Search

Co-administration of two or more drugs simultaneously can result in adverse drug reactions. Identifying drug-drug interactions (DDIs) is necessary, especially for drug development and for repurposing old drugs. DDI prediction can be viewed as a matrix completion task, for which matrix factorization (MF) appears as a suitable solution. This paper presents a novel Graph Regularized Probabilistic Matrix Factorization (GRPMF) method, which incorporates expert knowledge through a novel graph-based regularization strategy within an MF framework. An efficient and sounded optimization algorithm is proposed to solve the resulting non-convex problem in an alternating fashion. The performance of the proposed method is evaluated through the DrugBank dataset, and comparisons are provided against state-of-the-art techniques. The results demonstrate the superior performance of GRPMF when compared to its counterparts.

International audience; Denoising, detrending, deconvolution: usual restoration tasks, traditionally decoupled. Coupled formulations entail complex ill-posed inverse problems. We propose PENDANTSS for joint trend removal and blind deconvolution of sparse peak-like signals. It blends a parsimonious prior with the hypothesis that smooth trend and noise can somewhat be separated by low-pass filtering. We combine the generalized quasi-norm ratio SOOT/SPOQ sparse penalties $\ell_p/\ell_q$ with the BEADS ternary assisted source separation algorithm. This results in a both convergent and efficient tool, with a novel Trust-Region block alternating variable metric forward-backward approach. It outperforms comparable methods, when applied to typically peaked analytical chemistry signals. Reproducible code is provided.

International audience; Current solutions for bacteria and viruses identification are based on time-consuming technics with complex preparation procedures. In the present work, we revealed label-free the presence of free viral particles and bacteria with a computational two-photon fluorescence (C-TPF) strategy. Six bacteria were tested: Escherichia coli, Staphylococcus epidermidis, Proteus vulgaris, Pseudomonas fluorescens, Bacillus subtilis and Clostridium perfringens. The two families of viral particles were the herpes virus with the cytomegalovirus (CMV, 300 nm of diameter) and the coronavirus with the SARS-CoV-2 (100 nm of diameter). The instrumental and computational pipeline FAMOUS optimized the produced 3D images. The origin of the fluorescence emission was discussed for bacteria regarding to their two-photon excitation spectra and attributed to the metabolic indicators (FAD and NADH). The optical and computational strategy constitute a new approach for imaging label-free viral particles and bacteria and paves the way to a new understanding of viral or bacterial ways of infection.

In this paper, we introduce a variational Bayesian algorithm (VBA) for image blind deconvolution. Our generic framework incorporates smoothness priors on the unknown blur/image and possible affine constraints (e.g., sum to one) on the blur kernel. One of our main contributions is the integration of VBA within a neural network paradigm, following an unrolling methodology. The proposed architecture is trained in a supervised fashion, which allows us to optimally set two key hyperparameters of the VBA model and lead to further improvements in terms of resulting visual quality. Various experiments involving grayscale/color images and diverse kernel shapes, are performed. The numerical examples illustrate the high performance of our approach when compared to state-of-the-art techniques based on optimization, Bayesian estimation, or deep learning., Comment: 13 pages

International audience; A wide class of problems involves the minimization of a coercive and differentiable function F on R N whose gradient cannot be evaluated in an exact manner. In such context, many existing convergence results from standard gradientbased optimization literature cannot be directly applied and robustness to errors in the gradient is not necessarily guaranteed. This work is dedicated to investigating the convergence of Majorization-Minimization (MM) schemes when stochastic errors affect the gradient terms. We introduce a general stochastic optimization framework, called SABRINA (StochAstic suBspace majoRIzation-miNimization Algorithm) that encompasses MM quadratic schemes possibly enhanced with a subspace acceleration strategy. New asymptotical results are built for the stochastic process generated by SABRINA. Two sets of numerical experiments in the field of machine learning and image processing are presented to support our theoretical results and illustrate the good performance of SABRINA with respect to state-of-the-art gradient-based stochastic optimization methods.

International audience; This work addresses the problem of hyperspectral image classification when the number of labeled samples is very small (few shot learning). Our work is based on the recently proposed framework of convolutional transform learning. In this work, we propose a semisupervised version of deep convolutional transform learning. We compare with four recent studies which are tailored for solving the few-shot learning problem in hyperspectral classification. Results show that our method can improve over the state-of-the-art.

