Scientific Publications of the University of Toulouse II Le Mirail
Not a member yet
92205 research outputs found
Sort by
Can Anatomical Information Guide the Performance of Convolutional Neural Networks for Classifying Neurodegenerative Diseases Using Brain MRI?
No full textInternational audienceConvolutional neural networks (CNNs) have been gaining outstanding success in classifying neurodegenerative diseases using brain magnetic resonance imaging (MRI) images. Given the highly informative content of brain MRI and the black-box nature of CNNs, it can be challenging to disentangle the decision-making process and identify cerebral regions that are relevant for the classification. In this study, we aimed to investigate the behavior of a 3D CNN for the classification of multiple system atrophy (MSA), a rare atypical parkinsonian syndrome, by exploiting a priori anatomical information related to this neurodegenerative disorder. We considered three regions of interest (ROIs) affected by pathological changes as input to a 3D CNN in combination with the T1-weighted MRI images. By comparing different CNN implementations and introducing a specific module accounting for the anatomical information, we achieved high accuracy (78–89%) in discriminating MSA patients from healthy controls. Furthermore, performances varied depending on the ROI used, with lower sensitivity for MSA patients when considering bigger regions. Instead, modifying the CNN with the anatomical gate module, an anatomically oriented attention mechanism, improved the classification using smaller regions. These findings represent an encouraging starting point to exploit fully disease-specific a priori knowledge and enhance classification performance while supporting better interpretability and advancing deep learning-based aid-to-diagnosis tools in clinical practice
Modelling the impact of lot priorities on production cycle times: a data-driven approach
No full textInternational audienceThe production of electronic chips from silicon wafers in semiconductor manufacturing is widely regarded as one of the most complex industrial processes. This production context is characterised by high global demand, intensive utilisation of costly equipment to ensure profitability, and consequently, a high level of saturation that generates long queuing waiting times. Combined with the hundreds of operations required to produce a single lot, this leads to cycle times spanning several weeks or even months. To meet the varying cycle time requirements of different products in such a saturated environment, a priority queuing discipline is implemented, using a mix of different priority classes. In this context, and using data-driven models, this paper characterises the impact of the priority mix on the cycle times of products in each priority class within a wafer manufacturing facility. A fluid analytical queuing model is proposed, which extrapolates the effects of different priority mixes beyond the observed data range, using priority scores estimated from historical data. The model reveals a linear relationship between the ratio of hot lots and the speed-up of priority classes, defined as the inverse of their relative mean queuing waiting time. The results are consistent with findings reported in the literature
Data-driven queueing modelling: a simulation case study of emergency department crowding
No full textInternational audienceObjectives Emergency department crowding refers to a complex state of congestion associated with a set of performance indicators such as occupation levels, waiting times and specific scores. Among current methods to model it, an objective gap exists between forecasting machine learning methods, focusing on prediction precision and queueing and simulation methods, focusing on capturing correctly the effect of decision variables for evaluation and optimisation purposes. The objective of the present analysis is to implement and numerically validate a novel data-driven queueing methodology that can bridge this gap and to show its applicability in a simulation case study. Methods A statistical modelling of the queueing processes, particularly patient departure rates and probabilities, is developed to cross the gap defined above. Using the data from a major emergency department of eastern France, the resultant data-driven queueing network model is validated and applied through a synchronous simulation algorithm. Results The model obtained considers the complex effects of patient arrivals and doctor and nurse allocations while offering an unbiased and accurate measure of long-term crowding. Its application with the case study quantifies the impact of the opening of new Unscheduled Care Services on emergency department crowding. Discussion The new data-driven queueing methodology is able to model and quantify complex crowding effects at a detailed level in an emergency department. Conclusions This study shows an alternative approach successfully bridging the modelling gap by establishing a model that can effectively predict system crowding dynamics under the influence of multiple key variables
Analyse didactique de la pertinence d’un dispositif de formation visant à développer une démarche réflexive chez les professeurs des écoles stagiaires
