajectory during insertion.”

Mardi 5 Mai 2020 15:00-16:00

"Paper presentation : Deep Double Descent: Where Bigger Models and More Data Hurt"

Présenté par : John Baxter

“We show that a variety of modern deep learning tasks exhibit a “double-descent” phenomenon where, as we increase model size, performance first gets worse and then gets better. Moreover, we show that double descent occurs not just as a function of model size, but also as a function of the number of training epochs. We unify the above phenomena by defining a new complexity measure we call the effective model complexity and conjecture a generalized double descent with respect to this measure. Furthermore, our notion of model complexity allows us to identify certain regimes where increasing (even quadrupling) the number of train samples actually hurts test performance.”

Mardi 5 Mai 2020 15:00-16:00

"Paper presentation : Development of an Interactive Dashboard to Analyze Cognitive Workload of Surgical Teams During Complex Procedural Care"

Présenté par : Arnaud Huaulmé

“In the surgical setting, team members constantly deal with a high-demand operative environment that requires simultaneously processing a large amount of information. In certain situations, high demands imposed by surgical tasks and other sources may exceed team member's cognitive capacity, leading to cognitive overload which may place patient safety at risk. In the present study, we describe a novel approach to integrate an objective measure of team member's cognitive load with procedural, behavioral and contextual data from real-life cardiac surgeries. We used heart rate variability analysis, capturing data simultaneously from multiple team members (surgeon, anesthesiologist and perfusionist) in a real-time and unobtrusive manner. Using audio-video recordings, behavioral coding and a hierarchical surgical process model, we integrated multiple data sources to create an interactive surgical dashboard, enabling the analysis of the cognitive load imposed by specific steps, substeps and/or tasks. The described approach enables us to detect cognitive load fluctuations over time, under specific conditions (e.g., emergencies, teaching) and in situations that are prone to errors. This in-depth understanding of the relationship between cognitive load, task demands and error occurrence is essential for the development of cognitive support systems to recognize and mitigate errors during complex surgical care in the operating room.”

Mardi 28 Avril 2020 15:00-16:00

"Paper presentation : Biologically-inspired dynamical systems for movement generation: automatic real-time goal adaptation and obstacle avoidance"

Présenté par : Julien Berthelemy

“Dynamical systems can generate movement trajectories that are robust against perturbations. This article presents an improved modification of the original dynamic movement primitive (DMP) framework by Ijspeert et al [1], [2]. The new equations can generalize movements to new targets without singularities and large accelerations. Furthermore, the new equations can represent a movement in 3D task space without depending on the choice of coordinate system (invariance under invertible affine transformations). Our modified DMP is motivated from biological data (spinal-cord stimulation in frogs) and human behavioral experiments. We further extend the formalism to obstacle avoidance by exploiting the robustness against perturbations: an additional term is added to the differential equations to make the robot steer around an obstacle. This additional term empirically describes human obstacle avoidance. We demonstrate the feasibility of our approach using the Sarcos Slave robot arm: after learning a single placing movement, the robot placed a cup between two arbitrarily given positions and avoided approaching obstacles.”

Mardi 14 Avril 2020 15:00-16:00

"Academic background and PhD project"

Présenté par : Delphine Tanguy

“A quick summary of my academic background followed by a presentation of my thesis project: the neural basis of disinhibition in neurodegenerative diseases. This will be a first introduction to my PhD project and a description of the protocols used, without any results presentation.”

Mardi 7 Avril 2020 15:00-16:00

"Overview of my past projects"

Présenté par : Alfonso Estudillo

“In this meeting I am going to introduce a bit of myself, where I came from and what I did in my master and PhD theses. I am going to also show some projects I have involved with and finally the current project at MediCIS.”

Vendredi 6 Mars 2020 13:00-14:00

"History of Medecine, Study of medical and surgical knowledge and practices over time"

Présenté par : Alicia Diot

“The History of Medicine is the study of medical and surgical knowledge and practices over time. Thanks to the discoveries of traces of ancient medical knowledge, we understand that medical practice has always existed, the fight against pain and infirmity is as old as Humanity. What has evolved over the centuries are the means used in this fight, from magic to modern scanners, from shamans to specialized surgeons … Medicine is a discipline at the crossroads between Art and Science. From the origin of civilizations, it has been considered as an art, in the measure of intuition, empiricism, psychology … Over time and with progress, it has become a science, based on the understanding of the human and its pathologies, on the identification of therapeutics, and the discovery of new techniques. Thus, while the art of healing is as old as mankind, Medicine as we know it today only dates back to the beginning of the 19th century. Since then, progress has continued at an impressive speed that keeps accelerating. It therefore seems difficult to retrace the evolution of Medicine in an exhaustive way, however, I will outline the main stages and dominant discoveries that have given rise to today’s medicine. It is important to understand that the division into chronological periods that I use in this presentation remains very artificial, the evolutions are actually very intertwined. Likewise, the great names that I cite as having largely contributed to the evolution of Medicine, are ultimately only a tiny part of the crowd that has participated in this evolution.”

Mardi 11 Fevrier 2020 13:00-14:00

"Topology-Aware Activation Layer for Neural Network Image Segmentation"

Présenté par : John Baxter

“One of the recent developments in deep learning is the ability to train extremely deep residual neural networks, knowing that each residual block produces only marginal changes to the data. The accumulation of these changes result in the network's improved performance, analogous to a complex but trainable iterative algorithm. This intuition can be merged with the underlying theory of probabilistic graphical models in which these iterative algorithms are common and share the underlying probabilistic and information theoretic basis as deep learning. Prior models have been proposed with limitations on the number of iterations allowed in the solution algorithm due to the linear memory growth during the training process. This paper presents a structured activation layer which implements a conditional random field along with an arbitrary iteration message-passing marginal probability estimation algorithm which requires constant, rather than linear, memory with respect to the number of iterations. In this activation layer, the segmentation labels can be specified hierarchically, incorporating a level of abstract structure and some basic geometrical knowledge directly and easily into the network. Thus, this layer allows for the separation of abstract knowledge brought in by the network designer (in the form of the hierarchical structure) from probabilistic priors learned by the neural network. Preliminary results comparing this activation function to softmax and a similar non-hierarchical activation function indicate that it significantly improves performance in segmentation problems.”

Mardi 11 Fevrier 2020 13:00-14:00

"Localisation of the Subthalamic Nucleus in MRI via Convolution Neural Networks for Deep Brain Stimulation Planning"

Présenté par : John Baxter

“Deep brain stimulation (DBS) is an interventional treatment for Parkinson's disease in which electrodes are placed into specific locations in the basal ganglia to alleviate symptoms such as tremor and dyskinesia. Due to the small size and low contrast of specific targets of interest, such as the subthalamic nucleus (STN), localisation of these structures from pre-operative MRI is of great value. These localisation approaches are often atlas-based, using registration algorithms to align patient images with a prior annotated atlas. However, these methods require a large amount of time, especially for correctly estimating deformation fields, or are prone to error for smaller structures such as the STN. This paper investigates two deep learning frameworks for the localisation of the stn from T1- and T2-weighted MRI using convolutional neural networks which estimate its centroid. These methods are compared against an atlas-based segmentation using the ParkMedAtlis v3 atlas, showing an improvement in localisation error in the range of ~0.5-1.3 mm with a reduction of orders of magnitude of computation time. This method of STN localisation will allow us in future to automatically identify the STN for DBS surgical planning as well as define a greatly reduced region-of-interest for more accurate segmentation of the STN.”

Mardi 28 Janvier 2020 13:00-14:00

"Clinical and technological study of the spine biomechanics, scoliosis-oriented"

Présenté par : Kevin Le Mut

“Scoliosis is the largest three-dimensional deformation in the human body. It causes many health issues and there is a lack of data to correct them properly. Surgeons can only use their experience to operate. The role of biomechanics and finite element simulations is to provide more data and to analyze the constraints within the spine.”

Mardi 21 Janvier 2020 13:00-14:00

"Convolutional Neural Network for Localizing the Subthalamic Nucleus Using One Second Micro-electrode Recordings"

Présenté par : Maxime Peralta

“Micro-electrode recording (MER) is a powerful way of localizing target structures during neurosurgical procedures such as the implantation of deep brain stimulation electrodes which is a common treatment for Parkinson's disease and other neurological disorders. Although allowing better clinical results than guiding only with per-operative imaging, MER is not unanimously used in clinic, notably because of its long duration which can lead to complications during the operation, and because it requries a high degree of expertise. Different approaches towards automating MER analysis and structure localization have been proposed on the last decade, mainly focused on feature engineering. While the accuracies obtained are good in certain configurations, one issue with handcrafted features is that they do not necessarily capture more subtle differences in MER that are detected by expert neurophysiologists by ear. In this presentation, i'll show you a deep learning based pipeline for subthalamic nucleus (STN) localization with micro-electrode recordings which is motivated by the human auditory system. The proposed Convolutional Neural Network (CNN) shows improved accuracy over two comparative networks for locating the STN from one second MERs.”

Mardi 7 janvier 2020 13:00-14:00

"Kinematic data analysis for skill assessment in surgery: a review"

Présenté par : Julien Berthelemy

Mardi 22 Octobre 2019 13:00-14:00

"Elephants in the Room: How should we analyse trainee performance on simulator-based scenarios?"

Présenté par : Roy Eagleson

Mardi 08 Octobre 2019 13:00-14:00

"What will happen to you if you go in medical device industry"

Présenté par : Ehouarn Maguet

“I will share with you some part of the daily life of a software engineer in a medical device industry. I will talk about medical standard (and try not to be boring), audit, impact of software and hardware on regulation, a resolution bug example, and some anecdotes.”

Mardi 01 Octobre 2019 13:00-14:00

"Cognitive Oriented Design and Assessment of Augmented Reality in Medicine"

Présenté par : Pierre Jannin

“Mixed and Augmented Reality (AR) is a technology that aims to augment the perception of reality by adding computer-generated information. Such technology has been applied for medical procedures (especially surgery) the past 20 years to help physicians understand and perform procedures, as well as for training. However, as outlined in different review papers, AR has not yet reached its full potential usefulness and implementation in the medical workflow. Many reasons for this have been expressed; one is that the AR system design approach is mainly technology driven, resulting in the need to develop relevant and clinically valuable AR applications. Another reason is the lack of relevant assessments of proposed AR systems to validate their benefits in all aspects of medical care. In this presentation, I will suggest a cognitive-oriented approach for both optimal design and assessment of AR systems. This approach is based on two main statements: (1) ensuring proper definition and modeling of medical objective(s) and (2) considering understanding as the main motivation for AR.”