International audience; Conventional Convolutional Neural Network (CNN) based clustering formulations are based on the encoder-decoder based framework, where the clustering loss is incorporated after the encoder network. The problem with this approach is that it requires training an additional decoder network; this, in turn, means learning additional weights which can lead to over-fitting in data constrained scenarios. This work introduces a Deep Convolutional Transform Learning (DCTL) based clustering framework. The advantage of our proposed formulation is that we do not require learning the additional decoder network. Therefore our formulation is less prone to over-fitting. Comparison with state-of-the-art deep learning based clustering solutions on benchmark image datasets shows that our proposed method improves over the rest in challenging scenarios where there are many clusters with limited samples.

International audience; State-space models (SSM) are central to describe time-varying complex systems in countless signal processing applications such as remote sensing, networks, biomedicine, and finance to name a few. Inference and prediction in SSMs are possible when the model parameters are known, which is rarely the case. The estimation of these parameters is crucial, not only for performing statistical analysis, but also for uncovering the underlying structure of complex phenomena. In this paper, we focus on the linear-Gaussian model, arguably the most celebrated SSM, and particularly in the challenging task of estimating the transition matrix that encodes the Markovian dependencies in the evolution of the multi-variate state. We introduce a novel perspective by relating this matrix to the adjacency matrix of a directed graph, also interpreted as the causal relationship among state dimensions in the Granger-causality sense. Under this perspective, we propose a new method called GraphEM based on the well sounded expectation-maximization (EM) methodology for inferring the transition matrix jointly with the smoothing/filtering of the observed data. We propose an advanced convex optimization solver relying on a consensus-based implementation of a proximal splitting strategy for solving the M-step. This approach enables an efficient and versatile processing of various sophisticated priors on the graph structure, such as parsimony constraints, while benefiting from convergence guarantees. We demonstrate the good performance and the interpretable results of GraphEM by means of two sets of numerical examples.

International audience; Optimization problems arising in signal and image processing involve an increasingly large number of variables. In addition to the curse of dimensionality, another difficulty to overcome is that the cost function usually reads as the sum of several loss/regularization terms, which are non-necessarily smooth and possibly composed with large-size linear operators. Proximal splitting approaches are fundamental tools to address such problems, with demonstrated efficiency in many applicative fields. In this paper, we present a new distributed algorithm for computing the proximity operator of a sum of non-necessarily smooth convex functions composed with arbitrary linear operators. Our algorithm relies on a primal-dual splitting strategy, and benefits from established convergence guaranties. Each involved function is associated with a node of a hypergraph, with the ability to communicate with neighboring nodes sharing the same hyperedge. Thanks to this structure, our method can be efficiently implemented on modern parallel computing architectures, distributing the computations on multiple nodes or machines, with controlled requirements for synchronization steps. Good numerical performance and scalability properties are demonstrated on a problem of video sequence denoising. Our code implemented in Julia is made available at \url{https://github.com/MarinENSTA/distributed_julia_denoising}.

International audience; In interventional radiology, 3D reconstruction of devices such as needles can increase the precision of procedures. Doing so with CBCT is time-consuming and increases the X-ray dose. Needles being sparse, a compressed-sensing reconstruction approach seems viable. In this paper, we thus investigate the benefits of directional total variation as an adequate prior for anisotropic devices. We introduce a decomposition method that allows several a priori directions to be considered at once as well as excludes the anatomical background that is not sparse. The capacity of the method is illustrated in simulations of limited-angle acquisitions. It allows good reconstruction of the needles from a small angular coverage, even if it cannot recover the anatomical background.

The SARS-COV-2 pandemic has put pressure on intensive care units, so that identifying predictors of disease severity is a priority. We collect 58 clinical and biological variables, and chest CT scan data, from 1003 coronavirus-infected patients from two French hospitals. We train a deep learning model based on CT scans to predict severity. We then construct the multimodal AI-severity score that includes 5 clinical and biological variables (age, sex, oxygenation, urea, platelet) in addition to the deep learning model. We show that neural network analysis of CT-scans brings unique prognosis information, although it is correlated with other markers of severity (oxygenation, LDH, and CRP) explaining the measurable but limited 0.03 increase of AUC obtained when adding CT-scan information to clinical variables. Here, we show that when comparing AI-severity with 11 existing severity scores, we find significantly improved prognosis performance; AI-severity can therefore rapidly become a reference scoring approach., The SARS-COV-2 pandemic has put pressure on intensive care units, so that predicting severe deterioration early is a priority. Here, the authors develop a multimodal severity score including clinical and imaging features that has significantly improved prognostic performance in two validation datasets compared to previous scores.

