Skip to main content

Michel Dojat

Grenoble University, Grenoble Institut des Neurosciences, Faculty Member

Followers

49

Following

18

Co-authors

16

Public Views

Michel Dojat, Research Director at Inserm (Fr), is expert in neuroimaging, neuroinformatics, medical image analysis and medical informatics. His research is focused on the exploration of human vision using fMRI and development of methods for image processing and neuroscience data management. He is IEEE senior member, editorial board member of the Artificial Intelligence in Medicine journal and regularly serves on review of international journals (Neuroimage, TMI, …) and of the scientific committees of international conferences (AIM, MICCAI, ISBI, …).
Address: Grenoble Institut des Neurosciences
INSERM U1216
Bâtiment Edmond Safra
Chemin Fortuné Ferrini
38700 La Tronche

less

Interests

Uploads

Papers by Michel Dojat

Nodal statistics-based equivalence relation for graph collections

arXiv (Cornell University), Oct 3, 2022

Node role explainability in complex networks is very difficult, yet is crucial in different appli... more Node role explainability in complex networks is very difficult, yet is crucial in different application domains such as social science, neurosciences or computer science. Many efforts have been made on the quantification of hubs revealing particular nodes in a network using a given structural property. Yet, in several applications, when multiple instances of networks are available and several structural properties appear to be relevant, the identification of node roles remains largely unexplored. Inspired by the node automorphically equivalence relation, we define an equivalence relation on graph nodes associated with any collection of nodal statistics (i.e. any functions on the node-set). This allows us to define new graph global measures: the power coefficient, and the orthogonality score to evaluate the parsimony and heterogeneity of a given nodal statistics collection. In addition, we introduce a new method based on structural patterns to compare graphs that have the same vertices set. This method assigns a value to a node to determine its role distinctiveness in a graph family. Extensive numerical results of our method are conducted on both generative graph models and real data concerning human brain functional connectivity. The differences in nodal statistics are shown to be dependent on the underlying graph structure. Comparisons between generative models and real networks combining two different nodal statistics reveal the complexity of human brain functional connectivity with differences at both global and nodal levels. Using a group of 200 healthy controls connectivity networks, our method computes high correspondence scores among the whole population, to detect homotopy, and finally quantify differences between comatose patients and healthy controls.

Anisotropic Hybrid Networks for Liver Tumor Segmentation with Uncertainty Quantification

Lecture notes in computer science, 2023

Exploring continual learning strategies in artificial neural networks through graph-based analysis of connectivity: insights from a brain-inspired perspective

HAL (Le Centre pour la Communication Scientifique Directe), Oct 19, 2023

Artificial Neural Networks (ANNs) aim at mimicking information processing in biological networks.... more Artificial Neural Networks (ANNs) aim at mimicking information processing in biological networks. In cognitive neuroscience, graph modeling is a powerful framework widely used to study brain structural and functional connectivity. Yet, the extension of graph modeling to ANNs has been poorly explored especially in term of functional connectivity (i.e. the contextual change of the activity's units in networks). From the perspective of designing more robust and interpretable ANNs, we study how a brain-inspired graph-based approach can be extended and used to investigate their properties and behaviors. We focus our study on different continual learning strategies inspired by the human brain and modeled with ANNs. We show that graph modeling offers a simple and elegant framework to deeply investigate ANNs, compare their performances and explore deleterious behaviors such as catastrophic forgetting.

Multiple sclerosis new lesions segmentation challenge

Multiple sclerosis new lesions segmentation challenge: Structured description of the challenge de... more Multiple sclerosis new lesions segmentation challenge: Structured description of the challenge design CHALLENGE ORGANIZATION Title Use the title to convey the essential information on the challenge mission. Multiple sclerosis new lesions segmentation challenge Challenge acronym Preferable, provide a short acronym of the challenge (if any).

Realistic model for temporal reasoning in real-time patient monitoring

Applied Artificial Intelligence, Apr 1, 1996

Time is a central factor in patient monitoring. Introduction of domaindependent knowledge is esse... more Time is a central factor in patient monitoring. Introduction of domaindependent knowledge is essential to ensure efficiency of time managers especially when embedded into systems that interact with real world. We present a realistic temporal reasoning model based on two basic cognitive mechanisms: aggregation of similar observed situations and forgetting of non-relevant information. We describe in detail how we represented the proposed model and how, by refinement of domain-independent temporal entities and inferences, we added domain-specific knowledge to manage a clinical therapy. The model allows clinical observations to be incrementally interpreted as they are acquired by an intelligent system, mainly reactive in its reasoning, for the management of patients receiving respiratory support. * similar to conscious repression. ** similar to unconscious repression without a Freudian dimension.

Review Article - Mechanical ventilation: changing concepts

Mechanical ventilation is routinely delivered to patients admitted in intensive care units to red... more Mechanical ventilation is routinely delivered to patients admitted in intensive care units to reduce work of breathing, improve oxygenation, or correct respiratory acidosis. Although traditional modes of mechanical ventilation achieve many of these goals, they have important limitations. Alternative modes are supposed to handle some of these limitations and are now available on modern ventilators. This article reviews general aspects of functioning and limitations of traditional modes of mechanical ventilation, and the poten tial interest of some new promising modes.