Mardi 17 Septembre 2019 13:00-14:00

"Augmented Reality Needle Guidance System and Multi-Modal Brain Connectivity Visualization for Pre-Surgical Planning"

Présenté par : Denis Kikinov

“I will introduce my ongoing industrial project: a lightweight training simulator that allows clinicians and trainees to practise intramuscular injections. The system allows viewing of the underlying muscles when injecting with a syringe, provides EMG feedback and provides visual guidance in the case that the user wants to target a specific muscle. Our prototype is composed of a synthetic arm model (comprised of muscles, bones, simulated fat and skin), real syringes and needles, and android tablet, and a support frame. Although a prototype has been created, it has yet to be tested among clinicians. In my presentation I will also introduce our new mobile device AR application based on data derived from advanced imaging processing of neuroimaging data for both a healthy brain, and one with a brain tumour. The application uses a multi-modal approach implementing the structural and functional connectivity of the specified patient’s anatomy and visualizes their brain in augmented reality. This approach allows the user to interact in real time with the neuroanatomy in the context of its associated SC and FC. The project is very lightweight, requiring only an Android device and a couple pieces of paper.”

Mardi 10 Septembre 2019 13:00-14:00

"Improving activity recognition based on a priori about surgeon biomechanical behavior"

Présenté par : Zakaria Kaddour

“Robot assisted surgery has been a rising research field in the past few year with the introduction of the da Vinci Research Kit and similar tools and in order to supply better assistance the robot requires knowledge about the task performed by the surgeon. In this presentation I will talk about my previous work on the anatomy training VR platform and I move to talking about my current work which revolves around defining a constraint based on kinematic data collected from the surgeon during the surgery and use prior knowledge about the task as a whole and on the different gestures comprising it to improve activity recognition. This work can be done through the application of a LSTM deep learning model based on the kinematic data introduced with the priori.”

Mardi 04 Juin 2019 13:00-14:00

"Situation awareness in the “Virtual Operating Room of Errors”: a pilot study"

Présenté par : Marie-Stéphanie Bracq

“Training healthcare professionals in non-technical skills has revealed a crucial issue for patient safety and quality of care. Simulation offers good opportunities to follow that goal. Technologies for simulation are emerging, among which, virtual reality (VR). Yet, few VR simulators address non-technical skills. The “Virtual Operating Room of errors” scenario aims at training situation awareness in a VR environment.”

Lundi 27 Mai 2019 13:00-14:00

"Sub-cortical shape alteration analysis for automated Parkinson Disease Diagnosis with Machine Learning"

Présenté par : Maxime Peralta

“Brain morphometry is still considered as a secondary source of information for the diagnosis of Parkinson Disease. Indeed, clinical teams mostly make their diagnosis with clinical biomarkers. However, the increasing accuracy of MRI acquisition encourages research toward finding reliable and easy to acquire morphometric biomarkers to diagnose, track and monitor the progression of the disease. Among the techniques explored to analyze the morphometric variations occurring in Parkinson Disease, shape analysis has proved itself more reliable and consistent than volume analysis, thickness analysis and volume analysis. Several studies tried to evaluate the significance of subcortical shape displacements as a diagnosis biomarker for Parkinson Disease, leading to often contradictory results.

I will present you, in this noon sharing, a fully automated pipeline for sub-cortical alteration analysis involving, for the fist time in the litterature, machine learning. This pipeline is able to successfully classify several cohorts of PD patients with various disease severity, as well as an healthy control cohort, thus showing that crucial informations correlated to disease severity are embedded into bilateral caudate nucleus and putamen.”

Mardi 14 Mai 2019 13:00-14:00

"Overview of automatic classification methods to assess the skills of the surgeon during training using the kinematics data"

Présenté par : Julien Berthélemy

“I will present a few studies about the automatic assessment of the surgeon's skills based on recordings of the kinematics data during a training exercice. Every article I will present uses a different method of machine learning or deep learning on the same JIGSAWS dataset. After that, I will compare the results of those classification methods and discuss them in order to develop my own method based on the ones giving the most promising results.”

Mardi 30 Avril 2019 13:00-14:00

"How brain connectivity could help to predict DBS clinical outcomes ?"

Présenté par : Tristan Moreau

“Detailed anatomical knowledge and quantification of white matter properties are of particular therapeutic importance for Deep Brain Stimulation (DBS), an invasive neuromodulatory therapy which aims to balance disorganized motor circuits by applying high-frequency current to a target nucleus within basal ganglia. This presentation will take the form of a discussion around some interesting papers.”

Mardi 02 Avril 2019 13:00-14:00

"Introduction aux modèles graphiques probabilistes"

Présenté par : John Baxter

“Ma présentation est une introduction aux modèles graphiques probabilistes, spécifiquement les modèles max-flow, et leur intégration dans le cadre de l’apprentissage profond. Ces modèles nous permettent d’ajouter directement l’information topologique dans un réseau de neurones artificiels pendant son apprentissage et son application. La présentation inclura une explication brève sur les modèles max-flow et leurs solutions approximatives, les problèmes avec la combinaison de ces modèles et de l’apprentissage automatique, et DeepFlow qui vise à résoudre ces problèmes pour la segmentation d’images.”

Mardi 26 Mars 2019 13:00-14:00

"Future work regarding CARS conference and MediCIS' website"

Présenté par : Pierre Jannin

“This week's lab meeting will mainly be about two subjects : the organisation of the CARS conference and the evolution of the MediCIS website.

As you all know, the CARS conference is coming up which means we should take time to discuss once again about how each of us will take part in this project as well as present what has already been done.

When it comes to the MediCIS website, the goal of this meeting will be to discuss what could or should be done to renew it.”

Mardi 19 Mars 2019 13:00-14:00

"Treating chronic pain using neuronavigated Transcranial Magnetic Stimulation (TMS)"

Présenté par : Syneika

“Up to 7% of the general population suffers from neuropathic pain which comes from various sources such as neurological diseases or peripheral nerve lesions. However, more than half of these patients say they are not relieved by their medical treatment. Neuronavigated Transcranial Magnetic Stimulation (TMS) is a non-invasive and virtually painless therapy that modulates the activity of precise neural networks of the cerebral cortex using high intensity magnetic pulses. TMS has demonstrated its effectiveness in the treatment of several neurological and psychiatric diseases.

Syneika offers a complete and accurate solution of neuronavigated Transcranial Magnetic Stimulation, and allows each doctor to individualize his treatment by taking into account the anatomy of each patient. A solution that can also be used very simply by the healthcare team.”

Mardi 5 Fevrier 2019 13:00-14:00

"Automatic annotation of surgical procedures"

Présenté par : Arnaud Huaulme

“Annotation of surgical activities becomes increasingly important for many recent applications such as surgical workflow analysis, surgical situation awareness and the design of the operating room of the future, especially to train machine learning methods in order to develop intelligent assistance. Currently, annotation is mostly performed by observers with medical background and is incredibly costly and time-consuming, creating a major bottleneck for the above-mentioned technologies. However, meaningful information about interaction between objects is inherently available in virtual-reality environments. We propose a strategy to convert automatically this information into annotations in order to provide as output individual surgical process models. We implemented our approach through a peg-transfer task simulator and compared it to manual annotations performed by six members of the lab.”

Jeudi 31 Janvier 2019 13:00-14:00

"Comprendre et maîtriser les facteurs humains et organisationnels pouvant influencer les résultats individuels du chirurgien"

Présenté par : Antoine Duclos

“Cette présentation abordera la mesure, l’interprétation et la maîtrise des résultats individuels du chirurgien. Un premier axe méthodologique ciblera le développement d’indicateurs et d’outils graphiques pour guider les chirurgiens vers une meilleure compréhension de leur performance. Un second axe questionnera la courbe de performance des chirurgiens tout au long de leur carrière avec l’espoir d’identifier des facteurs humains et organisationnels susceptibles d’améliorer la qualité, la sécurité et l’efficience des procédures réalisées sur le patient dans le contexte du bloc opératoire.”

Mardi 22 Janvier 2019 13:00-14:00

"Medical data Translation for semantic storage"

Présenté par : Marine Brenet

“I will be presenting my work for the Medirad Project. For this project many medical data will be used. My server will translate all these data from european countries to store them and allow easy and powerfull querring.”

Mardi 15 Janvier 2019 13:00-14:00

"Presentation of the pyDBS project"

Présenté par : Boubacar Ndack

“For the next noon sharing, I will present the pyDBS project. As you know, the project involves developing an image guidance and assistance environment for the Deep Brain Stimulation (DBS) neurosurgery procedure. This procedure involves implanting electrical stimulation electrodes deep into the brain, in order to target specific anatomical areas that will respond to the applied electrical stimulation.

I'll do a little reminder of the deep brain stimulation. After that, I'll present our work on this project.”

Mardi 8 Janvier 2019 13:00-14:00

"Overview of my first year of PhD"

Présenté par : Maxime Peralta

“For this noon sharing, i will quickly present the context and the objectives of my thesis, before presenting you the work i did so far. The first project is about clinical data compression using neural networks. The second one is the work done by Antoine Ackaouy and I on electrophysiological signal analysis. Finally, i'll present you the work i did in the team of Ali Khan at Robarts Imaging (London, Ontario, Canada), and the perspectives of this project.”

Mardi 11 Décembre 2018 13:00-14:00

"In beer we must trust!"

Présenté par : Ehouarn Maguet

“I will attempt to get you more interested in one central element of our society, beer and beer processing. France is historically a country of wine, but recently the world wide trend of craft beer invaded our country, and it tries to prove that beer can be as complex and creative as wine. I will talk about raw ingredients, some historical anecdotes, industrial beer VS local production, beer styles classification, etc. In a second part I will present you the three main steps of beer production, and try to convince you to brew your own beer at home !”

Mercredi 5 Décembre 2018 13:00-14:00

"Appel à participation et volontariat pour CARS 2019"

Présenté par : Pierre Jannin

“Pierre Jannin will make a presentation about the next edition of CARS which will take place in Rennes in June 2019. The goal of this meeting is to discuss the organisation of the next edition of CARS as well as how the members of the MediCIS team and the whole LTSI institute can take part in this project.”

Mardi 27 Novembre 2018 13:00-14:00

"Virtual Reality for Healthcare."