International audience; This article introduces a new Penalized Majorization-Minimization Subspace algorithm (P-MMS) for solving smooth, constrained optimization problems. In short, our approach consists of embedding a subspace algorithm in an inexact exterior penalty procedure. The subspace strategy, combined with a Majoration-Minimization step-size search, takes great advantage of the smoothness of the penalized cost function, while the penalty method allows to handle a wide range of constraints. The main drawback of exterior penalty approaches, namely ill-conditioning for large values of the penalty parameter, is overcome by using a trust-regionlike technique. The convergence of the resulting algorithm is analyzed. Numerical experiments carried out on two large-scale image recovery applications demonstrate that, compared with state-of-the-art algorithms, the proposed method performs well in terms of computational time.

Most optimization problems arising in imaging science involve high-dimensional linear operators and their adjoints. In the implementations of these operators, approximations may be introduced for various practical considerations (e.g., memory limitation, computational cost, convergence speed), leading to an adjoint mismatch. This occurs for the X-ray tomographic inverse problems found in Computed Tomography (CT), where the adjoint of the measurement operator (called projector) is often replaced by a surrogate operator. The resulting adjoint mismatch can jeopardize the convergence properties of iterative schemes used for image recovery. In this paper, we study the theoretical behavior of a panel of primal-dual proximal algorithms, which rely on forward-backward-(forward) splitting schemes, when an adjoint mismatch occurs. We analyze these algorithms by focusing on the resolution of possibly non-smooth convex penalized minimization problems in an infinite-dimensional setting. By using tools from fixed point theory, we show that they can solve monotone inclusions that go beyond minimization problems. Such findings indicate these algorithms can be seen as a generalization of classical primal-dual formulations. The applicability of our findings are also demonstrated through two numerical experiments in the context of CT image reconstruction.

International audience; In a medical use, ultrastructure of muscle is currently revealed by images of resected samples achieved thanks to Electron Microscopy (EM), requiring freezing and paraffin sections, with a set of histological, molecular and biochemical analyses. The resection, slicing and labelling steps cause an alteration of the phenotypic and volumetric information compared to their initial integrity. Starting from this statement, we have developed an original pipeline resting on the imbrication of an optical and computational strategy for imaging 3D biomedical structures without resorting to slicing, freezing and labelling steps. The assembly of myosin of a whole muscle is revealed thanks to the second harmonic generation (SHG) recorded with a multiphoton microscope, all along the entire 180 µm of thickness of the Extensor digitorum longus (EDL) of a wild mouse. During the SHG recording, the Point-Spread-Function (PSF) of the multiphoton microscope is recorded all along the imaging depth. This step highlights an axial broadening of the PSF while maintaining a constant planar PSF throughout the whole depth of the recordings. Then, a fitting algorithm estimated a mathematical model of the PSF, highlighting its variability into the whole image depth. Finally, a computational image restauration is led thanks to the fast image deblurring algorithm BD3MG accounting for the depth variant PSF. The non-stationary distortions all along the recorded image is employed thus correcting accurately the image distortion. The axial organization of the myosin is revealed for the first time, highlighting tubular organization of myosin into the myofibrils.

International audience; In Drug-Drug-Interaction (DDI), the task is to predict the (adverse) effect of administering two drugs simultaneously. Currently, the techniques proposed in this direction are generally based on either shallow learning paradigms like Random Decision Forest (RDF), Logistic Regression (LR), Support Vector Machines (SVM), etc., or deep Convolutional Neural Networks (CNNs). However, specific works combine traditional machine learning (ML) algorithms such as RDF, LR, SVM, and deep learning paradigms such as CNNs in a piecemeal fashion which might not be optimal. Hence, the present work proposes a framework that presents a joint end-to-end solution. We propose a Siamese-like architecture with two processing channels' networks based on deep convolutional transform learning. Common fused representations as well as channel-wise representations are learnt, in addition with the transform across them. The final representation is passed to a decision forest to give final predictions. The proposed method is thus a supervised end-to-end multi-channel fusion framework that (i) learns unique and interpretable filters in contrast with CNNs, and (ii) jointly learns and optimizes decision forest in contrast with state-of-the-art piecemeal approach. We apply this technique to identify DDIs among 1059 drugs from the DrugBank database showing superiority of our method compared to the state-of-the-art(s).