Temporal Structures And Medical KnowledgeRepresentation

WIT Transactions on Information and Communication Technologies, 1970

Many realistic applications are temporally rich domains. The classic formalisms based on atomic t... more Many realistic applications are temporally rich domains. The classic formalisms based on atomic temporal primitives, such as interval or point, are insufficient to reflect the diversity and the evolution of real world situations. We propose to build, out of an initial set of domainindependent temporal structures and relations, more sophisticated domain-dependent layers incorporating the semantic of the considered application. The temporal constraint propagation is reduced to the needs of the discourse to model. Thus, the system is able to reason in real-time. Our approach is applied in order to represent time aspects in diagnosis and therapeutic tasks involved in a patient monitoring process.

MSSEG-2: Un challenge d'imagerie médicale sur VIP

HAL (Le Centre pour la Communication Scientifique Directe), Dec 13, 2021

Using a general theory of time and change in patient monitoring: Experiment and evaluation

Computers in Biology and Medicine, Sep 1, 1997

In this paper, we propose to use one of the well-known general theories of time and change, namel... more In this paper, we propose to use one of the well-known general theories of time and change, namely the Event Calculus [1], to represent temporal aspects in intelligent medical monitoring systems. In particular, we explore the application of CEC [2] (an efficient implementation of the Event Calculus) to the management of mechanical ventilation. First, we present the prototype we have built, which has been extensively tested on patient's data from real clinical cases. Then, we provide a thorough evaluation of the obtained results, pointing out both strengths and weaknesses of the approach, and identifying a number of extensions which can be extremely useful to scale up the medical application of the approach.

Chapter 8 - Consciousness without a cortex

HAL (Le Centre pour la Communication Scientifique Directe), 2021

La conscience comme fonction émergente du cerveau humain Jusque dans la terminologie zoologique, ... more La conscience comme fonction émergente du cerveau humain Jusque dans la terminologie zoologique, Homo sapiens 1 , la conscience a été pour une large part considérée comme une spécificité humaine. Débordant de ce courant dominant de la pensée occidentale, les approches philosophiques initiées dans la deuxième moitié du XX ème siècle, l'antispécisme d'un côté (Singer, 1975) et la singularité informatique de l'autre (Chalmers, 2010; Good, 1965), partagent de plus en plus la conviction que certains animaux et certaines machines jouissent de facultés comparables à celles d'un humain conscient. Les enjeux moraux sont de taille. Pendant plusieurs siècles, le débat est resté souterrain pour l'ensemble des scientifiques français. Dans la droite ligne de Descartes, la physiologie expérimentale héritée de Claude Bernard (Bernard, 1865) a considéré le corps humain comme un mécanisme semblable à celui d'autres animaux. Le philosophe mathématicien, dans une vision dualiste séparant matière et esprit, accordait, mais à l'humain seul, une parcelle divine, l'âme. Sa vision de l'âme restait en accord avec la religion du Dieu incarné, créateur d'un Homme à son image, libre et supérieur à la nature et aux entités qui la composent. Ces questions spirituelles sont restées hors du champ mono-disciplinaire du scientifique et hors de portée de ses outils. Il s'en est donc tenu à l'animal-machine. Cependant la place privilégiée de l'homme a été relativisée par la théorie de l'évolution des espèces (Darwin, 1859). La neuropsychologie, appuyée par la neuroimagerie, a mis directement en lien des réseaux corticaux communs à différentes espèces et certaines facultés cognitives et affectives. Par ailleurs, sur l'aspect fonctionnel, les capacités des machines actuelles, en termes de capacité mémorielle et de traitement de l'information, les rapprochent plus des organismes vivants que des horloges des siècles passés. Descartes proposait que, contrairement à l'homme, pour des machines « qui eussent des organes et la figure d'un singe, ou de quelque animal sans raison, nous n'aurions aucun moyen pour reconnaître qu'elles ne seraient pas en tout de même nature que les animaux » (Descartes, 1637). Il soulignait ainsi la dichotomie entre l'Homme et de simples mécanismes naturels ou artificiels. Le célèbre canard de Vaucanson, un automate reproduisant les mouvements d'un canard, aurait été construit relativement à cette posture, à laquelle se sont opposés d'autres penseurs, comme Perrault dans son « Avertissement » introduisant « La mechanique des animaux » (Perrault, 1680) 2 , puis Diderot (Diderot and (Le Rond) D'Alembert, 1751-72) qui accordait volontiers une âme aux animaux dans la célèbre Encyclopédie. A leur suite, associant âme et conscience, nous reconsidérons dans quelle mesure, dans le contexte des neurosciences et de l'informatique du XXI ème siècle, la conscience resterait ou non une spécificité humaine. Et si non, à quel type de conscience les entités sans cortex pourraient-elles prétendre ? Nous nous plaçons d'entrée dans une perspective matérialiste, moniste, partagée par la majorité des neuroscientifiques (voir enquête, annexe 1). Dans cette perspective,