Présenté par : Marie Le Duff

“Technology is changing the face of healthcare as we know it. Although VR tends to be associated with the entertainment genre and indeed set sails to enhance the demanding gamer’s experience, virtual environment are now becoming a real game-changer in healthcare, and has wide applications ranging from education, training, diagnosis and therapy. I will start my presentation with a brief introduction to VR technology and its history, and then present VR uses in medical environments. You will also be invited to try some VR experiences.”

Mardi 20 Novembre 2018 13:00-14:00

"Analysis and quantification of quality in laparoscopy based on spatial and temporal metrics."

Présenté par : Arthur Derathé

“We work on the evaluation of the surgical skills of confirmed surgeons during a real procedure based on the annotation of the temporal and the spatial aspects of the laparoscopic video, as well as the annotation of a quality score. Our methodology is to observe the behaviour of a machine learning algorithm as we optimize its hyper parameters. Of course our goal is to get the best prediction of the quality score, but I will discuss more closely how the metrics cluster (temporal or spatial) in input and the cross validation strategy influences the algorithm prediction. Finally I will introduce the next phase of my work, which is a close analysis of how do the input metrics behave in relation to the quality score, and how we can translate these analysis to the clinical realm.”

Mardi 13 Novembre 2018 13:00-14:00

"Special presentation : CARS."

Présenté par : Pierre Jannin

“After a quick and general presentation about what CARS is, we will discuss how the members of the MediCIS team and the whole LTSI institute can take part in this project.”

Mardi 16 octobre 2018 13:00-14:00

"Evaluation of surgical skills using physiological signals and machine learning, organisation of my work."

Présenté par : Thibaut Dognon

“The evaluation of surgical skills is usually based on observation which is limited both by the need for experts to be present and by the subjectivity of the examiner. The position of instruments or physiological measures recorded during a surgical procedure might be helpful to go beyond the subjectivity of the expert’s supervision. The goal of this study is to obtain unbiased, quantifiable and reproductible metrics to assess surgical skills.

As the internship has ended and the PhD starts I will be presenting what has already been done, how I plan to organise my work as well as what I have learned during the internship that will benefit me for the completion of this project.”

Mardi 9 octobre 2018 13:00-14:00

"The use of virtual reality for non-technical skills."

Présenté par : Marie-Stephanie Bracq

I will present my work on the evaluation of scrub nurses' non-technical skills in a virtual environment. More specifically, I will present the results of my first study on the acceptability of the virtual training environment by experts (scrub nurses from the neurosurgery department of the University Hospital) as well as by non-experts (some of whom are lab members). I will then adress the next steps with the “operating room of errors” scenario, which is currently being finalised, and finally with the very rapid evocation of a scenario on communication currently being developed.

Mardi 25 Septembre 2018 13:00-14:00

"Special Noon Sharing : 90 to 120 seconds to pitch a crazy idea."

Présenté par : Tout le laboratoire

Vendredi 29 juin 2018 14:00-15:00

"Goals, guidelines, and best practices: a framework for designing augmented reality ultrasound guidance systems; or if you build it, will they come?!"

Présenté par : Golafsoun Ameri

Real-time ultrasound has become a crucial aspect of several image-guided interventions. One of the main constraints of such an approach is the difficulty in interpretability of the limited field of view of the image, a problem that has recently been addressed using mixed reality environments, such as augmented reality (AR). The growing popularity and maturity of AR has led to a series of informal guidelines to direct development of new systems and to facilitate regulatory approval. However, the goals of AR image guidance systems and the guidelines for their development have not been thoroughly discussed. The purpose of this talk is to identify the goals, guidelines, and best practices for the development of AR ultrasound guidance and encourage discussion on the effective development of such systems.

Jeudi 28 juin 2018 13:00-14:00

"Analysis of surgical skills using physiological signals."

Présenté par : Thibaut Dognon

With the increasing number of surgical procedures performed every year, the precise evaluation of surgical skills is crucial. This evaluation usually consists in a trainee performing a specific task and an examiner evaluating the quality of the procedure with a standardized score (NOTSS, GOALS, OSATS, etc.). This approach is limited both by the need for experts to be present and by the subjectivity of the examiner. Information taken from the position of instruments or physiological measures might be helpful to go beyond the subjectivity of the expert’s supervision. EEG signals have already been used to predict the performance of subjects during the training of sports-related skills. They have also been used to monitor the training improvements in a motor task or visuomotor sequence learning. Using machine or deep learning approaches and complex features from the EEG signals both in sensor and source space, this study aims at obtaining unbiased, quantifiable and reproductible metrics to assess surgical skills.

Mardi 12 juin 2018 13:00-14:00

"Sequential surgical signatures in micro-suturing task."

Présenté par : Arnaud Huaulmé

Surgical processes are generally only studied by identifying differences in populations such as participants or level of expertise. But the similarity between populations is also important in understanding the process. We therefore proposed to study these two aspects. In this presentation, we will demonstrate the concept of sequential surgical signature thank to a pattern mining method applied to a micro-surgical suturing task data set.

Jeudi 7 juin 2018 13:00-14:00

"Electroencephalography Source Connectivity: Aiming for High Resolution of Brain Networks in Time and Space."

Présenté par : Mahmoud Hassan

The human brain is a large-scale network the function of which depends on dynamic interactions between spatially distributed regions. In the rapidly evolving field of network neuroscience, two unresolved challenges hold the promise of potential breakthroughs. First, functional brain networks should be identified using noninvasive and easy-to-use neuroimaging techniques. Second, the time-space resolution of these techniques should be good enough to assess the dynamics of the identified networks. Emerging evidence suggests that the electroencephalography (EEG) source-connectivity method may offer solutions to both issues, provided that scalp EEG signals are appropriately processed. Therefore, this technique's performance strongly depends on signal processing that involves various methods, such as preprocessing approaches, inverse solutions, statistical couplings between signals, and network science.

Jeudi 31 mai 2018 13:00-14:00

"Special Noon Sharing: Your work in 90 seconds."

Présenté par : Tout le laboratoire.

Jeudi 17 et jeudi 24 mai 2018 13:00-14:00

"Interaction in medical image segmentation."

Présenté par : John Baxter

Interaction in medical image segmentation:

Human-computer interaction as a tool-kit is diverse but also infrequently used to provide the context behind movements in computer science. New developments in hardware and algorithms tend to take precedence in explaining these advancements more so than an understanding of how they are actually used. The goal of this set of presentations is to introduce concepts from human-computer interaction and how it applies to segmentation programs in order to help spur a more out-of-the-box thought process to a familiar topic.

Part 1: Olabarragia & Smeulders, “Interaction in the segmentation of medical images: a survey”, (2001)

After introducing concepts from HCI, we will explore the first major work in understanding medical image segmentation (and ultimately medical image processing in general) from the perspective of human-computer interaction written by Olabarragia & Smeulders in 2001 for Medical Image Analysis. At this point in the development of segmentation algorithms, many of the most common interaction mechanisms have both been developed and were also about to be developed, making this article a particularly good snapshot into understanding what goes into developing a philosophical framework to understand a complex computational process.

Part 2: Baxter et al., “The Semiotics of Medical Image Segmentation”, (2017)

My review paper, written in 2017 for Medical Image Analysis, is largely a response to that of Olabarragia & Smeulders. Specifically, this paper proposes a framework that explains Olabarragia & Smeulders’ concluisions while at the same time being more rigorously justified in order to be applicable for a longer span of time. The framework is built on the idea that segmentation can be understood as an exchange of information between a human and a computer, mediated by signs, and applying semiotics and cognitive science to this understanding can give us a rubric as to how to analyze and improve these signs.

Mardi 3 avril 2018 13:00-14:00

"Festivals de printemps et d'été à Rennes et autour."

Présenté par : Gurvan Lecuyer

My presentation is about the main music festivals of the spring and summer in Rennes, Brittany and around. The purpose of this presentation is to introduce newcomers to this incredible occasion to have fun while listening good music. I will present the lineup of each festivals and the experience you can expect for those I've attended.

Mardi 27 mars 2018 13:00-14:00

"Presentation of the differents ongoing projects of MediCIS."

Présenté par : Pierre JANNIN

Mardi 20 mars 2018 13:00-14:00

"Presentation of the "Délégation à la Recherche Clinique et à l'Innovation (DRCI)"."

Présenté par : Loïc FIN

Lundi 12 mars 2018 13:00-14:00

"Simulation, performance and human factors in obstetrics and gynecology."

Présenté par : Maela LE LOUS

Mardi 20 février 2018 13:00-14:00

"Exploring Fusion Models in Computer Vision for Medical Image Computing (part 2)."

Présenté par : Duygu Sarikaya

Mardi 13 février 2018 13:00-14:00

"Exploring Fusion Models in Computer Vision for Medical Image Computing."

Présenté par : Duygu Sarikaya

Automatic understanding of medical images has been an active research area, as a need for universally accepted, standard and validated metrics is addressed in the community. The recent advances of computer vision and machine learning algorithms coupled with the use of digital imaging modalities has led to improvements in automatic understanding of medical imaging data. In this talk, we explore fusion models in computer vision for medical image computing. We propose solutions to the open problems of robot-assisted surgery (RAS) video understanding, and segmentation of the tumor and anatomical structures of the brain. Modeling the gestures and skill level of surgeons in robot-assisted surgery (RAS) videos presents an interesting problem. Early identification of technical competence in surgical skills is expected to help tailor training to personalized needs of surgeons in training. The insights drawn may be applied in effective skill acquisition, objective skill assessment, real-time feedback, and human-robot collaborative surgeries. Characterization of anatomical structure of the brain in magnetic resonance human brain images (MRI) has also gained high interest in recent years. There is a strong need for a computer-aided system that will automatically and accurately define the tumor and anatomical structures of the brain. A universally-accepted, standardized automatic mapping and segmentation of brain volumetry would be a significant improvement for diagnosis, modeling personalized progression, and treatment monitoring of neurologic conditions. We show that exploiting the relationship across different tasks, jointly reasoning multi-tasks, and taking advantage of a combination of shared and task-specific representations, or using smart fusion of a priori knowledge inferred from multiple data sources for these problems perform astonishingly better than their single-task counterparts.

Mardi 6 février 2018 13:00-14:00

"Study of anatomo-clinical correlation in Deep Brain Stimulation by Machine Learning: A presentation of my thesis."