Gliomas are among the most common types of central nervous system (CNS) tumors. A prompt diagnosis of the glioma subtype is crucial to estimate the prognosis and personalize the treatment strategy. The objective of this study was to develop a radiomics pipeline based on the clinical Magnetic Resonance Imaging (MRI) scans to noninvasively predict the glioma subtype, as defined based on the tumor grade, isocitrate dehydrogenase (IDH) mutation status, and 1p/19q codeletion status. A total of 212 patients from the public retrospective The Cancer Genome Atlas Low Grade Glioma (TCGA-LGG) and The Cancer Genome Atlas Glioblastoma Multiforme (TCGA-GBM) datasets were used for the experiments and analyses. Different settings in the radiomics pipeline were investigated to improve the classification, including the Z-score normalization, the feature extraction strategy, the image filter applied to the MRI images, the introduction of clinical information, ComBat harmonization, the classifier chain strategy, etc. Based on numerous experiments, we finally reached an optimal pipeline for classifying the glioma tumors. We then tested this final radiomics pipeline on the hold-out test data with 51 randomly sampled random seeds for reliable and robust conclusions. The results showed that, after tuning the radiomics pipeline, the mean AUC improved from 0.8935 (±0.0351) to 0.9319 (±0.0386), from 0.8676 (±0.0421) to 0.9283 (±0.0333), and from 0.6473 (±0.1074) to 0.8196 (±0.0702) in the test data for predicting the tumor grade, IDH mutation, and 1p/19q codeletion status, respectively. The mean accuracy for predicting the five glioma subtypes also improved from 0.5772 (±0.0816) to 0.6716 (±0.0655). Finally, we analyzed the characteristics of the radiomic features that best distinguished the glioma grade, the IDH mutation, and the 1p/19q codeletion status, respectively. Apart from the promising prediction of the glioma subtype, this study also provides a better understanding of the radiomics model development and interpretability. The results in this paper are replicable with our python codes publicly available in github.

This paper addresses the problem of image reconstruction for region-of-interest (ROI) computed tomography (CT). While model-based iterative methods can be used for such a problem, their practicability is often limited due to tedious parameterization and slow convergence. In addition, inadequate solutions can be obtained when the retained priors do not perfectly fit the solution space. Deep learning methods offer an alternative approach that is fast, leverages information from large data sets, and thus can reach high reconstruction quality. However, these methods usually rely on black boxes not accounting for the physics of the imaging system, and their lack of interpretability is often deplored. At the crossroads of both methods, unfolded deep learning techniques have been recently proposed. They incorporate the physics of the model and iterative optimization algorithms into a neural network design, leading to superior performance in various applications. This paper introduces a novel, unfolded deep learning approach called U-RDBFB designed for ROI CT reconstruction from limited data. Few-view truncated data are efficiently handled thanks to a robust non-convex data fidelity function combined with sparsityinducing regularization functions. Iterations of a block dual forward-backward (DBFB) algorithm, embedded in an iterative reweighted scheme, are then unrolled over a neural network architecture, allowing the learning of various parameters in a supervised manner. Our experiments show an improvement over various state-of-the-art methods, including model-based iterative schemes, deep learning architectures, and deep unfolding methods.

PURPOSE Discretizing tomographic forward and backward operations is a crucial step in the design of model-based reconstruction algorithms. Standard projectors rely on linear interpolation, whose adjoint introduces discretization errors during backprojection. More advanced techniques are obtained through geometric footprint models that may present a high computational cost and an inner logic that is not suitable for implementation on massively parallel computing architectures. In this work, we take a fresh look at the discretization of resampling transforms and focus on the issue of magnification-induced local sampling variations by introducing a new magnification-driven interpolation approach for tomography. METHODS Starting from the existing literature on spline interpolation for magnification purposes, we provide a mathematical formulation for discretizing a one-dimensional homography. We then extend our approach to two-dimensional representations in order to account for the geometry of cone-beam computed tomography with a flat panel detector. Our new method relies on the decomposition of signals onto a space generated by nonuniform B-splines so as to capture the spatially varying magnification that locally affects sampling. We propose various degrees of approximations for a rapid implementation of the proposed approach. Our framework allows us to define a novel family of projector/backprojector pairs parameterized by the order of the employed B-splines. The state-of-the-art distance-driven interpolation appears to fit into this family thus providing new insight and computational layout for this scheme. The question of data resampling at the detector level is handled and integrated with reconstruction in a single framework. RESULTS Results on both synthetic data and real data using a quality assurance phantom, were performed to validate our approach. We show experimentally that our approximate implementations are associated with reduced complexity while achieving a near-optimal performance. In contrast with linear interpolation, B-splines guarantee full usage of all data samples, and thus the X-ray dose, leading to more uniform noise properties. In addition, higher-order B-splines allow analytical and iterative reconstruction to reach higher resolution. These benefits appear more significant when downsampling frames acquired by X-ray flat-panel detectors with small pixels. CONCLUSIONS Magnification-driven B-spline interpolation is shown to provide high-accuracy projection operators with good-quality adjoints for iterative reconstruction. It equally applies to backprojection for analytical reconstruction and detector data downsampling.

