Laboratoire d'Informatique Fondamentale et Appliquée de Tours

Knowledge of whole heart anatomy is a prerequisite for many clinical applications. Whole heart segmentation (WHS), which delineates substructures of the heart, can be very valuable for modeling and analysis of the anatomy and functions of the heart. However, automating this segmentation can be challenging due to the large variation of the heart shape, and different image qualities of the clinical data. To achieve this goal, an initial set of training data is generally needed for constructing priors or for training. Furthermore, it is difficult to perform comparisons between different methods, largely due to differences in the datasets and evaluation metrics used. This manuscript presents the methodologies and evaluation results for the WHS algorithms selected from the submissions to the Multi-Modality Whole Heart Segmentation (MM-WHS) challenge, in conjunction with MICCAI 2017. The challenge provided 120 three-dimensional cardiac images covering the whole heart, including 60 CT and 60 MRI volumes, all acquired in clinical environments with manual delineation. Ten algorithms for CT data and eleven algorithms for MRI data, submitted from twelve groups, have been evaluated. The results showed that the performance of CT WHS was generally better than that of MRI WHS. The segmentation of the substructures for different categories of patients could present different levels of challenge due to the difference in imaging and variations of heart shapes. The deep learning (DL)-based methods demonstrated great potential, though several of them reported poor results in the blinded evaluation. Their performance could vary greatly across different network structures and training strategies. The conventional algorithms, mainly based on multi-atlas segmentation, demonstrated good performance, though the accuracy and computational efficiency could be limited. The challenge, including provision of the annotated training data and the blinded evaluation for submitted algorithms on the test data, continues as an ongoing benchmarking resource via its homepage (www.sdspeople.fudan.edu.cn/zhuangxiahai/0/mmwhs/).

This title presents a large variety of models and algorithms dedicated to the resource-constrained project scheduling problem (RCPSP), which aims at scheduling at minimal duration a set of activities subject to precedence constraints and limited resource availabilities. In the first part, the standard variant of RCPSP is presented and analyzed as a combinatorial optimization problem. Constraint programming and integer linear programming formulations are given. Relaxations based on these formulations and also on related scheduling problems are presented. Exact methods and heuristics are surveyed. Computational experiments, aiming at providing an empirical insight on the difficulty of the problem, are provided. The second part of the book focuses on several other variants of the RCPSP and on their solution methods. Each variant takes account of real-life characteristics which are not considered in the standard version, such as possible interruptions of activities, production and consumption of resources, cost-based approaches and uncertainty considerations. The last part presents industrial case studies where the RCPSP plays a central part. Applications are presented in various domains such as assembly shop and rolling ingots production scheduling, project management in information technology companies and instruction scheduling for VLIW processor architectures.

Spermatozoa are among the most diversified cells in the animal kingdom, but the underlying evolutionary forces affecting intraspecific variation in sperm morphology are poorly understood. It has been hypothesized that sperm competition is a potent selection pressure on sperm variation within species. Here, we examine intraspecific variation in total sperm length of 22 wild passerine bird species (21 genera, 11 families) in relation to the risk of sperm competition, as expressed by the frequency of extrapair paternity and relative testis size. We demonstrate, by using phylogenetic comparative methods, that between-male variation in sperm length within species is closely and negatively linked to the risk of sperm competition. This relationship was even stronger when only considering species in which data on sperm length and extrapair paternity originated from the same populations. Intramale variation in sperm length within species was also negatively, although nonsignificantly, related to sperm competition risk. Our findings suggest that postcopulatory sexual selection is a powerful evolutionary force reducing the intraspecific phenotypic variation in sperm-size traits, potentially driving the diversification of sperm morphology across populations and species.

The biological mechanisms involved in SARS-CoV-2 infection are only partially understood. Thus we explored the plasma metabolome of patients infected with SARS-CoV-2 to search for diagnostic and/or prognostic biomarkers and to improve the knowledge of metabolic disturbance in this infection. We analyzed the plasma metabolome of 55 patients infected with SARS-CoV-2 and 45 controls by LC-HRMS at the time of viral diagnosis (D0). We first evaluated the ability to predict the diagnosis from the metabotype at D0 in an independent population. Next, we assessed the feasibility of predicting the disease evolution at the 7th and 15th day. Plasma metabolome allowed us to generate a discriminant multivariate model to predict the diagnosis of SARS-CoV-2 in an independent population (accuracy > 74%, sensitivity, specificity > 75%). We identified the role of the cytosine and tryptophan-nicotinamide pathways in this discrimination. However, metabolomic exploration modestly explained the disease evolution. Here, we present the first metabolomic study in SARS-CoV-2 patients which showed a high reliable prediction of early diagnosis. We have highlighted the role of the tryptophan-nicotinamide pathway clearly linked to inflammatory signals and microbiota, and the involvement of cytosine, previously described as a coordinator of cell metabolism in SARS-CoV-2. These findings could open new therapeutic perspectives as indirect targets.

