Publications

2021

• Processus d’appropriation et de mémorisation de raccourcis gestuels sur trackpad : Etude longitudinale des stratégies et usages des utilisateurs et impact d’une aide visuo-sémantique
  
  • Safin Stéphane
  • Maitrallin Marie
  • Fruchard Bruno
  • Lecolinet Eric
  , 2021, pp.14:1-12. De nombreuses techniques d’interaction gestuelle ont été proposées pour offrir des raccourcis aux utilisateurs. Evaluer leur efficacité réelle en termes de mémorisation est une question difficile, la plupart des études étant effectuée en laboratoire pendant une durée limitée, donc dans des conditions assez éloignées d’un usage réel. Dans cet article, nous détaillons une étude longitudinale portant sur l’appropriation spontanée d’une technique de raccourcis gestuels sur trackpad, et l’effet d’une aide visuo-sémantique sur les stratégies de mémorisation, les performances de rappel et l’usage des raccourcis. Six participants (dont trois disposant de l’aide visuo-sémantique) ont été suivis pendant six semaines. Les résultats suggèrent que la présence d’une aide visuo-sémantique modifie les stratégies de mémorisation, et permet aux usagers de mémoriser un plus grand nombre de raccourcis. L’étude souligne aussi l’importance de l’activité de configuration du logiciel comme constitutif de son appropriation, et suggère que les raccourcis pourraient être moins utiles pour leur rapidité d’activation que pour leur facilité d’usage. (10.1145/3450522.3451332)
  DOI : 10.1145/3450522.3451332
• Des leaders d’équipes virtuels pour encourager le développement du système de mémoire transactive
  
  • Biancardi Beatrice
  • Giaccaglia Ivan
  • Ravenet Brian
  • Varni Giovanna
  , 2021, pp.4:1-7. Des équipes plongées dans une tâche de résolution de problèmes pourraient bénéficier de soutien afin de développer de bonnes représentations mentales de leurs connaissances qui faciliteraient leur travail. Dans cet article, nous initions le travail de conception d’un agent conversationnel animé jouant le rôle de leader et capable d’aider une équipe à développer son système de mémoire transactive (SMT). Nous avons mené une étude de la littérature afin d’identifier des styles de leaderships et de comportements que notre leader virtuel pourrait utiliser et nous avons mené une première étude en ligne sur la perception qu’ont nos participants de l’influence de ces comportements sur l’évolution du SMT d’une équipe en activité. Nos résultats suggèrent un effet positif des deux styles employés par l’agent sur la perception du SMT des équipes, sans différence majeure entre les deux. Dans le futur, nous prévoyons de comparer avec d’autres types d’interventions (message et leader humain) afin de continuer notre travail de conception d’un agent aidant. (10.1145/3451148.3458639)
  DOI : 10.1145/3451148.3458639
• Automatic size and pose homogenization with Spatial Transformer Network to improve and accelerate pediatric segmentation
  
  • La Barbera Giammarco
  • Gori Pietro
  • Boussaid Haithem
  • Belucci Bruno
  • Delmonte Alessandro
  • Goulin Jeanne
  • Sarnacki Sabine
  • Rouet Laurence
  • Bloch Isabelle
  , 2021, pp.1773-1776. Due to a high heterogeneity in pose and size and to a limited number of available data, segmentation of pediatric images is challenging for deep learning methods. In this work, we propose a new CNN architecture that is pose and scale invariant thanks to the use of Spatial Transformer Network (STN). Our architecture is composed of three sequential modules that are estimated together during training: (i) a regression module to estimate a similarity matrix to normalize the input image to a reference one; (ii) a differentiable module to find the region of interest to segment; (iii) a segmentation module, based on the popular UNet architecture, to delineate the object. Unlike the original UNet, which strives to learn a complex mapping, including pose and scale variations, from a finite training dataset, our segmentation module learns a simpler mapping focusing on images with normalized pose and size. Furthermore, the use of an automatic bounding box detection through STN allows saving time and especially memory, while keeping similar performance. We test the proposed method in kidney and renal tumor segmentation on abdominal pediatric CT scanners. Results indicate that the estimated STN homogenization of size and pose accelerates the segmentation (25h), compared to standard data-augmentation (33h), while obtaining a similar quality for the kidney (88.01% of Dice score) and improving the renal tumor delineation (from 85.52% to 87.12%). (10.1109/ISBI48211.2021.9434090)
  DOI : 10.1109/ISBI48211.2021.9434090
• Self-Concordant Analysis of Generalized Linear Bandits with Forgetting
  