International audience; The objective of this letter is to propose a novel computational method to learn the state of an appliance (ON / OFF) given the aggregate power consumption recorded by the smart-meter. We formulate a multi-label classification problem where the classes correspond to the appliances. The proposed approach is based on our recently introduced framework of convolutional transform learning. We propose a deep supervised version of it relying on an original multi-label cost. Comparisons with state-of-the-art techniques show that our proposed method improves over the benchmarks on popular non-intrusive load monitoring datasets.

International audience; In this work, we introduce a new modeling and inferential tool for dynamical processing of time series. The approach is called recurrent dictionary learning (RDL). The proposed model reads as a linear Gaussian Markovian state-space model involving two linear operators, the state evolution and the observation matrices, that we assumed to be unknown. These two unknown operators (that can be seen interpreted as dictionaries) and the sequence of hidden states are jointly learnt via an expectation-maximization algorithm. The RDL model gathers several advantages, namely online processing, probabilistic inference, and a high model expressiveness which is usually typical of neural networks. RDL is particularly well suited for stock forecasting. Its performance is illustrated on two problems: next day forecasting (regression problem) and next day trading (classification problem), given past stock market observations. Experimental results show that our proposed method excels over state-of-the-art stock analysis models such as CNN-TA, MFNN, and LSTM.

International audience; Restoration of analytical chemistry data from degraded physical acquisitions is an important task for chemists to obtain accurate component analysis and sound interpretation. The high-dimensional nature of these signals and the large amount of data to be processed call for fast and efficient reconstruction methods. Existing works have primarily relied on optimization algorithms to solve a penalized formulation. Although very powerful, such methods can be computationally heavy, and hyperparameter tuning can be a tedious task for non-experts. Another family of approaches explored recently consists in adopting deep learning to perform the signal recovery task in a supervised fashion. Although fast, thanks to their formulations amenable to GPU implementations, these methods usually need large annotated databases and are not explainable. In this work, we propose to combine the best of both worlds, by proposing unfolded Majorization-Minimization (MM) algorithms with the aim to reach fast and accurate methods for sparse spectroscopy signal restoration. Two state-of-the-art iterative MM algorithms are unfolded onto deep network architectures. This allows both the deployment of GPU-friendly tools for accelerated implementation, as well as the introduction of a supervised learning strategy for tuning automatically the regularization parameter. The effectiveness of our approach is demonstrated on the restoration of a large dataset of realistic mass spectrometry data.

Curbing hate speech is undoubtedly a major challenge for onlinemicroblogging platforms like Twitter. While there have been studiesaround hate speech detection, it is not clear how hate speech findsits way into an online discussion. It is important for a contentmoderator to not only identify which tweet is hateful, but alsoto predict which tweet will be responsible for accumulating hatespeech. This would help in prioritizing tweets that need constantmonitoring. Our analysis reveals that for hate speech to manifestin an ongoing discussion, the source tweet may not necessarily behateful; rather, there are plenty of such non-hateful tweets whichgradually invoke hateful replies, resulting in the entire reply threadsbecoming provocative.In this paper,we define a novel problem – given a source tweet anda few of its initial replies, the task is to forecast the hate intensity ofupcoming replies. To this end, we curate a novel dataset constituting∼ 4.5𝑘 contemporary tweets and their entire reply threads. Our preliminaryanalysis confirms that the evolution patterns along timeof hate intensity among reply threads have highly diverse patterns,and there is no significant correlation between the hate intensity ofthe source tweets and that of their reply threads. We employ sevenstate-of-the-art dynamic models (either statistical signal processingor deep learning based) and show that they fail badly to forecast thehate intensity. We then propose DESSERT, a novel deep state-spacemodel that leverages the function approximation capability of deepneural networks with the capacity to quantify the uncertainty ofstatistical signal processing models. Exhaustive experiments andablation study show that DESSERT outperforms all the baselinessubstantially. Further, its deployment in an advanced AI platformdesigned to monitor real-world problematic hateful content has improvedthe aggregated insights extracted for countering the spreadof online harms.

International audience; Finding the global minimum of a nonconvex optimization problem is a notoriously hard task appearing in numerous applications, from signal processing to machine learning. Simulated annealing (SA) is a family of stochastic optimization methods where an artificial temperature controls the exploration of the search space while preserving convergence to the global minima. SA is efficient, easy to implement, and theoretically sound, but suffers from a slow convergence rate. The purpose of this work is twofold. First, we provide a comprehensive overview on SA and its accelerated variants. Second, we propose a novel SA scheme called curious simulated annealing, combining the assets of two recent acceleration strategies. Theoretical guarantees of this algorithm are provided. Its performance with respect to existing methods is illustrated on practical examples.