No full textInternational audienceThe main aim of this research article is to demonstrate the relevance and appropriateness of a classroom visit by a trainee primary school teacher in France. The visit is structured around three phases: observation by a trainer of the implementation of a situation by a trainee primary school teacher, post-observation interview, production of a post-interview reflective piece of writing. The methodology is based, on the one hand, on the analysis of the interview that follows the classroom observation and, on the other, on the study of the professional writing drafted by the trainee. Throughout the analysis of the system, the aim is to detect changes in the trainee teacher's posture, and thus to highlight the trainee's ability to adopt a reflective posture in terms of how he or she brings his or her pupils to experience a mathematical situation.Cet article de recherche a pour principale finalité de montrer la pertinence et la cohérence d’un dispositif spécifique de visite en classe d’un enseignant stagiaire au primaire. Ce dispositif s’articule en trois temps : observation par un formateur de la mise en œuvre d’une situation mathématique par un professeur des écoles stagiaire, entretien post-observation, production d’un écrit réflexif post-entretien. La méthodologie repose d’une part sur l’analyse de l’entretien qui fait suite à l’observation de classe, d’autre part sur l’étude de l’écrit professionnel rédigé par le stagiaire. Il s’agit tout au long de l’analyse du dispositif de déceler les changements d’attitude de l’enseignant stagiaire et ainsi de mettre en lumière sa capacité à adopter une démarche réflexive quant à sa manière de faire vivre une situation mathématique à ses élèves
Contact-Robust Trajectory Planning via Parametric Sensitivity Analysis for Hybrid Robotic Systems
No full textInternational audienceIn this paper, we combine first-order approximations of hybrid systems (i.e., the so-called saltation matrix) with previous works on parametric sensitivity for continuous systems to propose a general framework for robust trajectory optimization of hybrid systems subject to parametric uncertainties. A method for computing parametric sensitivities of both continuous dynamics and hybrid events is presented. The obtained "hybrid parametric sensitivity" is then combined with sensitivity-based tubes that encapsulate all possible perturbed states and control trajectories given a known bounded range for the uncertain parameters. The proposed method is then applied to the problem of planning robust trajectories for legged robot systems, which allows obtaining trajectories that remain feasible w.r.t. the contact constraints even in presence of uncertainties in the dynamics, guard conditions, and reset maps. We also illustrate one of the fundamental limitations of first-order approximations, that is, the fact that the sensitivity reset time is fixed, and propose an extension to the sensitivity analysis that can form the basis for future developments
Predictors of brain death after hanging-induced cardiac arrest
No full textInternational audienceBrain death after hanging-induced cardiac arrest is a fatal complication about which few data are available. We aimed at identifying the early predictors of progression to brain death in patients with hanging-induced cardiac arrest
From the block to the blade: The contribution of lithic technology of the Villazette site (Creysse, Dordogne) to understanding the early stages of blade production in the Middle Magdalenian
No full textInternational audienceIn southwestern France, blades attributed to the Late Middle Magdalenian are occasionally found in the form of batches of finished objects in particular contexts such as painting cave like Labastide or Enlène in the Pyrenees. The lack of first stages of production, absent or poorly documented in “consumption” sites, represent a key obstacle to fully understand the chaîne opératoire. As such, investigating a lithic assemblage from “production” sites through refitting and technological analysis raises questions about technological segmentation of blade production.Located in the Aquitaine Basin, the site of Villazette (Creysse, Dordogne) is situated on the low terraces of the Dordogne River, in an area rich in high-quality ‘Bergeracois’ flint. Excavations have revealed a series of open-air occupations, including one layer related to the Late Middle Magdalenian phase showing evidence of blade production.The analysis of cores and other products confirms the presence of blades exceeding twenty centimeters, along with the production of smaller blades for toolmaking. Refitting studies provide insight into the early stages of blade production: raw blocks were collected on-site and shaped through the preparation of anterior and posterior crests, ensuring an efficient reduction sequence.Furthermore, the spatial distribution analysis reveals two concentrations with distinct functions, raising questions about the organization of the occupation. The discovery of this site underscores the archaeological potential of the Dordogne’s riverbanks for exploring the organization of Magdalenian groups
Bilevel gradient methods and the Morse parametric qualification condition
No full textInternational audienceWe introduce the Morse parametric qualification condition for bilevel programming. Generic semi-algebraic functions are Morse parametric in a piecewise sense. Thus, bilevel programs with a Morse parametric lower level constitute a relevant intermediate class between strongly convex and fully generic lower levels. In this framework, we study bilevel gradient algorithms with two strategies: the single-step multi-step strategy, which involves a sequence of steps on the lower-level problems followed by one step on the upper-level problem, and a differentiable programming strategy that optimizes a smooth approximation of the bilevel problem. While the first is shown to be a biased gradient method on the problem with rich properties, the second, inspired by meta-learning applications, is less stable but offers simplicity and ease of implementation
An iterative CP approach for handling min/max worload constraints in preemptive JSP
No full textInternational audienceAn iterative CP approach for handling min/max worload constraints in preemptive JS