This paper deals with a special case of Project Scheduling problem: there is a project to schedule, which is made up of activities linked by precedence relations. Each activity requires specific skills to be done. Moreover, resources are staff members who master fixed skill(s). Thus, each resource requirement of an activity corresponds to the number of persons doing the corresponding skill that must be assigned to the activity during its whole processing time. We search for an exact solution that minimizes the makespan, using a Branch-and-Bound method.

We introduce a novel superpoint-based transformer architecture for efficient semantic segmentation of large-scale 3D scenes. Our method incorporates a fast algorithm to partition point clouds into a hierarchical superpoint structure, which makes our preprocessing 7 times faster than existing superpoint-based approaches. Additionally, we leverage a self-attention mechanism to capture the relationships between superpoints at multiple scales, leading to state-of-the-art performance on three challenging benchmark datasets: S3DIS (76.0% mIoU 6-fold validation), KITTI-360 (63.5% on Val), and DALES (79.6%). With only 212k parameters, our approach is up to 200 times more compact than other state-of-the-art models while maintaining similar performance. Furthermore, our model can be trained on a single GPU in 3 hours for a fold of the S3DIS dataset, which is 7× to 70× fewer GPU-hours than the best-performing methods. Our code and models are accessible at github.com/drprojects/superpoint_transformer.

This paper presents a comprehensive framework for studying methods of pulse rate estimation relying on remote photoplethysmography (rPPG). There has been a remarkable development of rPPG techniques in recent years, and the publication of several surveys too, yet a sound assessment of their performance has been overlooked at best, whether not undeveloped. The methodological rationale behind the framework we propose is that in order to study, develop and compare new rPPG methods in a principled and reproducible way, the following conditions should be met: 1) a structured pipeline to monitor rPPG algorithms' input, output, and main control parameters; 2) the availability and the use of multiple datasets; and 3) a sound statistical assessment of methods' performance. The proposed framework is instantiated in the form of a Python package named pyVHR (short for Python tool for Virtual Heart Rate), which is made freely available on GitHub (github.com/phuselab/pyVHR). Here, to substantiate our approach, we evaluate eight well-known rPPG methods, through extensive experiments across five public video datasets, and subsequent nonparametric statistical analysis. Surprisingly, performances achieved by the four best methods, namely POS, CHROM, PCA and SSR, are not significantly different from a statistical standpoint higighting the importance of evaluate the different approaches with a statistical assessment.

In this paper, we introduce a new method to solve the unsupervised clustering problem, based on a modelling of the chemical recognition system of ants. This system allow ants to discriminate between nestmates and intruders, and thus to create homogeneous groups of individuals sharing a similar odor by continuously exchanging chemical cues. This phenomenon, known as "colonial closure", inspired us into developing a new clustering algorithm and then comparing it to a well-known method such as K-MEANS method. Our results show that our algorithm performs better than K-MEANS over artificial and real data sets, and furthermore our approach requires less initial information (such as number of classes, shape of classes, limitation in the types of attributes handled).

This work is aimed at the detection of adult images that appear in Internet. Skin detection is of the paramount importance in the detection of adult images. We build a maximum entropy model for this task. This model, called the First Order Model in this paper, is subject to constraints on the color gradients of neighboring pixels. Parameter estimation as well as optimization cannot be tackled without approximations. With Bethe tree approximation, parameter estimation is eradicated and the Belief Propagation algorithm permits to obtain exact and fast solution for skin probabilities at pixel locations. We show by the Receiver Operating Characteristics (ROC) curves that our skin detection improves the performance in the previous work in the context of skin pixel detecton rate and false positive rate. The output of skin detection is a grayscale skin map with the gray level indicating the belief of skin. We then calculate 9 simple features from this map which form a feature vector. We use the fit ellipses to catch the characteristics of skin distribution. Two fit ellipses are used for each skin map---the fit ellipse of all skin regions and the fit ellipse of the largest skin region. They are called respectively Global Fit Ellipse and Local Fit Ellipse in this paper. A multi-layer perceptron classifier is trained for these features. Plenty of experimental results are presented including photographs and a ROC curve calculated over a test set of 5,084 photographs, which show stimulating performance for such simple features.

Over the recent years, there has been an increasing interest in large-scale classification of remote sensing images. In this context, the Inria Aerial Image Labeling Benchmark has been released online in December 2016. In this paper, we discuss the outcomes of the first year of the benchmark contest, which consisted in dense labeling of aerial images into building / not building classes, covering areas of five cities not present in the training set. We present four methods with the highest numerical accuracies, all four being convolutional neural network approaches. It is remarkable that three of these methods use the U-net architecture, which has thus proven to become a new standard in image dense labeling.

Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.

Graphs are an extremely general and powerful data structure. In pattern recognition and computer vision, graphs are used to represent patterns to be recognized or classified. Detection of maximum common subgraph (MCS) is useful for matching, comparing and evaluate the similarity of patterns. MCS is a well known NP-complete problem for which optimal and suboptimal algorithms are known from the literature. Nevertheless, until now no effort has been done for characterizing their performance. The lack of a large database of graphs makes the task of comparing the performance of different graph matching algorithms difficult, and often the selection of an algorithm is made on the basis of a few experimental results available. In this paper, three optimal and well-known algorithms for maximum common subgraph detection are described. Moreover a large database containing various categories of pairs of graphs (e.g. random graphs, meshes, bounded valence graphs), is presented, and the performance

Global warming is a major threat to agriculture worldwide. Between 2008 and 2013, some coffee producing countries in South and Central America suffered from severe epidemics of coffee leaf rust (CLR), resulting in high economic losses with social implications for coffee growers. The climatic events not only favoured the development of the pathogen but also affected the physiological status of the coffee plant. The main objectives of the study were to evaluate how the physiological status of the coffee plant modified by different environmental conditions impact on the pathogenesis of CLR and to identify indicators of the physiological status able to predict rust incidence. Three rust susceptible genotypes (one inbred line and two hybrids) were grown in controlled conditions with a combination of thermal regime (TR), nitrogen and light intensity close to the field situation before being inoculated with the rust fungus Hemileia vastatrix. It has been demonstrated that a TR of 27 -22 °C resulted in 2 000 times higher sporulation than with a TR of 23-18 °C. It has been also shown that high light intensity combined with low nitrogen fertilisation modified the CLR pathogenesis resulting in huge sporulation. CLR sporulation was significantly lower in the F1 hybrids than in the inbred line. The hybrid vigour may have reduced disease incidence. Among the many parameters studied, parameters related to photosystem II and photosynthetic electron transport chain components appeared as indicators of the physiological status of the coffee plant able to predict rust sporulation intensity. Taken together, these results show that CLR sporulation not only depends on the TR but also on the physiological status of the coffee plant, which itself depends on agronomic conditions. Our work suggests that vigorous varieties combined with a shaded system and appropriate nitrogen fertilisation should be part of an agro-ecological approach to disease control.

The new version of the International Classification of Diseases (ICD-11) mentions the existence of four different profiles in the verbal part of the Autism Spectrum Disorder (ASD), describing them as combinations of either spared or impaired functional language and intellectual abilities. The aim of the present study was to put ASD heterogeneity to the forefront by exploring whether clear profiles related to language and intellectual abilities emerge when investigation is extended to the entire spectrum, focusing on verbal children. Our study proposed a systematic investigation of both language (specifically, structural language abilities) and intellectual abilities (specifically, nonverbal cognitive abilities) in 51 6- to 12-year-old verbal children with ASD based on explicitly motivated measures. For structural language abilities, sentence repetition and nonword repetition tasks were selected; for nonverbal cognitive abilities, we chose Raven's Progressive Matrices, as well as Matrix Reasoning and Block Design from the Wechsler Scales. An integrative approach based on cluster analyses revealed five distinct profiles. Among these five profiles, all four logically possible combinations of structural language and nonverbal abilities mentioned in the ICD-11 were detected. Three profiles emerged among children with normal language abilities and two emerged among language-impaired children. Crucially, the existence of discrepant profiles of abilities suggests that children with ASD can display impaired language in presence of spared nonverbal intelligence or spared language in the presence of impaired nonverbal intelligence, reinforcing the hypothesis of the existence of a separate language module in the brain. Autism Res 2020, 13: 1155-1167. © 2020 International Society for Autism Research, Wiley Periodicals, Inc. LAY SUMMARY: The present work put Autism Spectrum Disorder heterogeneity to the forefront by exploring whether clear profiles related to language and cognitive abilities emerge when investigation is extended to the entire spectrum (focusing on verbal children). The use of explicitly motivated measures of both language and cognitive abilities and of an unsupervised machine learning approach, the cluster analysis, (a) confirmed the existence of all four logically possible profiles evoked in the new ICD-11, (b) evoked the existence of (at least) a fifth profile of language/cognitive abilities, and (c) reinforced the hypothesis of a language module in the brain.