Based on its great successes in inference and denosing tasks, Dictionary Learning (DL) and its related sparse optimization formulations have garnered a lot of research interest. While most solutions have focused on single layer dictionaries, the recently improved Deep DL methods have also fallen short on a number of issues. We hence propose a novel Deep DL approach where each DL layer can be formulated and solved as a combination of one linear layer and a Recurrent Neural Network, where the RNN is flexibly regraded as a layer-associated learned metric. Our proposed work unveils new insights between the Neural Networks and Deep DL, and provides a novel, efficient and competitive approach to jointly learn the deep transforms and metrics. Extensive experiments are carried out to demonstrate that the proposed method can not only outperform existing Deep DL, but also state-of-the-art generic Convolutional Neural Networks., Accepted by ICASSP 2021. arXiv admin note: substantial text overlap with arXiv:2002.07898

Simple Summary As a rapid-development research field, radiomics-based analysis has been applied to many clinical problems. However, the reproducibility of the radiomics studies remain challenging especially when data suffers from scanner effects, a kind of non-biological variations introduced by different image acquiring settings. This study aims to investigate how the image preprocessing methods (N4 bias field correction and image resampling) and the harmonization methods (intensity normalization methods working on images and ComBat method working on radiomic features) help to remove the scanner effects and improve the radiomics reproducibility in brain MRI radiomics. Abstract In brain MRI radiomics studies, the non-biological variations introduced by different image acquisition settings, namely scanner effects, affect the reliability and reproducibility of the radiomics results. This paper assesses how the preprocessing methods (including N4 bias field correction and image resampling) and the harmonization methods (either the six intensity normalization methods working on brain MRI images or the ComBat method working on radiomic features) help to remove the scanner effects and improve the radiomic feature reproducibility in brain MRI radiomics. The analyses were based on in vitro datasets (homogeneous and heterogeneous phantom data) and in vivo datasets (brain MRI images collected from healthy volunteers and clinical patients with brain tumors). The results show that the ComBat method is essential and vital to remove scanner effects in brain MRI radiomic studies. Moreover, the intensity normalization methods, while not able to remove scanner effects at the radiomic feature level, still yield more comparable MRI images and improve the robustness of the harmonized features to the choice among ComBat implementations.

International audience; The current coronavirus pandemic (COVID-19) became a worldwide threat, infecting more than 42 million people since its outbreak in early 2020. Recent studies show that analyzing chest CT scans plays an essential role in assessing disease progression and facilitates early diagnosis. Automatic lesion segmentation constitutes a useful tool to complement more traditional healthcare system strategies to address the COVID-19 crisis. We introduce MASC-Net, a novel deep neural network that automatically detects COVID-19 related infected lung regions from chest CT scans. The proposed architecture consists of a multi-input encoder-decoder that aggregates high-level features extracted with variable-size receptive fields.

International audience; Acquiring ultrasound images of suspected lesion areas allows radiologists to monitor the cancer development of patients. The goal of this paper is to provide an automatic lesion segmentation tool for assisting them on the analysis of ultrasound images, by relying on recent neural network methods. Specifically, we perform a comparative study for the segmentation of 348 ultrasound image pairs acquired in 19 centers across France, displaying different tumor types. We show that, with a careful hyperparameter tuning, U-net outperforms other state-of-the-art networks, reaching a Dice coefficient of 0.929. We then propose to introduce group convolution into U-net architecture. This leads to a lightweight network named Lighter U-net @128 that achieves comparable segmentation performance with obviously reduced model size, hence paving the way for an embedded integration within hospital environment. We made our code publicly available for reproducibility purpose.