Présenté par : Maxime Peralta

Deep Brain Stimulation (DBS) is a surgical procedure used notably for conteract the motor effects of Parkinson Desease. Even though it shows promising results, several unknowns and difficulties remain. Artifical intelligence, through Machine Learning (ML) or Deep Learning (DL), is forseen to be one of the solution to come up with these issues. During this presentation, i will quickly introduce DBS and ML/DL. After that, i will present the objectives of my thesis by talking about the difficulties of using deep learning algorithms in neurosurgery and introducing tractography. I'll then present a study in progress with Clément Baumgarten and finish by introducing the objectives of my next study.

Mardi 30 janvier 2018 13:00-14:00

"Relationships between Expertise, Team Familiarity and Surgical Workflow Disruptions: an Observational Study."

Présenté par : Pierre-Louis Henaux

A workflow disruption (WD) is characterized by deviations from the natural progression of an operation, thus possibly compromising operation safety. Disruptions in the operating room (OR) can be subdivided into communication, usability, physical layout, environmental hazards, general interruptions and equipment failures. Communication failures in the OR can occur in approximately 30% of team exchanges and a third of these causes of WD endanger patient safety. Surgical errors are strongly related to number of WD. Teamwork has been shown as an essential factor to reduce WD. Research in other high-risk settings such as in aviation field highlighted the fact that limited team familiarity (TF) between teammates can lead to negative safety outcomes. The aim of this exploratory study was to examine the relationships between expertise, TF and WD with scrub nurses (SN) and neurosurgeons through an observation of elective neurosurgical procedures carried out by surgeons and SN with different levels of expertise and different degrees of familiarity with each other.

Mardi 23 janvier 2018 13:00-14:00

"Skill based assistance in robotic surgery."

Présenté par : Atsushi Nakazawa

Robotic surgery has gained increasing popularity in recent years. Intelligent robotic surgical assistance in surgical sub-tasks such as suturing has the potential to enhance surgeon's performance and reduce fatigue and operating time. Previous works in surgical automation use a replication of demonstrated motions or a planned path, but they are not necessarily skilled motions. Therefore, we aim to develop an assistance system in robotic surgery based on skill analysis. As a first step, we are working on environment recognition to get better understanding of the skills.

Mardi 9 janvier 2018 13:00-14:00

"Mental Representation during Preoperative Surgical Planning."

Présenté par : Pierre-Louis Henaux

Expert judgment is vital for safe surgery.
 Judgment is difficult to define. A critical aspect of it is the ability of accurately mentally represent the purpose of surgical procedure.
 Mental Representation (MR) determines the details of this procedure, plans what decisions would be made, as well as what approaches would be taken and whether a procedure fits within a surgeon confort zone.
 Therefore, MR is critical to our understanding of surgical expertise. There is a body of literature about decision making, surgical intuition, clinical reasoning, problems solving. However, we should understand the essence of all these non-technical skills. MR is a process which underpins all these cognitive skills. Through a qualitative study following a constructivist grounded theory methodology, the aim of this study was to explore how does the surgeon mentally represent the operative procedure preoperatively and how this MR changes over time with development of surgical expertise.

Mardi 19 décembre 2017 13:00-14:00

"Development of an intelligent surgical training system for Thoracentesis."

Présenté par : Hiren Nakawala

Surgical training improves patient care, helps to reduce surgical risks, increases surgeon’s confidence, and thus enhances overall patient safety. Current surgical training systems are more focused on developing technical skills, e.g. dexterity, of the surgeons while lacking the aspects of context-awareness and intra-operative real-time guidance. Context-aware intelligent training systems interpret the current surgical situation and help surgeons to train on surgical tasks. first part of the talk will introduce context-aware system framework which helps to identify surgical context e.g. instruments to aid in surgical training. The second part will focus on the deep learning for context awareness, in which context-aware system framework ensure robust perception capabilities to intepret surgical processes. The applications to be presented are in the area of percutaneous surgery and robot-assisted kidney surgery.

Mardi 5 décembre 2017 13:00-13:30

"Presentation of my thesis and the state of progress."

Présenté par : Gurvan Lecuyer

Presentation of my thesis subject: “Methods to facilitate the learning steps of recognition algorithm deep learning-based: application to surgery videos”. I will talk about the basis of deep learning and then I will explain the work I made since I began, finally I will detail what I'm going to do for the 2 remaining years

Mardi 5 décembre 2017 13:30-14:00

"Presentation and administrative points about ordering and mission procedures."

Présenté par : Noémie Buisard

Noémie Buisard will quickly present her role in the lab and address some administrative points about ordering and mission procedures. These points will only concern the MediCIS team.

Lundi 13 novembre 2017 13:00-14:00

"Knowledge-based support for surgical workflow analysis and recognition."

Présenté par : Olga Dergachyova

Computer assistance became indispensable part of modern surgical procedures. Desire of creating new generation of intelligent operating rooms incited researchers to explore problems of automatic perception and understanding of surgical situations. Situation awareness includes automatic recognition of surgical workflow. A great progress was achieved in recognition of surgical phases and gestures. Yet, there is still a blank between these two granularity levels in the hierarchy of surgical process. Very few research is focused on surgical activities carrying important semantic information vital for situation understanding. Two important factors impede the progress. First, automatic recognition and prediction of surgical activities is a highly challenging task due to short duration of activities, their great number and a very complex workflow with multitude of possible execution and sequencing ways. Secondly, very limited amount of clinical data provides not enough information for successful learning and accurate recognition.

In our opinion, before recognizing surgical activities a careful analysis of elements that compose activity is necessary in order to chose right signals and sensors that will facilitate recognition. We used a deep learning approach to assess the impact of different semantic elements of activity on its recognition. Through an in-depth study we determined a minimal set of elements sufficient for an accurate recognition. Information about operated anatomical structure and surgical instrument was shown to be the most important. We also addressed the problem of data deficiency proposing methods for transfer of knowledge from other domains or surgeries. The methods of word embedding and transfer learning were proposed. They demonstrated their effectiveness on the task of next activity prediction offering 22% increase in accuracy. In addition, pertinent observations about the surgical practice were made during the study. In this work, we also addressed the problem of insufficient and improper validation of recognition methods. We proposed new validation metrics and approaches for assessing the performance that connect methods to targeted applications and better characterize capacities of themethod. The work described in this these aims at clearing obstacles blocking the progress of the domain and proposes a new perspective on the problem of surgical workflow recognition.

Mardi 7 novembre 2017 13:00-14:00

"A Semantic Repository for Medical Data Storage"

Présenté par : Marine Brenet

For a European research project about radioactivity exposure we need to store and manage images and datas from medical examination from dozens of laboratories. My task is to build a semantic repository for store all datas and allow powerful querying for access to datas with different filters and user friendly for all scientists involved in the project. This semantic repository could be reused for others projects (Biobank, medical datas storage..).

I will introduce myself, how I am building the semantic repository and briefly how I will build a user friendly request software.

Mercredi 18 octobre 2017 14:00-15:00

"Surgical Process Modeling using Hierarchical Task Representations"

Présenté par : Prs. Roy Eagleson and Sandrine De Ribeaupierre from Western University London Ontario

Mardi 10 octobre 2017 13:00-14:00

"Generic Surgical Process Model Derivation using Test and Flip Net Synthesis"

Présenté par : Aurélien Lamercerie

There is a constant demand for computer tools to assist the management and conduct of surgical operation. Formal modeling of surgical practices is needed to facilitate the development of these tools. This justifies the need for surgical process model (SPM), based on semantics. One of the difficulties is to construct a generic model, which is not just the composition of somes instances.

Our proposal is an innovative approach to deriving a generic model (gSPM) from a somes instances of surgical process model (iSPM). This approach uses a net synthesis method based on the theory of regions. This technique generates a Test 'n Flip net and makes it possible to generalize the variability found in the instances by adding transitions absent from the original data.

We have experimented with our proposal on some cases of study with real data. These experiments validated the approach: we arrive at constructing a generic model with interesting levels of generalization. In addition, this work is part of the S3PM / Sunset project, which aims to develop a virtual reality training environment for the surgical process. The virtual reality environment requires a scenario corresponding to the model of the targeted surgical process. This expected scenario can be obtained by adapting the synthesis of the Test 'n Flip net and thus automatically constructing the scenario from somes examples of the target process.

Mardi 6 juin 2017 13:00-14:00

"Automatic Technical Surgical Skill Scoring from Motion Data"

Présenté par : Chantal Julliard

Mardi 16 mai 2017 13:00-14:00

"A Mixed-Reality Surgical Trainer of Laser Fetal Minimally Invasive Surgery for Evaluating Novel Technology"

Présenté par : Allan Javaux, de l'Université de Leuven (Bruxelles, Belgique)

Fetal surgery is considered one of the most challenging disciplines within Minimally Invasive Surgery (MIS). Indeed, fragility of the anatomic features, the poor visibility, the limited maneuverability and the extreme requirements in terms of instrument handling and positioning precision contribute to this factor. In order to assist the surgeons for a specific laser surgery treatment Twin-Twin Transfusion Syndrome (TTTS), our team is developing (would remove smart) flexible instruments and dedicated techniques for robotic assistance. To evaluate these novel technologies in an early stage of the design, a mixed-reality surgical simulator was developed. Before making too strong conclusions about the technology under evaluation, the simulator must first be validated. Validity based on qualitative data, such as Face and Content Validity, has already been confirmed. However, the validation based on quantitative data, specifically instrument motions, is still undergoing. Through collaboration with Rennes we hope to come up with a mature and sound manner to accomplish a validation based on quantitative data.

Mardi 6 avril 2017 14:00-15:00

"Finding the needle in the haystack: detection, tracking and registration in biomedical imaging"

Présenté par : Raphael Sznitman, head of the Ophthalmic Technology Laboratory (ARTORG Center, University of Bern, Switzerland)

From centimeter-sized observations visible in endoscopy to nanometer large intra-cellular structures discernible with Electron Microscopes, searching and locating objects of interest in images is a central problem in medical image computing. If anything, the need for efficient object detection techniques has never been higher due to the advent of cheaper and ever more sophisticated imaging devices, able of amassing unprecedented quantities of data. And while established search paradigms are showing their limits, faster methods capable of dealing with larger quantities of data are now indispensable. In this talk, I will present a computational framework for object detection with efficiency at its core. This framework, a Bayesian formulation of the traditional “twenty questions” game, considers a sequential strategy for evaluating different parts of the image data in order to locate the object effectively. In this context, we will see how dynamic programming and information theory can be used to characterize a provably-optimal search strategy that is both simple to compute and greedy in nature. Using these results, I will then show how this framework can be used to solve traditional object detection and tracking problems, as well as non-rigid registration of multimodal multi-scale image data, allowing for more accurate solutions and large problems to be tackled.