  • Russac Yoan
  • Faury Louis
  • Cappé Olivier
  • Garivier Aurélien
  , 2021. Contextual sequential decision problems with categorical or numerical observations are ubiquitous and Generalized Linear Bandits (GLB) offer a solid theoretical framework to address them. In contrast to the case of linear bandits, existing algorithms for GLB have two drawbacks undermining their applicability. First, they rely on excessively pessimistic concentration bounds due to the non-linear nature of the model. Second, they require either non-convex projection steps or burn-in phases to enforce boundedness of the estimators. Both of these issues are worsened when considering non-stationary models, in which the GLB parameter may vary with time. In this work, we focus on self-concordant GLB (which include logistic and Poisson regression) with forgetting achieved either by the use of a sliding window or exponential weights. We propose a novel confidence-based algorithm for the maximum-likehood estimator with forgetting and analyze its perfomance in abruptly changing environments. These results as well as the accompanying numerical simulations highlight the potential of the proposed approach to address non-stationarity in GLB. (10.48550/arXiv.2011.00819)
  DOI : 10.48550/arXiv.2011.00819
• Cooperating Networks To Enforce A Similarity Constraint In Paired But Unregistered Multimodal Liver Segmentation
  
  • Couteaux Vincent
  • Trintignac Mathilde
  • Nempont Olivier
  • Pizaine Guillaume
  • Vlachomitrou Anna Sesilia
  • Valette Pierre-Jean
  • Milot Laurent
  • Bloch Isabelle
  , 2021, pp.753-756. We propose a method for segmenting two unregistered images from different modalities of the same patient and study at once, while enforcing a similarity constraint between the two segmentation masks. Our method relies on a segmentation network and a registration network, cooperating to get accurate and consistent segmentation masks across modalities, while forcing the segmentor to use all information available. Experiments on a dataset of T1 and T2-weighted liver MRI show that our method enables to get more similar segmentation masks across modalities than manual annotations, without deteriorating the performance (Dice =0.95 for T1, 0.92 for T2). (10.1109/ISBI48211.2021.9433767)
  DOI : 10.1109/ISBI48211.2021.9433767
• Processus d’appropriation et de mémorisation de raccourcis gestuels sur trackpad : Etude longitudinale des stratégies et usages des utilisateurs et impact d’une aide visuo-sémantique
  
  • Safin Stéphane
  • Maitrallin Marie
  • Fruchard Bruno
  • Lecolinet Eric
  , 2021, pp.1-12. (10.1145/3450522.3451332)
  DOI : 10.1145/3450522.3451332
• Nonlinear Functional Output Regression: a Dictionary Approach
  
  • Bouche Dimitri
  • Clausel Marianne
  • Roueff François
  • d'Alché-Buc Florence
  , 2021, PMLR: Volume 130. To address functional-output regression, we introduce projection learning, a novel dictionary-based approach that learns to predict a projection of the output function on a dictionary while minimizing a functional loss. Projection learning makes it possible to use non orthogonal dictionaries and can then be combined with dictionary learning. It is thus much more flexible than expansion-based approaches relying on vectorial losses. Using reproducing kernel Hilbert spaces of vector-valued functions, this general method is instantiated as kernel-based projection learning (KPL). For the functional square loss, we propose two closed-form estimators, one for fully observed output functions and the other for partially observed ones. Both are backed theoretically by an excess risk analysis. Then, in the more general setting of integral losses based on differentiable ground losses, KPL is implemented using first-order optimization for both fully and partially observed output functions. Eventually, several robustness aspects of the proposed algorithms are highlighted on a toy dataset; and a study on two real datasets shows that they are competitive compared to other nonlinear approaches while keeping the computational cost significantly lower.
• Validation Of A Bi-Energetic Spectrum Approximation In Bone Mineral Density Measurement With A DXA Digital Twin
  
  • Haddadi Karine
  • Muller Serge
  • Bloch Isabelle
  , 2021, pp.380-384. (10.1109/ISBI48211.2021.9433940)
  DOI : 10.1109/ISBI48211.2021.9433940
• Learning Fair Scoring Functions: Bipartite Ranking under ROC-based Fairness Constraints
  