Over the last decade, the combination of compressed sensing (CS) with acquisition over multiple receiver coils in magnetic resonance imaging (MRI) has allowed the emergence of faster scans while maintaining a good signal-to-noise ratio (SNR). Self-calibrating techniques, such as ESPiRIT, have become the standard approach to estimating the coil sensitivity maps prior to the reconstruction stage. In this work, we proceed differently and introduce a new calibration-less multi-coil CS reconstruction method. Calibration-less techniques no longer require the prior extraction of sensitivity maps to perform multi-coil image reconstruction but usually alternate estimation sensitivity map estimation and image reconstruction. Here, to get rid of the nonconvexity of the latter approach we reconstruct as many MR images as the number of coils. To compensate for the ill-posedness of this inverse problem, we leverage structured sparsity of the multi-coil images in a wavelet transform domain while adapting to variations in SNR across coils owing to the OSCAR (octagonal shrinkage and clustering algorithm for regression) regularization. Coil-specific complex-valued MR images are thus obtained by minimizing a convex but nonsmooth objective function using the proximal primal-dual Condat-Vù algorithm. Comparison and validation on retrospective Cartesian and non-Cartesian studies based on the Brain fastMRI data set demonstrate that the proposed reconstruction method outperforms the state-of-the-art (ℓ1-ESPIRiT, calibration-less AC-LORAKS and CaLM methods) significantly on magnitude images for the T1 and FLAIR contrasts. Additionally, further validation operated on 8 to 20-fold prospectively accelerated high-resolution ex vivo human brain MRI data collected at 7 Tesla confirms the retrospective results. Overall, OSCAR-based regularization preserves phase information more accurately (both visually and quantitatively) compared to other approaches, an asset that can only be assessed on real prospective experiments.

International audience; Adaptive importance sampling (AIS) methods are increasingly used for the approximation of distributions and related intractable integrals in the context of Bayesian inference. Population Monte Carlo (PMC) algorithms are a subclass of AIS methods, widely used due to their ease in the adaptation. In this paper, we propose a novel algorithm that exploits the benefits of the PMC framework and includes more efficient adaptive mechanisms, exploiting geometric information of the target distribution. In particular, the novel algorithm adapts the location and scale parameters of a set of importance densities (proposals). At each iteration, the location parameters are adapted by combining a versatile resampling strategy (i.e., using the information of previous weighted samples) with an advanced optimization-based scheme. Local second-order information of the target distribution is incorporated through a preconditioning matrix acting as a scaling metric onto a gradient direction. A damped Newton approach is adopted to ensure robustness of the scheme. The resulting metric is also used to update the scale parameters of the proposals. We discuss several key theoretical foundations for the proposed approach. Finally, we show the successful performance of the proposed method in three numerical examples, involving challenging distributions.

This work addresses the problem of analyzing multi-channel time series data %. In this paper, we by proposing an unsupervised fusion framework based on %the recently proposed convolutional transform learning. Each channel is processed by a separate 1D convolutional transform; the output of all the channels are fused by a fully connected layer of transform learning. The training procedure takes advantage of the proximal interpretation of activation functions. We apply the developed framework to multi-channel financial data for stock forecasting and trading. We compare our proposed formulation with benchmark deep time series analysis networks. The results show that our method yields considerably better results than those compared against., Comment: Accepted at EUSIPCO 2020

International audience; This work addresses the problem of completing a partially observed matrix where the entries are either ones or zeroes. This is typically called one-bit matrix completion or binary matrix completion. In this problem, the association among the rows and among the columns can be modeled through graph Laplacians. Since the Laplacians cannot be computed from the incomplete matrix, they must be simultaneously estimated while completing the matrix. We model the problem as graph regularized binary matrix completion where the graphs need to be learnt from the data. We proposed an algorithm based on an alternating minimization scheme, taking advantage of an efficient proximity-based inner solver. The algorithm is applied to the problem of collaborative filtering. Experiments on benchmark datasets with state-of-the-art techniques in collaborative filtering show that the proposed method improves over the rest by a considerable margin.

MotivationInvestigation of existing drugs is an effective alternative to discovery of new drugs for treating diseases. This task of drug re-positioning can be assisted by various kinds of computational methods to predict the best indication for a drug given the open-source biological datasets. Owing to the fact that similar drugs tend to have common pathways and disease indications, the association matrix is assumed to be of low-rank structure. Hence, the problem of drug-disease association prediction can been modelled as a low-rank matrix-completion problem.ResultsIn this work, we propose a novel matrix completion framework which makes use of the sideinformation associated with drugs/diseases for the prediction of drug-disease indications modelled as neighborhood graph: Graph regularized 1-bit matrix compeltion (GR1BMC). The algorithm is specially designed for binary data and uses parallel proximal algorithm to solve the aforesaid minimization problem taking into account all the constraints including the neighborhood graph incorporation and restricting predicted scores within the specified range. The results of the proposed algorithm have been validated on two standard drug-disease association databases (Fdataset and Cdataset) by evaluating the AUC across the 10-fold cross validation splits. The usage of the method is also evaluated through a case study where top 5 indications are predicted for novel drugs and diseases, which then are verified with the CTD database. The results of these experiments demonstrate the practical usage and superiority of the proposed approach over the benchmark methods.Contactaanchalm@iiitd.ac.in