Mardi 28 mars 2017 13:00-14:00

"Overview over image guided interventions and related applications in Trondheim"

Présenté par : Ingerid Reinertsen, senior research scientist from SINTEF (Trondheim, Norway)

The presentation will focus on image guided interventions in the context of neurosurgery, laparascopic surgery, bronchoscopy and anaesthesiology. Related topics are software development, image processing and machine learning applications.

Mardi 14 mars 2017 13:00-14:00

"Ingénierie des Ontologies Formelles: Les Vices et les Vertus"

Présenté par : Bernard Gibaud

La présentation originale a été faite par Stefan Schulz (Graz, Autriche) durant les Journées Françaises sur les ontologies octobre dernier. Lien vers la présentation : Lien vers le site de l'auteur :

Mardi 7 mars 2017 13:00-14:00

"Les enjeux de l'intelligence artificielle"

Discussion entre les membres de l'équipe

Mardi 28 février 2017 13:00-14:00

"Recognition of Low-Level Surgical Activities is Possible: a Smart Choice of Sensors is the Key!"

Présenté par : Olga Dergachyova

The present work aims at discovering which sensors and signals would facilitate recognition of low-level surgical activities, usually represented by the following elements: verb describing a performed action, surgical instrument and operated anatomical structure. We hypothesized that activity recognition does not require sensors for all three elements. We conducted a large-scale study using deep learning on semantic data of 154 interventions from four different surgical procedures. The results showed that the instrument and the verb encode similar information. Thus, only one is needed to be tracked, preferably the instrument. The anatomical structure provides some unique cues. Thereby, a sensor capturing it is indispensable. For all studied procedures, a combination of two elements, one being the structure, is sufficient to confidently recognize activity. To the best of our knowledge, this is the first study examining the importance of input signals and information to the process of low-level surgical activity recognition.

Mardi 10 janvier 2017 13:00-14:00

"Semantic Surgical Workflows"

Présenté par : Javier Rojas Balderrama, PhD

Jeudi 15 décembre 2016 13:00-14:00

"Détection automatique de déviations chirurgicales et identification de comportements chirurgicaux par modélisation et analyse des processus chirurgicaux"

Présenté par : Arnaud Huaulmé

Les événements indésirables (EIs) sont devenus une vraie préoccupation du monde médical, leur réduction étant recherchée pour assurer la meilleure sécurité possible pour les patients. Les événements indésirables sont, selon la HAS, ‘‘des situations qui s'écartent de procédures ou de résultats escomptés dans une situation habituelle et qui sont ou qui seraient potentiellement sources de dommages’’. Alors que les événements indésirables postopératoires sont étudiés depuis de nombreuses années, ceux ayant lieu au cours des opérations ne le sont que depuis récemment, comme le montre la récente classification des événements indésirables intraopératoires par Kaafarani et al. publiée en 2014. Cependant, la classification d'événements indésirables intraopératoires n'est que la première étape pour comprendre les comportements chirurgicaux qui les entraînent.

Dans cette thèse, nous présenterons des méthodes pour détecter l'apparition de déviations dues à l'apparition d'événements indésirables intraopératoires et pour identifier des comportements chirurgicaux à partir de modèle de processus chirurgicaux.

Ce travail a nécessité de concevoir et développer une modélisation formelle de la rectopexie et des événements indésirables qui sont associés à cette procédure chirurgicale grâce à la mise en place d'ontologies. Cette modélisation formelle nous a permis de bien appréhender le principe de cette opération et de fournir un vocabulaire permettant une annotation détaillée de vidéos endoscopiques de rectopexies, afin de créer des modèles de processus chirurgicaux en jeu.

Grâce à l'annotation des vidéos chirurgicales basée sur cette modélisation, nous avons développé une méthode de détection automatique des déviations dues à l'apparition d'événements indésirables. Cette méthode est basée sur un alignement temporel non linéaire multidimensionnel, que nous avons développé, suivi d'un modèle semi-Markovien caché que nous avons entraîné pour déterminer s'il existe des déviations par rapport à une chirurgie de référence et si celles-ci sont dues à des événements indésirables.

Cette détection de déviations dues aux événements indésirables est la première étape afin de comprendre les raisons de leurs apparitions. Nous émettons l'hypothèse que leurs apparitions peuvent être expliquées par une succession d’activités, c'est-à-dire un pattern. Pour répondre à cette hypothèse, nous avons mis en place une méthode de découverte de patterns permettant d'identifier les comportements chirurgicaux spécifiques à différents critères. Cette identification de comportements chirurgicaux est réalisée par une classification ascendante hiérarchique avec la mise en place d'une nouvelle métrique basée sur les patterns partagés entre les chirurgies. Afin de valider notre méthode, nous l'avons comparé à deux études mettant en évidence des différences de comportements chirurgicaux, par exemple entre différents sites chirurgicaux ou entre deux types de procédures de la même opération. Une fois la méthode validée, nous avons utilisé notre méthode afin de montrer s'il existait des comportements chirurgicaux spécifiques à des données préopératoires et à l'apparition d'événements indésirables.

Mardi 6 décembre 2016 13:00-14:00

"Modelling and learning surgical gestures for man-machine collaborative training"

Présenté par : Chantal Julliard

Mardi 29 novembre 2016 13:00-14:00

"UHF MRI for stereotactic surgery"

Présenté par : Jonathan C. Lau, MD, Neurosurgery Resident at Western University (London, Canada)

We are pleased to receive Jonathan C. Lau, MD visiting Rennes. Having a background in Computer Science and Biomedical Engineering, Jonathan is an active collaborator with the VASST laboratory of Terry Peters. His current research goals are to develop and validate image processing software for neurosurgical applications to facilitate improved characterization of neurosurgical illnesses using computational modeling techniques and the development of a framework for improved surgical planning and intra-operative neuronavigation.

Vendredi 17 novembre 2016 15:00-14:00

"Measuring the acceptance of S3PM by scrub nurses"

Présenté par : Marie-Stéphanie Bracq

Marie-Stéphanie Bracq is a newcomer PhD student preparing a thesis in collaboration of CRPCC (Psychology Research Center, Cognition and Communication, University of Rennes 2) and MediCIS team, co-directed by Pierre Jannin and Estelle Michinov. Her thesis is about the assessment of scrub nurses non-technical skills in a collaborative learning virtual environment. Her PhD work lies in the context of SUNSET project (Scrub Nurse Non-Technical Skills Training System) which is the continuation of S3PM project (Synthesis and Simulation of Surgical Process Models). The objective of her current work is to measure acceptance of S3PM virtual environment by surgical scrub nurses before deploying it for training. Marie-Stéphanie will make a presentation introducing her thesis. The title of her talk is “Measuring the acceptance of S3PM by scrub nurses”. The idea of measuring the acceptance of the proposed training technology is to be able to predict its use intentions and understand possible brakes stopping from its application. She will also present 2 models: 1) TAM: Technology Acceptance Model by Dabis (1989), and 2) UTAUT: Unified Theory of Acceptance and Use of Technology by Vankatesh (2003, 2012), as well as inqueries used for expiriments with S3PM.

Mardi 15 novembre 2016 13:00-14:00

"Intelligent surgical robotic system based on the multidisciplinary computational anatomy"

Présenté par : Professeur Makoto Hashizume (Université de Kyushu, Japon)

Nous avons le plaisir d'accueillir le Professeur Hashizume de l'Université de Kyushu. Il nous présentera la partie robotique d'un énorme projet national qui regroupe toutes les principales équipes de recherche japonaises de notre domaine, projet dont il est le porteur.

Mardi 8 novembre 2016 13:00-14:00

"Angiogram Image Processing for Catheter Guidance in Thoracoabdominal Region"

Présenté par : Professor Hideaki Haneishi (Center for Frontier Medical Engineering, Chiba University, Japan)

Interventional radiology (IVR) with catheter is a useful, less invasive procedure for diagnosis and treatment of diseases in thoracoabdominal organs such as liver. However, such organs are always moving due to respiration, which makes the catheter operation difficult. Patients breath holding is a solution but some patients including elder people sometimes cannot hold the breath. Image processing for visualizing high contrast vessel pattern without breath holding is highly desired in this field. Digital subtract angiography (DSA) is one of imaging methods using X-ray image and is commonly conducted to clearly visualize the vessels during intervention with a catheter. But DSA requires breath-holding. Thus, non-breath-hold DSA is very attractive.

Two methods for angiogram image processing are presented for catheter guidance in thoracoabdominal region. One is roadmap generation from angiogram motion picture. The other is respiration-synchronized real-time generation of angiogram from fluoroscopy images. Both methods utilize fluoroscopy and/or angiography motion pictures and generate blood vessel enhanced images to assist physician’s catheter control and guidance of catheter feeding. The first method was tested by 53 patient data and the second method was tested 12 patient data. Through the experiments with patient data, the effectiveness of the proposed methods was confirmed.

We present two methods of angiogram generation. One is roadmap generation from angiogram motion picture. The other is respiration-synchronized real-time generation of angiogram from fluoroscopy images. In terms of the first method, we use consecutive angiographic images including respiratory motion and heartbeat. The method consists of blood vessel enhancement and artifact removal. Robust principal component analysis (RPCA) is used to enhance blood vessel information from consecutive angiographic images. RPCA separates the consecutive images into a low-rank component and a sparse component. The information of contrast media is included into the sparse component. We applied the proposed method to 53 sets of angiographic images and confirmed that the proposed method using RPCA more effectively enhanced the blood vessels than the conventional subtraction method. In terms of the second method, a respiratory synchronized DSA generation based on a respiratory phase matching is performed. Pre-operative angiographic images under natural respiration and intra-operative fluoroscopic image under natural respiration during the operation are used. For each fluoroscopic image, the proposed method chooses an angiographic image in the most similar respiratory phase by a pattern matching technique in real time. We conducted two kinds of experiments to confirm that the proposed method can be used as the free breathing DSA and the respiratory synchronized roadmap. The result showed that the proposed method can produce fine DSA without breath-holding and the processing can be done in real-time.

Respiratory phase matching method was proposed for generation of free breathing DSA. Acceptable fine DSA could be obtained without breath-holding by using proposed method. We also have proposed a blood vessel enhancement method using consecutive angiographic images with respiratory motion and heartbeat. The proposed method is composed of robust principal component analysis and a deformation technique to extract blood vessel pattern from angiographic images. We applied the proposed method to 53 patient data. In most cases we confirmed that artifacts by difference between live images and mask images were reduced by the proposed method.