  • Vogel Robin
  • Bellet Aurélien
  • Clémençon Stéphan
  , 2021, pp.1-35. Many applications of AI involve scoring individuals using a learned function of their attributes. These predictive risk scores are then used to take decisions based on whether the score exceeds a certain threshold, which may vary depending on the context. The level of delegation granted to such systems in critical applications like credit lending and medical diagnosis will heavily depend on how questions of fairness can be answered. In this paper, we study fairness for the problem of learning scoring functions from binary labeled data, a classic learning task known as bipartite ranking. We argue that the functional nature of the ROC curve, the gold standard measure of ranking accuracy in this context, leads to several ways of formulating fairness constraints. We introduce general families of fairness definitions based on the AUC and on ROC curves, and show that our ROC-based constraints can be instantiated such that classifiers obtained by thresholding the scoring function satisfy classification fairness for a desired range of thresholds. We establish generalization bounds for scoring functions learned under such constraints, design practical learning algorithms and show the relevance our approach with numerical experiments on real and synthetic data.
• The uplink reception and downlink transmission in MU-MIMO for 5G
  
  • Askri Aymen
  , 2021. Multiple-input multiple-output (MIMO) technologies were developed to increase system capacity and offer better link reliability. They allow a dense network architecture that will allow many users to connect in the same area without experiencing slowdowns. 5G networks and beyond will use these MIMO technologies with many small antennas allowing the beam to be focused on a given area. Coupled with high-frequency bands, the use of these antennas will significantly increase throughput.In such systems, multi-user (MU)-MIMO detection in the uplink reception and MU-MIMO precoding in the downlink transmission enable separating user data streams and pre-cancelling interference. However, some challenges have to be met under realistic conditions such as the reasonable complexity of the decoding and precoding processes, the erroneous channel knowledge, and the adjacent cell interference. This thesis addresses all these limitations above for the uplink reception and the downlink transmission in MU-MIMO systems.In the uplink reception, we study the well-known sphere decoding (SD) algorithm for MIMO detection. We seek to reduce its complexity which increases exponentially with the number of antennas and the constellation size. Thus, we profit from recent advances in neural networks (NNs) to develop the low-complexity NN assisted SD. We also propose the block recursive MIMO decoding, which achieves almost the maximum likelihood (ML) performance. Using deep neural networks (DNNs), we suggest a new and low complex scheme for signal processing and cloud-RAN (C-RAN) detection. This DNN scheme aims to mimic the whole transmission in uplink C-RAN, which considers the quantization constraints at the radio remote units (RRUs) and the corrupted observations at the central processor (CP).In the downlink transmission, we study the non-linear vector perturbation (VP) precoding. We design the combined VP to serve multiple users with different modulation coding schemes (MCSs). We also introduce the block VP algorithm, which merges both linear and non-linear precoding to offer a tunable tradeoff between complexity and performance. To deal with the erroneous channel state information (CSI) in the downlink precoding, we develop the new CSI accuracy indicator reporting to design a novel precoder that is less sensitive to CSI errors.
• Cooperative multi-sensor detection under variable-length coding
  
  • Hamad Mustapha
  • Wigger Michèle
  • Sarkiss Mireille
  , 2021, pp.1-5. We investigate the testing-against-independence problem over a cooperative MAC with two sensors and a single detector under an average rate constraint on the sensors-detector links. For this setup, we design a variable-length coding scheme that maximizes the achievable type-II error exponent when the type-I error probability is limited to $\epsilon$. Similarly to the single-link result, we show here that the optimal error exponent depends on $\epsilon$ and that variable-length coding allows to increase the rates over the optimal fixed-length coding scheme by the factor $(1-\epsilon)^{-1}$. (10.1109/ITW46852.2021.9457665)
  DOI : 10.1109/ITW46852.2021.9457665
• Zero-Error Sum Modulo Two with a Common Observation
  
  • Sefidgaran Milad
  • Tchamkerten Aslan
  , 2020, pp.1-5. This paper investigates the classical modulo two sum problem in source coding, but with a common observation: a transmitter observes (X,Z), the other transmitter observes (Y,Z), and the receiver wants to compute X ⊕Y without error. Through a coupling argument, this paper establishes a new lower bound on the sum-rate when X -Z -Y forms a Markov chain. (10.1109/ITW46852.2021.9457672)
  DOI : 10.1109/ITW46852.2021.9457672
• On the Capacity of the Continuous-Space SSFM Model of Optical Fiber
  