A Multicenter Preclinical MRI Study: Definition of Rat Brain Relaxometry Reference Maps

bioRxiv (Cold Spring Harbor Laboratory), Feb 7, 2020

Nodal-statistics-based equivalence relation for graph collections

Physical review, Jan 18, 2023

Node role explainability in complex networks is very difficult, yet is crucial in different appli... more Node role explainability in complex networks is very difficult, yet is crucial in different application domains such as social science, neurosciences or computer science. Many efforts have been made on the quantification of hubs revealing particular nodes in a network using a given structural property. Yet, in several applications, when multiple instances of networks are available and several structural properties appear to be relevant, the identification of node roles remains largely unexplored. Inspired by the node automorphically equivalence relation, we define an equivalence relation on graph nodes associated with any collection of nodal statistics (i.e. any functions on the node-set). This allows us to define new graph global measures: the power coefficient, and the orthogonality score to evaluate the parsimony and heterogeneity of a given nodal statistics collection. In addition, we introduce a new method based on structural patterns to compare graphs that have the same vertices set. This method assigns a value to a node to determine its role distinctiveness in a graph family. Extensive numerical results of our method are conducted on both generative graph models and real data concerning human brain functional connectivity. The differences in nodal statistics are shown to be dependent on the underlying graph structure. Comparisons between generative models and real networks combining two different nodal statistics reveal the complexity of human brain functional connectivity with differences at both global and nodal levels. Using a group of 200 healthy controls connectivity networks, our method computes high correspondence scores among the whole population, to detect homotopy, and finally quantify differences between comatose patients and healthy controls.

Pourquoi et comment favoriser le partage en neuro-imagerie ?

Annales des Mines - Réalités industrielles, Jul 12, 2021

Généralement, la distance entre les distributions de données que l'on veut séparer est faible ave... more

Effects of background and contour luminance on the hue and brightness of the Watercolor effect

doi: bioRxiv preprint greater hue filling-in at higher luminances of the interior contour. For th... more doi: bioRxiv preprint greater hue filling-in at higher luminances of the interior contour. For the brightness judgments, the higher background luminance generated less brightness filling-in at higher luminances of the interior contour. The results indicate different effects of the inner contour and background on the induction of the brightness and coloration percepts of the WCE, suggesting that they are mediated by different mechanisms.

Fast Uncertainty Quantification for Deep Learning-based MR Brain Segmentation

HAL (Le Centre pour la Communication Scientifique Directe), Jan 24, 2022

Quantifying the uncertainty attached to Deep Learning models predictions can help their interpret... more Quantifying the uncertainty attached to Deep Learning models predictions can help their interpretation, and thus their acceptance in critical fields. Yet, current standard approaches rely on multi-steps approaches, increasing the inference time and memory cost. In clinical routine, the automated prediction has to integrate into the clinical consultation timeframe, raising the need for faster and more efficient uncertainty quantification methods. In this work, we propose a novel model, named as BEHT, and evaluate it on an automated segmentation task of White-Matter Hyperintensities from T2-weighted FLAIR MRI sequences of Multiple-Sclerosis (MS) patients. We demonstrate that this approach outputs predictive uncertainty much faster than the state-of-the-art Monte Carlo Dropout approach, with a similar-and even slightly better-accuracy. Interestingly, our approach distinguishes 2 distinct sources of uncertainties, namely aleatoric and epistemic uncertainties.

The Superior Colliculus in De Novo Parkinson's Disease: A Prodromal Dysfunction?

Social Science Research Network, 2018

Network Embedding For Brain Connectivity

In Neurosciences, networks are currently used for representing the brain connections system with ... more In Neurosciences, networks are currently used for representing the brain connections system with the purpose of determining the specific characteristics of the brain itself. However, discriminating between a healthy human brain network and a pathological one using common network descriptors could be misleading. For this reason, we explored network embedding techniques with the purpose of brain connectivity networks comparison. We proposed first the definition of representative graph for healthy brain connectivity. Then, two classification procedures through embedding are introduced, achieving good accuracy results in different datasets. Moreover, the intriguing power of this technique is given by the possibility of visualizing networks in a low-dimensional space, facilitating the interpretation of the differences between networks under diverse conditions e.g. normal or pathological.

Artificial intelligence in neuroimaging: four challenges to improve interpretation of brain images

Artificial Intelligence in Medicine, Feb 1, 2004

Artificial Intelligence in Medicine in Europe AIME’03

Artificial Intelligence in Medicine, May 1, 2005

Modulation of visual hallucinations originating from deafferented occipital cortex by robotized transcranial magnetic stimulation

Clinical Neurophysiology, Aug 1, 2020

This is a PDF file of an article that has undergone enhancements after acceptance, such as the ad... more This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.