As future works for the first method, we will evaluate the clinical value. As future works for the second method we will apply the method to clinical data more and more and obtained feedback from physicians side so that the method will be improved for clinical practice.

Jeudi 17 novembre 2016 14:00-16:00

"Pré-soutenance de thèse : Imagerie fonctionnelle intraopératoire naviguée pour l'optimisation de la chirurgie des tumeurs cérébrales"

Présenté par : Frederic Monge

Les gliomes sont des tumeurs cérébrales primitives représentant le deuxième cancer le plus fréquent chez l’enfant et la troisième cause de mortalité chez l’adulte jeune. Il a été démontré qu’une chirurgie d’exérèse tumorale permet d’augmenter la qualité et la durée de vie du patient, voir même dans certains cas, d’obtenir sa guérison. L’identification intraopératoire des résidus tumoraux permettrait au chirurgien de vérifier, in situ, la qualité de son geste d’exérèse. Une nouvelle modalité d’imagerie intraopératoire a été proposée comme solution pour détecter les tumeurs résiduelles. Elle se base sur l’utilisation d’une sonde nucléaire combinée à un système de localisation optique. Cette nouvelle modalité, appelée imagerie surfacique de positons (ISP), permet la génération d’images de la distribution surfacique d’un radiotraceur comme le 18F d’une zone d’intérêt scannée. L’ISP n’étant actuellement pas utilisée en clinique, nous proposons pour la première fois une étude de faisabilité de son utilisation pour l’optimisation de la chirurgie des tumeurs cérébrales. Nous montrons les limites de l’utilisation potentielle de l’ISP dans un contexte neurochirurgical par des études expérimentales en considérant des facteurs intraopératoires pouvant influencer la qualité des images générées par le système. Les contributions présentées se concentrent sur trois axes. Dans un premier temps, nous souhaitons obtenir des images fonctionnelles d’ISP avec un temps de calcul faible. L’application de modèles d’acquisition aux mesures permettent d’améliorer la qualité des images, au détriment d’un temps de calcul élevé. Nous suggérons un nouveau modèle d’acquisition dédié au contexte intraopératoire, permettant l’amélioration de la résolution et du contraste des images pour un temps de calcul réduit. Un deuxième axe est dédié à l’étude de l’impact d’une acquisition intraopératoire sur les performances du système. Nous proposons l’estimation des paramètres intrinsèques de la sonde nucléaire utilisée et de l’influence de la hauteur du processus de scan sur leurs valeurs. Le dernier axe est consacré à la validation de modèles d’acquisition dans le contexte intraopératoire. Nous présentons une étude comparative des performances de modèles d’acquisition en considérant l’impact potentiel de la vitesse de scan durant l’acquisition. L’ensemble de ces travaux a contribué à l’étude de faisabilité de l’utilisation d’un système d’ISP en intraopératoire, proposée pour la première fois en neurochirurgie.

Mardi 25 octobre 2016 13:00-14:00

"Surgical Vision Augmentation in da Vinci Robotic Prostatectomy"

Présenté par : Xiongbiao Luo

Da Vinci robotic prostatectomy that is increasingly performed for minimally invasive prostatic tumor resection uses stereoscopic laparoscopes to intuitively visualize the organ surface in the body and manipulate various surgical tools under the insufficient light sources and the relatively narrow field of view of the binocular stereoscopic laparoscope that usually limits the visualization on the organ being operated on and its anatomical surroundings. Surgical vision augmentation in accordance with non-uniform illumination correction, fast visibility-contrast fusion defogging, motion magnification, and surgical filed 3D reconstruction techniques to maintain and augment a clear field of in-situ direct vision in robotic prostatectomy is paramount not only for safety by preventing inadvertent injury, but also to improve precision and reduce operative time.

Mardi 27 septembre 2016 13:00-14:00

"Séquencement d'actions en environnement virtuel collaboratif"

Présenté par : Guillaume Claude

Un environnement virtuel interactif à lui seul ne permet pas de répondre à tous les besoins liés à son utilisation. Dans le cadre de la formation par Réalité Virtuelle, un cadrage de l’activité de l’apprenant est important pour offrir une formation efficace. Le problème est le même dans les fictions interactives : seul, l’environnement n’offre qu’un intérêt limité. Nous nous intéressons ici au problème de la spécification du séquencement des actions dans un environnement collaboratif. Il s’agit de définir puis de contrôler ce qui peut ou doit se passer au cours de la simulation dans un contexte potentiellement multiutilisateur. Ceci passe, entre autres, par la spécification (puis l’exécution) d’un ensemble de scénarios possibles lors d’une session de simulation ainsi que par la distribution des actions réalisables entre les différents acteurs (réels ou virtuels) intervenant dans la simulation.

Cette présentation reprend les travaux de thèse de Guillaume, réalisée dans le cadre du projet de rechercheS3PM du labex Cominlabs.

Mardi 20 septembre 2016 13:00-14:00

"Integration of Medical Ontologies into the Framework of a Top-level Ontology"

Présenté par : Chantal Julliard

Chantal did her studies in Mechanical Engineering at the Karlsruhe Institute of Technologies in Germany focusing on Biomedical Engineering. In her Master’s thesis she worked on medical ontologies and during her PhD she will work on Modelling Surgical Gestures for a Man-Machine Collaborative Training. Within her presentation she will share her master project and explain why and how to design a good ontology, particularly for laparoscopic surgeries.

Mardi 13 septembre 2016 13:00-14:00

"Deep learning: Covolutional Neural Networks"

Présenté par : Olga Dergachyova

Le deep learning (apprentissage profond en français) est un ensemble de méthodes d'apprentissage automatiques tentant de modéliser avec un haut niveau d’abstraction des données grâce à un graph profond avec multiple couches de traitement composées de nombreuses transformations linéaires et non-linéaires. Ces tecniques ont démontré qu'elles étaient capables de produire d'excellents résultats pour différentes problématiques dans des champs d'application tels que la vision par ordinateur, la reconnaissance automatique de la parole et du langage naturel.

Dans ma présentation je vais me concentrer sur le réseau neuronal convolutif (CNN), le type de réseau de neurones le plus utilisé actuellement. Je vais introduire le concept général, parler de la structure du réseau et de son apprentissage. Ensuite je présenterai les problèmatiques et les directions de recherche les plus vives dans ce domaine, ainsi que des exemples d'application.

Mardi 05 juillet 2016 13:00-14:00

"Intepersonal non technical skills in neurosurgery"

Présenté par : Pierre-Louis Hénaux

The talk will consist of three parts:

  • Presentation of “PACTE project (evaluation of teamworking in emergency operative room)”
  • Quantitative evaluation of non technical skills in neurosurgery
  • Presentation of “Influence of team familiarity in Neurosurgical OR”

Mardi 24 mai 2016 13:00-14:00

"Représentation sémantique des biomarqueurs d'imagerie"

Présenté par : Emna Amdouni

Le National Institute of Health a souligné le besoin de définir un vocabulaire lié au domaine des biomarqueurs d’imagerie afin de lever les ambiguïtés qui affectent ce concept. Plusieurs initiatives en ce sens ont émané de la part de diverses communautés (recherche en imagerie, radiologie, génétique, gestion des connaissances, etc.). Toutefois, jusqu’à aujourd’hui, aucun de ces travaux n’a réussi à répondre parfaitement à cette question. Notre travail vise à proposer une nouvelle façon d’aborder ce problème de représentation des biomarqueurs d’imagerie fondée sur les technologies du web sémantique 3.0 et consistant à définir explicitement les entités du domaine à couvrir grâce à des ontologies. Dans notre travail, nous proposons une ontologie générique intitulée «Imaging Biomarker Ontology» (IBO) qui définit les trois principaux aspects liés au concept de biomarqueur d’imagerie et qui sont: la qualité biologique mesurée, l’instrument de mesure et l’outil d’aide à la décision. Cette ontologie est une extension d’autres ontologies provenant de l’«Open Biological and Biomedical Ontologies Foundry». Elle a été, également, utilisée pour représenter des biomarqueurs d’imagerie (quantitatifs et qualitatifs) employés dans le domaine des gliomes cérébraux. Actuellement, nous travaillons sur la représentation sémantique des structures pathologiques et anatomiques qui sont présentes dans les IRMs cérébraux. Pour ce faire, nous avons choisi comme référence la terminologie VASARI qui fait intervenir des scores qui décrivent des qualités qualitatives et quantitatives.

Mardi 10 mai 2016 13:00-14:00

"Reconnaissance de gestes pour une visualisation optimale d’injections naviguées"

Présenté par : Sébastien Muller

Le but est l’identification de changements de phases chirurgicales lors d’interventions simples avec un nouvel ustensile (Multiguide) pour guider des injections et/ou biopsies. Les actions qui en découlent sont simplement de mettre à jour le système de navigation avec l’information la plus pertinente pour le chirurgien.Des atributs ont été définis basés sur des interventions simulées sur fantôme. Leur pouvoir de classification a été testé sur 25 séries réparties sur 5 opérateurs (3 chirugiens, 2 ingénieurs ayant connaissance de l’intervention). Nous avons testé la validité interne et externe des classifications. Des résultat préliminaires seront présentés.

Mardi 25 avril 2016 13:00-14:00

"Pattern discovery for surgical skills analysis"

Présenté par : Arnaud Huaulmé

Mardi 29 mars 2016 13:00-14:00

"Objective evaluation and feedback for training in deep brain stimulation planning"

Présenté par : Matthew Holden

Develop of a virtual training course with an objective evaluation is of a high importance for improvement of planning skills of trainees in DBS. The software measures the performance of trainee based on two criteria : quality of trajectory and planning process, and provides an useful feedback to guide the trainee towards a good trajectory and acquire more experience in trajectory definition.

Mardi 15 mars 2016 13:00-14:00

"ADCAS: Automatic Detection in Computer-Assisted Surgery"

Présenté par : Olga Dergachyova

ADCAS is a platform developed within the MediCIS team for detection of surgical workflow. The software is coded in C++ with use of Python scripts. Its pipeline includes 4 applications: data formatting, training, recognition and evaluation. The data formatting serves to process and transform the signals acquired in the OR in a appropriate and convenient format for the further use. It also creates two data sets: for training and for tests. The training application extracts features from input data in order to describe it and then learns a model of a chosen classifier to be able to distinguish surgical processes. The recognition application, the heart of the pipeline, predictes the workflow label of each test sample. The evaluation application allows to asses the performancies of the trained model and the choosen strategy. The whole platforme is highly modular based on a plugin system.