  • Sefidgaran Milad
  • Yousefi Mansoor
  , 2021, pp.1-5. The limit of a discrete-time model of the optical fiber described by the split-step Fourier method (SSFM) when the number of segments in distance K tends to infinity is considered. It is shown that if K≥P2/3 and P→∞, where P is the average input power, the capacity of the resulting continuous-space lossless model is lower bounded by 12log2(1+SNR)−12+o(1), where o(1) tends to zero with the signal-to-noise ratio SNR. This implies that at least half of the signal degrees-of-freedom remain asymptotically in this model. (10.1109/ITW46852.2021.9457619)
  DOI : 10.1109/ITW46852.2021.9457619
• Joint Sensing and Communication over Memoryless Broadcast Channels
  
  • Ahmadipour Mehrasa
  • Wigger Michele
  • Kobayashi Mari
  , 2021, pp.1-5. (10.1109/ITW46852.2021.9457571)
  DOI : 10.1109/ITW46852.2021.9457571
• FARF: A Fair and Adaptive Random Forests Classifier
  
  • Zhang Wenbin
  • Bifet Albert
  • Zhang Xiangliang
  • Weiss Jeremy C.
  • Nejdl Wolfgang
  , 2021, 12713, pp.245--256. As Artificial Intelligence (AI) is used in more applications, the need to consider and mitigate biases from the learned models has followed. Most works in developing fair learning algorithms focus on the offline setting. However, in many real-world applications data comes in an online fashion and needs to be processed on the fly. Moreover, in practical application, there is a trade-off between accuracy and fairness that needs to be accounted for, but current methods often have multiple hyper-parameters with non-trivial interaction to achieve fairness. In this paper, we propose a flexible ensemble algorithm for fair decision-making in the more challenging context of evolving online settings. This algorithm, called FARF (Fair and Adaptive Random Forests), is based on using online component classifiers and updating them according to the current distribution, that also accounts for fairness and a single hyper-parameters that alters fairness-accuracy balance. Experiments on real-world discriminated data streams demonstrate the utility of FARF. (10.1007/978-3-030-75765-6_20)
  DOI : 10.1007/978-3-030-75765-6_20
• Approximating Probability Distributions by ReLU Networks
  
  • Mukherjee Manuj
  • Tchamkerten Aslan
  • Yousefi Mansoor
  , 2020, pp.1-5. How many neurons are needed to approximate a target probability distribution using a neural network with a given input distribution and approximation error? This paper examines this question for the case when the input distribution is uniform, and the target distribution belongs to the class of histogram distributions. We obtain a new upper bound on the number of required neurons, which is strictly better than previously existing upper bounds. The key ingredient in this improvement is an efficient construction of the neural nets representing piecewise linear functions. We also obtain a lower bound on the minimum number of neurons needed to approximate the histogram distributions. (10.1109/ITW46852.2021.9457598)
  DOI : 10.1109/ITW46852.2021.9457598
• Random User Activity with Mixed Delay Traffic
  
  • Nikbakht Homa
  • Wigger Michèle
  • Shamai Shitz Shlomo
  , 2021, pp.1-5. This paper analyses the multiplexing gain (MG) achievable over a general interference network with random user activity and random arrival of mixed-delay traffic. The mixeddelay traffic is composed of delay-tolerant traffic and delaysensitive traffic where only the former can benefit from receiver cooperation since the latter is subject to stringent decoding delays. Two setups are considered. In the first setup, each active transmitter always has delay-tolerant data to send and delaysensitive data arrival is random. In the second setup, both delaytolerant and delay-sensitive data arrivals are random, and only one of them is present at any given transmitter. The MG regions of both setups are completely characterized for Wyner's softhandoff network. For Wyner's symmetric linear and hexagonal networks inner bounds on the MG region are presented. (10.1109/ITW46852.2021.9457661)
  DOI : 10.1109/ITW46852.2021.9457661
• Performance benchmarking of state-of-the-art software switches for NFV
  