Abstract This paper deals with the routing problem of health care sta in a home health care problem. Given a list of patients needing several cares, the problem is to assign cares to care workers. Some cares have to be performed by several persons and some cares cannot be performed with others. If a patient needs several cares, he may want to be treated by the same person. Moreover, some skills constraints and time windows have to be satis ed. We show that this problem is equivalent to a multiple traveling salesman problem with time windows (mTSPWT) with some speci c constraints. For solving this problem, we propose an integer linear program with some technical improvements. 1

UNLABELLED: Drugs interacting with opioid or N-methyl-D-aspartate (NMDA) receptors may have differing effects on post-surgical sensory changes, such as central inhibition or spinal excitation. We compared the effect of supplementing isoflurane/N2O/O2 anesthesia with an opioid agonist (fentanyl [n = 15]) or two drugs inhibiting the NMDA system differently (magnesium, ketamine [n = 15 in each group]) on sensory changes after abdominal hysterectomy. Electric sensation, pain detection, and pain tolerance thresholds were determined (preoperatively and 1, 4, 24 h, and 5 days postoperatively) in arm, thoracic, incision, and leg dermatomes together with pain scores and cumulative morphine consumption. Thresholds relative to the arm were derived to unmask segmental sensory changes hidden by generalized changes. Absolute thresholds were increased 1-24 h, returning to baseline on Day 5, without overall differences among drugs. Fentanyl thresholds were lower 1 h and higher 5 days postoperatively compared with magnesium and ketamine; thresholds were lower at 24 h for magnesium versus ketamine. Relative thresholds increased compared with baseline only with fentanyl (1-4 h); none decreased. Pain scores and morphine consumption were similar. Thus, all adjuvants suppressed spinal sensitization after surgery. Fentanyl showed the most, and magnesium the least, central sensory inhibition up to 5 days postoperatively, with different patterns of inhibition directly postsurgery versus later. Differences in sensory processing were not reflected in clinical measures. IMPLICATIONS: We studied the effects on postsurgical sensory processing of general anesthesia supplemented by drugs affecting opioid or N-methyl-D-aspartate receptors using sensory thresholds. Generalized central sensory inhibition, differently affected by the drugs, predominated after surgery. All drugs suppressed spinal excitation. Clinical pain measures did not reflect sensory change.

We introduce the electric vehicle routing problem with public-private recharging strategy in which vehicles may recharge en route at public charging infrastructure as well as at a privately-owned depot. To hedge against uncertain demand at public charging stations, we design routing policies that anticipate station queue dynamics. We leverage a decomposition to identify good routing policies, including the optimal static policy and fixed-route-based rollout policies that dynamically respond to observed queues. The decomposition also enables us to establish dual bounds, providing a measure of goodness for our routing policies. In computational experiments using real instances from industry, we show the value of our policies to be within 10% of a dual bound. Furthermore, we demonstrate that our policies significantly outperform the industry-standard routing strategy in which vehicle recharging generally occurs at a central depot. Our methods stand to reduce the operating costs associated with electric vehicles, facilitating the transition from internal-combustion engine vehicles.

Production scheduling and vehicle routing are both well-studied problems in literature. A coordinated approach that does not solve these interrelated problems sequentially improves the overall performance. An integrated approach becomes imperative in the current competitive business environment as it helps to save costs and to achieve on-time deliveries. This paper aims to provide a review on integrated scheduling and outbound vehicle routing problems where vehicle routing decisions are explicitly taken into account, including 65 studies since 2010. The various problem characteristics, constraints and solution methods used in the literature are summarized and a general model will be presented. Some suggestions for future research directions are given.

Remote photoplethysmography (rPPG) aspires to automatically estimate heart rate (HR) variability from videos in realistic environments. A number of effective methods relying on data-driven, model-based and statistical approaches have emerged in the past two decades. They exhibit increasing ability to estimate the blood volume pulse (BVP) signal upon which BPMs (Beats per Minute) can be estimated. Furthermore, learning-based rPPG methods have been recently proposed. The present pyVHR framework represents a multi-stage pipeline covering the whole process for extracting and analyzing HR fluctuations. It is designed for both theoretical studies and practical applications in contexts where wearable sensors are inconvenient to use. Namely, pyVHR supports either the development, assessment and statistical analysis of novel rPPG methods, either traditional or learning-based, or simply the sound comparison of well-established methods on multiple datasets. It is built up on accelerated Python libraries for video and signal processing as well as equipped with parallel/accelerated ad-hoc procedures paving the way to online processing on a GPU. The whole accelerated process can be safely run in real-time for 30 fps HD videos with an average speedup of around 5. This paper is shaped in the form of a gentle tutorial presentation of the framework.

The question of data reliability is of first importance to assess the quality of manually annotated corpora. Although Cohen ' s is the prevailing reliability measure used in NLP, alternative statistics have been proposed. This paper presents an experimental study with four measures (Cohen's , Scott's , binary and weighted Krippendorff ' s ) on three tasks: emotion, opinion and coreference annotation. The reported studies investigate the factors of influence (annotator bias, category prevalence, number of coders, number of categories) that should affect reliability estimation. Results show that the use of a weighted measure restricts this influence on ordinal annotations. They suggest that weighted is the most reliable metrics for such an annotation scheme.