Mardi 23 février 2016 13:00-14:00

"Advancing personalized treatments"

Présenté par : Valeria De Luca

Valeria is Postdoc researcher in Computer Vision Laboratory and ETH Zurich (Switzerland) working on machine learning in neurointensive care. The presented us her two projects: one made during her PhD thesis and one she is actually managing. The first project on liver ultrasound proposed generalized solutions for cancer treatments. The second project is for neurointensive care (ICU Cockpit) solution to specific clinical questions.

Mardi 9 février 2016 13:00-14:00

"Reconnaissance de gestes pour une visualisation optimale d'injections naviguées"

Présenté par : Sébastien Muller

Sébastien Muller vient du département de technologie médicale à la SINTEF et de l'institut de neuromédecine de l'université technologique NTNU à Trondheim, Norvège. Il est à Rennes afin de travailler sur de la modélisation de processus chirurgicaux dans le cadre d'une intervention appliquant un nouvel outil chirurgical pour guider des injections et biopsies. Le travail portera sur de la reconnaissance de gestes pour reconnaître les phases de l'intervention et mettre à jour le système de navigation avec l'information la plus pertinente pour le chirurgien. Durant son discours il nous a présenté le système de navigation CustusX et un outil développé à la SINTEF destiné à l’injection du toxine botulique dans un ganglion cervicale responsable de migraines fréquentes. Le défi est de reconnaître automatiquement les étapes principales de l’intervention afin de pouvoir mettre à jour la visualisation du navigateur selon les gestes du chirurgien.

Mardi 2 février 2016 13:00-14:00

"Introduction to formal language"

Présenté par : Yulong Zhao

Formal language theory is concerned with the specification and manipulation of sets of strings of symbols, i.e., languages. Formal language theory is the heart of modern compiler architectures, regular expressions, parsers, web-scrappers, natural language processing (NLP), state machines based on markov chains, pattern matching.

Mardi 26 janvier 2016 13:00-14:00

"Real-Time Workflow Analysis for Automatic Guidance in Image-Guided Needle Intervention Training"

Présenté par : Matthew Holden

To augment expert supervision during medical intervention training, the use of computer-assisted training systems has been proposed. To be an asset in the training process, such systems should provide feedback to the trainee in the form of either skill evaluation or context-sensitive instruction. In this work, we address a key challenge in providing context-sensitive instruction during an intervention: identifying which task in the intervention's workflow the operator is performing. To this end, we have developed an intervention-independent algorithm which determines the ongoing task in real-time for interventions with arbitrary task ordering and repetition. The algorithm has been validated on two intervention datasets: freehand lumbar puncture and ultrasound-guided epidural. Relative to expert observer consistency, the proposed algorithm exhibits a mean accuracy of 93%. We suggest that this algorithm has sufficient task identification accuracy for use within a system which provides automated instruction to augment expert supervision in the training process. The proposed algorithm is implemented within the open-source Perk Tutor training platform ( for image-guided intervention training.

Jeudi 17 décembre 2015 13:00-14:00

"Sequence Analysis for Surgical Process Modeling and Mining"

Présenté par : Germain Forestier

The goal of Surgical Process Modeling (SPM) is to support surgery through the quantitative analysis and understanding of operating room activities. This field is in constant growth with the increasing amount of data acquired in operating rooms. In this talk, I will give an overview of the work we carried out in the past few years on using sequence analysis to exact knowledge from recordings of surgical activities. I will highlight how we used and adapted Dynamic Time Warping (DTW) algorithm to automatically identify and understand complex surgical behaviors. I will also present our recent work focusing on the prediction of surgical activities and surgical phases. Finally, I will briefly discuss my other research interests including ontology engineering and pathology image analysis using deep learning.

Jeudi 3 décembre 2015 13:00-14:00

"SurgeTrack, SurgePlan et SurgeLive : une suite logicielle basée SPM"

Présenté par : Guillaume Dardenne

La modélisation des procédures chirurgicales (Surgical Process Modelling, SPM) ouvre des perspectives extrêmement intéressantes tant sur la formation des chirurgiens, que sur l'assistance optimisée au bloc opératoire, ou bien encore la mesure d'indicateurs qualité pertinents associés à une procédure chirurgicale. Outre ces perspectives, l'utilisation de ces modèles en routine clinique ou pour la formation est encore timide. Ceci peut s'expliquer par la non-standardisation de ces modèles et le manque d'outils adaptés qui répondent réellement aux exigences et aux contraintes du milieu chirurgical. Il existe par conséquent tout un travail de recherche à mener sur la standardisation de ces modèles mais également sur le développement d'outils innovants utilisables dans le contexte chirurgical et s'appuyant sur des modèles à priori de chirurgie. Nous allons ici vous présenter la suite logicielle SurgeSuite développée au sein de l'IRT b<>com en collaboration avec l'équipe MEDICIS. Cette suite est composée des logiciels : SurgeTrack pour l'enregistrement et l'analyse de modèles chirurgicaux ; SurgePlan pour la préparation d'une chirurgie ; et enfin, SurgeLive pour l'assistance au bloc opératoire. Nous réaliserons une démonstration des logicielles SurgeTrack et SurgePlan en cours de test sur plusieurs sites ; et nous vous présenterons le concept associé au futur logiciel SurgeLive.

Jeudi 26 novembre 2015 13:00-14:00

"Knowledge support for multi-granularity and real-time modelling and recognition of surgical processes from multimodal sensors"

Présenté par : Olga Dergachyova

The operating room is a high-risk and dynamic environment. Thus, improved integration of high technology, along with enhancement of teamwork, and coordination between services are essential to improve patient safety and comfort, augment efficiency and reduce the cost of care. A new paradigm called Operating Room of the Future has been introduced into the clinical environment. It aims to extend the perception and action of surgical staff. The realization of the this concept requires a context-aware system which understands actions and events having place in OR and descover the surgical workflow. In this presentation we describe knowledge-based method for surgical wokrflow detection and propose novel metrcis adapted for assesment of systems performing wokflow analysis.

Mardi 10 novembre 2015 13:00-14:00

"Qualitative assessment of Non Technical Skills in neurosurgical Operating Room: a focus on cognitive skills."

Présenté par : Pierre-Louis Henaux.

Non Technical Skills (NTS) in surgery can be assessed by external rating scales (NOTAS, NOTSS, NOTECHS). These scales seem to be well adapted to assess the interpersonal skills but incomplete to fully evaluate cognitive skills. Cognitive skills are indeed often non-verbalized and lack behavioral markers to decrypt them. Each surgeon has a different character and do not necessarily express orally all questions he may have during the intervention. Furthermore the emotional context and workload (mental readiness and workload distribution) of the surgeon may not be known by the team during the surgery. For example during an operation, the surgeon may have a family annoyance that will impact his mental readiness. It may be in a period of extreme fatigue or even himself having a disease (more or less severe) without be off work. Furthermore he may also have other professional events with deadlines to keep (teaching or research) forcing him to adapt his workload distribution and decreasing his mental readiness. Another example is interventions over a single day concatenation with a first intervention that goes wrong and a patient who has an adverse event (vital event or serious functional impairment) while the surgeon is performing a second surgery. All these factors can not be accurately described with simple generic scales. So it seems essential, for having a complete analysis of the NTS, to include a qualitative analysis of interviews with the surgeon after his intervention. A constructivist grounded theory study is an alternative to accurately analyze all these cognitive skills involved during the procedure.

Mardi 13 octobre 2015 13:00-14:00

"Image-Guided Intervention Training Using the Open-Source Perk Tutor Platform."

Présenté par : Matthew Holden.

With the recent shift of medical education to a competency-based model, the use of objective measures of skill and automatic feedback systems has become increasingly important. This presentation will outline the Perk Tutor platform for image-guided intervention training ( The Perk Tutor is a free, open-source resource designed to facilitate medical education research and the development of image-guided intervention training curricula. The platform has been validated for both performance evaluation and automatic workflow analysis in a variety of applications, including spinal, central line, and prostate interventions.

Mardi 29 septembre 2015 13:00-14:00

"Detection of deviation in rectopexy surgery thanks to NLTS and HsMM."

Présenté par : Arnaud Huaulmé.

Une opération chirurgicale ne se déroule pas toujours comme escompté. Celle-ci peut être perturbée par les spécificités du patient qui viennent modifier le planning, mais aussi par des événements indésirables pouvant avoir de lourdes conséquences. A partir des modèles individuels des procédures de rectopexie (iSPM), nous essayerons de détecter ces éléments en combinant l'utilisation d'une méthode d'alignement de séquence non-linéaire (NLTS) et des modèles semi-markovien cachés (HsMM).

Mardi 22 septembre 2015 13:00-14:00

"Chemistry-inspired Adaptive Stream Processing".

Présenté par : Javier Rojas Balderrama.


Mardi 30 juin 2015 13:00-14:00

"Gate simulation toolkit: Overview".

Présenté par : Frédéric Monge.

Geant4 is a toolkit for the simulation of the passage of particles through matter. Its areas of application include high energy, nuclear and accelerator physics, as well as studies in medical and space science.

GATE is an advanced opensource software developed by the international OpenGATE collaboration, based on Geant4 and dedicated to numerical simulations in medical imaging and radiotherapy. It currently supports simulations of Emission Tomography (Positron Emission Tomography - PET and Single Photon Emission Computed Tomography - SPECT), Computed Tomography (CT) and Radiotherapy experiments. Using an easy-to-learn macro mechanism to configurate simple or highly sophisticated experimental settings, GATE now plays a key role in the design of new medical imaging devices, in the optimization of acquisition protocols and in the development and assessment of image reconstruction algorithms and correction techniques. It can also be used for dose calculation in radiotherapy experiments.

Mardi 16 juin 2015 13:00-14:00

"Programmation GPU".

Présenté par : David Bouget, Ph.D.

David nous a exposé l'intérêt du passage du CPU au GPU pour les calculs parallèles notamment en comparant leurs structures et capacités de calcules. La base et le modèle de programmation ont été également présentés. Les configuration d'exécution, navigation entre les threads, accès au mémoire et fonctions kernel ont été examinés.

Jeudi 20 mars 2014 13:00-14:00

"Objective skill assessment and training method for laparoscopic surgery".