  • Zhang Tianzhu
  • Linguaglossa Leonardo
  • Giaccone Paolo
  • Iannone Luigi
  • Roberts James
  Computer Networks, Elsevier, 2021, 188, pp.107861. With the ultimate goal of replacing proprietary hardware appliances with Virtual Network Functions (VNFs) implemented in software, Network Function Virtualization (NFV) has gained popularity in the past few years. Software switches are widely employed to route traffic between VNFs and physical Network Interface Cards (NICs). It is thus of paramount importance to compare the performance of different switch designs and architectures. In this paper, we propose a methodology to compare fairly and comprehensively the performance of software switches. We first explore the design spaces of 7 state-of-the-art software switches and then compare their performance under four representative test scenarios. Each scenario corresponds to a specific case of routing NFV traffic between NICs and/or VNFs. In our experiments, we evaluate the throughput and latency between VNFs in two of the most popular virtualization environments, namely virtual machines (VMs) and containers. Our experimental results show that no single software switch prevails in all scenarios. It is, therefore, crucial to choose the most suitable solution for the given use case. At the same time, the presented results and analysis provide a more in-depth insight into the design tradeoffs and identify potential performance bottlenecks that could inspire new designs. (10.1016/j.comnet.2021.107861)
  DOI : 10.1016/j.comnet.2021.107861
• Binding data mining and expert knowledge for one-day-ahead prediction of hourly global solar radiation
  
  • Campo-Ávila José Del
  • Takilalte Abdelatif
  • Bifet Albert
  • López Llanos Mora
  Expert Systems with Applications, Elsevier, 2021, 167, pp.114147. AbstractA new methodology to predict one-day-ahead hourly solar global radiation is proposed in this paper. This information is very useful to address many real problems; for instance, energy-market decision making is one of the contexts where that information is essential to ensure the correct integration of grid-connected photovoltaic solar systems. The developed methodology is based on the contribution of different experts to obtain improved data-driven models when included in the data mining process. The modelling phase, when models are induced and new patterns can be identified, is the one that most benefits from that expert knowledge. In this case, it is achieved by combining clustering, regression and classification methods that exploit meteorological data (directly measured or predicted by weather services). The developed models have been embedded in a prediction system that offers reliable forecasts on next-day hourly global solar radiation. As a result of the automatic learning process including the knowledge of different experts, 14 different types of day were identified based on the shape of hourly solar radiation throughout a day. The conventional definitions of types of days, that usually consider 4 options, are updated with this new proposal. The next-day prediction of hourly global radiation is obtained in two phases: in the first one, the next-day type is obtained from among the 14 possible types of day; in the second one, values of hourly global radiation are obtained using the centroid of the predicted type of day and extraterrestrial solar radiation. The relative root mean square error of the prediction model is less than 20%, meaning a significant reduction compared to previous models. Moreover, the proposed models can be recognized in the context of eXplainable Artificial Intelligence. (10.1016/J.ESWA.2020.114147)
  DOI : 10.1016/J.ESWA.2020.114147
• Quand la trigonométrie saute aux yeux
  
  • Zayana Karim
  • Rabiet Victor
  CultureMath, ENS, 2021.
• Actes de la conférence CAID 2020
  
  • de Vieilleville François
  • May Stéphane
  • Lagrange Adrien
  • Dupuis A
  • Ruiloba Rosa
  • Ngolè Mboula Fred
  • Bitard-Feildel Tristan
  • Nogues Erwan
  • Larroche Corentin
  • Mazel Johan
  • Clémençon Stephan
  • Burgot Romain
  • Gaurier Alric
  • Hulot Louis
  • Isaac-Dognin Léo
  • Leichtnam Laetitia
  • Totel Eric
  • Prigent Nicolas
  • Mé Ludovic
  • Bernhard Rémi
  • Moëllic Pierre-Alain
  • Dutertre Jean-Max
  • Kapusta Katarzyna
  • Thouvenot Vincent
  • Bettan Olivier
  • Charrier Tristan
  • Bonnafoux Luc
  • Puig Francisco-Pierre
  • Lhoest Quentin
  • Renault Thomas
  • Benamira Adrien
  • Bonnet Benoit
  • Furon Teddy
  • Bas Patrick
  • Farcy Benjamin
  • Gil-Casals Silvia
  • Mattioli Juliette
  • Fiammante Marc
  • Lambert Marc
  • Bresson Roman
  • Cohen Johanne
  • Hullermeier Eyke
  • Labreuche Christophe
  • Sebag Michele
  • Thebaud Thomas
  • Larcher Anthony
  • Le Lan Gaël
  • Nour Nouredine
  • Belhaj-Soullami Reda
  • Buron Cédric L.R.
  • Peres Alain
  • Barbaresco Frédéric
  • D’acremont Antoine
  • Quin Guillaume
  • Baussard Alexandre
  • Fablet Ronan
  • Corbineau Marie-Caroline
  • Morge-Rollet Louis
  • Le Roy Frederic
  • Le Jeune Denis
  • Gautier Roland
  • Camus Benjamin
  • Monteux Eric
  • Vermet Mikaël
  • Goupilleau Alex
  • Ceillier Tugdual
  , 2021.
• Cache Updating Strategy Minimizing the Age of Information with Time-Varying Files' Popularities
  
  • Tang Haoyue
  • Ciblat Philippe
  • Wang Jintao
  • Wigger Michèle
  • Yates Roy D
  , 2021. We consider updating strategies for a local cache which downloads time-sensitive files from a remote server through a bandwidth-constrained link. The files are requested randomly from the cache by local users according to a popularity distribution which varies over time according to a Markov chain structure. We measure the freshness of the requested timesensitive files through their Age of Information (AoI). The goal is then to minimize the average AoI of all requested files by appropriately designing the local cache's downloading strategy. To achieve this goal, the original problem is relaxed and cast into a Constrained Markov Decision Problem (CMDP), which we solve using a Lagrangian approach and Linear Programming. Inspired by this solution for the relaxed problem, we propose a practical cache updating strategy that meets all the constraints of the original problem. Under certain assumptions, the practical updating strategy is shown to be optimal for the original problem in the asymptotic regime of a large number of files. For a finite number of files, we show the gain of our practical updating strategy over the traditional square-root-law strategy (which is optimal for fixed non time-varying file popularities) through numerical simulations.
• Bayesian Allocation Model: Marginal Likelihood-Based Model Selection for Count Tensors
  
  • Yldrm Sinan
  • Kurutmaz M. Burak
  • Barsbey Melih
  • Şimşekli Umut
  • Cemgil A. Taylan
  IEEE Journal of Selected Topics in Signal Processing, IEEE, 2021, 15 (3), pp.560-573. (10.1109/JSTSP.2020.3045297)
  DOI : 10.1109/JSTSP.2020.3045297
• Editorial: Game Theory for Networks
  
  • Song Ju Bin
  • Li Husheng
  • Coupechoux Marceau
  Mobile Networks and Applications, Springer Verlag, 2021, 26 (2), pp.489-490. (10.1007/s11036-019-01264-0)
  DOI : 10.1007/s11036-019-01264-0
• Automated brain MRI metrics in the EPIRMEX cohort of preterm newborns: Correlation with the neurodevelopmental outcome at 2 years
  
  • Morel Baptiste
  • Bertault Pierre
  • Favrais Géraldine
  • Tavernier Elsa
  • Tosello Barthelemy
  • Bednarek Nathalie
  • Barantin Laurent
  • Chadie Alexandra
  • Proisy Maia
  • Xu Yongchao
  • Bloch Isabelle
  • Sirinelli Dominique
  • Adamsbaum Catherine
  • Tauber Clovis
  • Saliba Elie
  Diagnostic and Interventional Imaging, Elsevier, 2021, 102 (4), pp.225-232. Purpose. The purpose of this study was to identify in the EPIRMEX cohort the correlations between MRI brain metrics, including diffuse excessive high signal intensities (DEHSI) obtained with an automated quantitative method and neurodevelopmental outcomes at 2 years. Materials and methods. A total of 390 very preterm infants (gestational age at birth ≤ 32 weeks) who underwent brain MRI at term equivalent age at 1.5T (n=338) or 3T (n=52) were prospectively included. Using a validated algorithm, automated metrics of the main brain surfaces (cortical and deep gray matter, white matter, cerebrospinal fluid) and DEHSI with three thresholds were obtained. Linear adjust regressions were performed to assess the correlation between brain metrics with the ages and stages questionnaire (ASQ) score at 2 years. Results. Basal ganglia and thalami, cortex and white matter surfaces positively and significantly correlated with the global ASQ score. For all ASQ sub-domains, basal ganglia and thalami surfaces significantly correlated with the scores. DEHSI was present in 289 premature newborns (74%) without any correlation with the ASQ score. Metrics of DEHSI were greater at 3T than at 1.5T. Conclusion. Brain MRI metrics obtained in our multicentric cohort correlate with the neurodevelopmental outcome at 2 years of age. The quantitative detection of DEHSI is not predictive of adverse outcomes. Our automated algorithm might easily provide useful predictive information in daily practice. (10.1016/j.diii.2020.10.009)
  DOI : 10.1016/j.diii.2020.10.009