Présenté par : Munenori UEMURA Ph.D., As.P.E.Jp. Center for Advanced Medical Innovation Kyushu University - Kyushu University Hospital, Fukuoka, Japan.

Mardi 18 mars 2014 13:00-14:00

"De l'indéxation d'évènements dans des films -- Application à la détection de violence".

Présenté par : Cedric Penet, PhD.

Mardi 28 janvier 2014 13:00-14:00

"Image Guided Surgery".

Présenté par : Louis Collins, Prof. PhD. Montreal Neurological Institute - McGill University, Quebec, Canada.

Mardi 21 janvier 2014 13:00-14:00

Des connaissances procédurales à la commande de robot chirurgicaux en chirurgie laparoscopique.

Présenté par : Fabien Despinoy, Phd student LIRMM - Département Robotique, équipe DEXTER.

Mardi 14 janvier 2014 13:00-14:00

Acquisition, visualisation et reconstruction 3D de données anatomiques issues de dissection : application aux fibres blanches cérébrales.

Présenté par : Barthélemy Serres, Phd - Post doc UMRS U930, Inserm/Université de Tours - “Imagerie et cerveau”.

Mardi 10 décembre 2013 13:00-14:00

Improving intracranial EEG by optimizing depth electrode trajectory planning

Présenté par : Rina Zelmann, PhD The McConnell Brain Imaging Center Montreal Neurological Institute, McGill University, Quebec, Canada.

Mardi 5 novembre 2013 13:00-14:00

Activités sur le projet CREDIBLE (fédération de données d'imagerie).

Présenté par : Bernard Gibaud.

Mardi 22 octobre 2013 13:00-14:00

Démonstration de pyDBS: interface dédiée création d'atlas et parcours de bases de donnée"

Présenté par : Gregory Laheurte.

Mardi 8 octobre 2013 13:00-14:00

Real time surgical tool recognition: a journey in Saarbrücken.

Présenté par : David Bouget.

Mardi 1er octobre 2013 13:00-14:00

Présentation du laboratoire de l'université Kyūshū à Fukuoka, Japon & futures collaboration

Présenté par : Pierre Jannin.

Mardi 17 septembre 2013 13:00-14:00

Présentation du séjour de 3 mois à la Technische Universität München, Allemagne

Présenté par : Frédéric Monge.

Mardi 10 septembre 2013 13:00-14:00

Présentation oral de MASTER II SIBM (Signaux et Images en Biologie et Médecine)

Présenté par : Pierre-Jean Le Reste.

Mardi 21 mai 2013 13:00-14:00

Nelles approches de partage de données dans les PACS et le projet ISIMED (IRT B-COM)

Présenté par : Bernard Gibaud.

Mardi 7 mai 2013 13:00-14:00

Empreintes connectionnelles et annotations sémantiques

Présenté par : Tristan Moreau.

Mardi 23 avril 2013 13:00-14:00

Présenté par : David Bouget.

Mardi 9 avril 2013 13:00-14:00

Les fractales en neurosciences.

Présenté par : Pierre-Jean Le Reste.

Mardi 26 mars 2013 13:00-14:00

Présenté par : Gregory Laheurte.

Mardi 12 mars 2013 13:00-14:00


Présenté par : Bernard Gibaud.

Mardi 26 février 2013 13:00-14:00

Présenté par : Florent Lalys.

Mardi 05 février 2013 13:00-14:00

Modeling potential fields and tissue activation in DBS

Présenté par : Tiziano D'Albis

Abstract: A first step towards a better understanding of clinical outcomes and side-effects of Deep Brain Stimulation (DBS) consists in the estimation of the volume of neural tissue affected by the stimulation. Since this estimation is currently experimentally difficult, computational models have been proposed to shade light on this topic. The problem can be subdivided in two steps: 1) modeling of the potential field generated by DBS electrodes within the patient brain and 2) modeling of the neural response to this field. In this talk I review the current approaches found in literature for solving these two problems, with particular enphasis on the modeling of the potential field.

Mardi 22 janvier 2013 13:00-14:00


Mardi 08 janvier 2013 13:00-14:00

Présentation des thèses sur l'étude de l'utilisabilité de systèmes de chirurgie guidée par l'image et sur la stimulation cérébrale profonde

Présenté par : Pierre Jannin.

Mardi 11 décembre 2012 13:00-14:00

“Secrétariat MediCis - L'administratif et vous”
Présenté par : Noémie Buisard

Mardi 27 novembre 2012 13:00-14:00

Anatomical patient-specific model for Deep-Brain Stimulation

Présenté par : Tiziano D'Albis

Abstract: High frequency Deep Brain Stimulation (DBS) has been demonstrated as an effective minimally invasive surgical treatment for treating motor related diseases, e.g. Parkinson Disease, Essential Tremor, Tourette’s syndrome. The quality of clinical improvement, as well as the presence of side effects, strongly depend on the accuracy of electrode placement. Stimulation targets, however, are often difficult to localize from patient images and the choice of an optimal implantation trajectory is a time-demanding and tedious task. To support suregeons during DBS surgical planning we build patient-specific and generic models of anatomical and clinical data and automatically compute electrode trajectories from these models. In this talk I will give a general overview of our work on DBS surgical planning, focusing on the construction of an anatomical patient-specific model from multimodal images.


Mardi 13 novembre 2012 13:00-14:00

Sujet de thèse: Imagerie intra-opératoire naviguée pour l'optimisation de la chirurgie des tumeurs cérébrales.

Présenté par : Frederic Monge.

Mardi 30 octobre 2012 13:00-14:00

Quels outils en neuroimagerie ?

Présenté par : Tristan Moreau.

Mardi 16 octobre 2012 13:00-14:00

Etude exploratoire des interactions durant des opérations de Stimulation Cérébrale Profonde

Présenté par : Martin Ragot.

Collaboration and special events

Thursday 21/03/2013 14:00-15:00: Modèles numériques pour l'évaluation objective de la qualité d'images médicales

Présenté par : Lu ZHANG-GE, Doctorante, LISA - Université d'Angers

Monday 18/03/2013 14:00-15:00: Advanced MRI Sequences

Présenté par : Giulio Gambarota, Professeur, LTSI - Rennes

Monday 11/03/2013 12:00-13:00: Organisations hautement fiables et prise de décision erronée :Retour d’expérience des bâtiments de surface de la Marine Nationale française

Présenté par : Sophie Le Bris et Dominique Martin. Centre de Recherche en Économie et Management, Unité Mixte de Recherche du CNRS qui associe des chercheurs du CNRS, de l’Université de Caen Basse-Normandie, et de l’Université de Rennes 1.

Thursday 14/02/2013 14:00-15:00: Approches cognitives symboliques et connaissances/formes a priori pour l’imagerie microscopique haut débit et haut contenu. Perspectives à travers des projets franco-singapouriens.

Présenté par : Daniel RACOCEANU. CNRS et Université Pierre et Marie Curie, Université Nationale de Singapour, Image & Pervasive Access Lab – IPAL, UMI CNRS, Singapour.


L’intervention proposée est structurée en 3 parties :

1. Présentation de l’Unité Mixte Internationale IPAL, basée à Singapour. URL : 2. Exploration d’une lame virtuelle en utilisant une approche in vitro - in silico – in cognito. Application à la gradation du cancer du sein en histopathologie. Cadre : projet ANR TecSan MICO (Microscopie Cognitive) et plus récemment, le projet FUI FlexMIm URL : 3. Approches probabilistes avec a priori de formes, analyse 2D – synthèse 3D pour le suivi et la différentiation des cellules souches neuronales. Reconstruction 3D de neurites / neurones. Cadre : Système de microscopie intégré pour le suivi des cellules souche neuronales et pour les structures complexes 3D – Projets A*STAR (Agency for Science Technology and Research) URL :

Thursday 05/07/2012 13:00-14:00 : "5D Image Analysis in Cardiac Image Analysis"

Dr. Shuo Li, Image Research Scientist, Project Manager, GE Healthcare, Adjunct Research Professor, Dept. of Medical Biophysics, Schulich School of Medicine and Dentistry, Adjunct Research Professor, Dept. of Medical Imaging, Schulich School of Medicine and Dentistry, University of Western Ontario.


Five-dimensional (5D) image processing has become a major challenge within the field of medical imaging. 5D images, i.e dynamic 3D images (4D) that are acquired at multiple times. A typical application in cardiac image analysis would include dynamic cardiac CT scans and/or gated cardiac MRI acquired at 3-monthly intervals and dynamic cell growth and shape change in weekly intervals. 4D images provide research a great assistance on analysis of the deformation and motion. The pattern of this deformation and motion can be used for recognition and diagnosis when another dimension is added. Another dimension usually comes from atlas, template, historical data and etc. After integrated with another dimensional data, the got the 5D we talk about. The correlations in 5th dimension (e.g. the pattern of deformation for dynamic cardiac CT) can be extracted for recognition, tracking and diagnosis. During presentation, a group of experiments and validation will be presented to demonstrate the strength of 5D processing in cardiac image analysis.


Friday 22/06/2012 13:00-14:00 : "Medical Robotics at Waseda University"

Dr. Hiroyuki ISHII, Takanishi-Lab, Waseda University, Dr. Takeharu HOSHI, Fujie-Lab, Waseda University.


  • Fujie-Lab (Medical and Assistive Robotics Laboratory)

Fujie-Lab has been developing new medical robots which are controlled using biomechanical properties of human, such as material mechanics, dynamics, thermo dynamics and fluid dynamics. This means that we transform experimental knowledge of surgeons into biomechanical quantitative properties. We will highlight some of the work carried out by us and will present how these new approach may help to improve surgical treatments.

  • Takanishi-Lab (RT for Medical Education)

Takanishi-Lab has been conducting research on robotic patient simulator for physicians training. We will present a new robotic simulator for suture/ligature procedures. This simulator with embedded sensors measures applied force during suturing/ligaturing and captures condition of the stitches to evaluate the performance scores for the suture/ligature process. We also will present a new humanoid for training of airway management. This humanoid mimics different kinds of patient’s motor functions in the TMJ (temporomandibular joint) and neck, where several sensors are implemented in the tongue and teeth of the humanoid to measure the applied force by trainee.

Wednesday 02/05/2012 13:00-14:00 : "Computer Assisted Interventions: Challenges in design, development, validation and deployment of novel techniques"

Pr. Nassir Navab : chair for Computer Aided Medical Procedures & Augmented Reality Institut für Informatik I16, TU Munich, Germany.

inserm rennes1 ltsi