International audience; Anti-angiogenic therapy with bevacizumab is a widely used therapeutic option for recurrent glioblastoma (GBM). Nevertheless, the therapeutic response remains highly heterogeneous among GBM patients with discordant outcomes. Recent data have shown that radiomics, an advanced recent imaging analysis method, can help to predict both prognosis and therapy in a multitude of solid tumours. The objective of this study was to identify novel biomarkers, extracted from MRI and clinical data, which could predict overall survival (OS) and progression-free survival (PFS) in GBM patients treated with bevacizumab using machine-learning algorithms. In a cohort of 194 recurrent GBM patients (age range 18–80), radiomics data from pre-treatment T2 FLAIR and gadolinium-injected MRI images along with clinical features were analysed. Binary classification models for OS at 9, 12, and 15 months were evaluated. Our classification models successfully stratified the OS. The AUCs were equal to 0.78, 0.85, and 0.76 on the test sets (0.79, 0.82, and 0.87 on the training sets) for the 9-, 12-, and 15-month endpoints, respectively. Regressions yielded a C-index of 0.64 (0.74) for OS and 0.57 (0.69) for PFS. These results suggest that radiomics could assist in the elaboration of a predictive model for treatment selection in recurrent GBM patients.

International audience; We consider the proximal gradient algorithm for solving penalized least-squares minimization problems arising in data science. This first-order algorithm is attractive due to its flexibility and minimal memory requirements allowing to tackle large-scale minimization problems involving non-smooth penalties. However, for problems such as X-ray computed tomography, the applicability of the algorithm is dominated by the cost of applying the forward linear operator and its adjoint at each iteration. In practice, the adjoint operator is thus often replaced by an alternative operator with the aim to reduce the overall computation burden and potentially improve conditioning issues. In this paper, we propose to analyze the effect of such an adjoint mismatch on the convergence of the proximal gradient algorithm in an infinite-dimensional setting, thus generalizing the existing results on PGA. We derive conditions on the stepsize and on the gradient of the smooth part of the objective function under which convergence of the algorithm to a fixed point is guaranteed. We also derive bounds on the error between this point and the solution to the original minimization problem. We illustrate our theoretical findings with two image reconstruction tasks in computed tomography.

International audience; This work introduces a new unsupervised representation learning technique called Deep Convolutional Transform Learning (DCTL). By stacking convolutional transforms, our approach is able to learn a set of independent kernels at different layers. The features extracted in an unsupervised manner can then be used to perform machine learning tasks, such as classification and clustering. The learning technique relies on a well-sounded alternating proximal minimization scheme with established convergence guarantees. Our experimental results show that the proposed DCTL technique outperforms its shallow version CTL, on several benchmark datasets.

International audience; Subspace clustering assumes that the data is separable into separate subspaces; this assumption may not always hold. For such cases, we assume that, even if the raw data is not separable into subspaces, one can learn a deep representation such that the learnt representation is separable into subspaces. To achieve the intended goal, we embed subspace clustering techniques (locally linear manifold clustering, sparse subspace clustering and low rank representation) into deep transform learning. The entire formulation is jointly learnt; giving rise to a new class of methods called deeply transformed subspace clustering (DTSC). To test the performance of the proposed techniques, benchmarking is performed on image clustering problems. Comparison with state-of-the-art clustering techniques shows that our formulation improves upon them.

Three-dimensional (3D), in vivo and in live imaging of living samples with a sub-micrometer resolution is a current necessity for biomedical researches which corresponds to a hot topic for microscopy engineers. It is likely that the role of computational approaches is yet underestimated and underused in 3D-microscopy. This paper is an illustration of the fundamental role that could be played by such strategies. Usual imaging depths in biomedical microscopy reach few hundreds of micrometers in optimal conditions when multiphoton approaches are favored. However, light scattering and absorption still damage image quality, especially as imaging depth increases. Our approach rests on the multi-parametric 3D-PSF estimation of the true 3DGaussian model of the PSF along multi-millimeter depth. Image acquisition in MPM generates stacks of 2D images constituting an overall 3D-image. 3D-volumes of images containing a single object are selected and isolated all along the depth using automatic morphological tools. Finally, our computational 3DGaussian shape-fitting algorithm named FIGARO is applied on each individual PSF and quantify PSF full-width at half maximum (FWHM) in the 3 dimensions simultaneously with PSF tilt angles. FWHM evolution in the 3 dimensions along the 2 mm depth highlights a highly significant effect of spherical aberrations. Starting from standard values of PSF measured at sample top compared to what expected in MPM, we show an increase with a factor three of the PSF FWHM in axial plane when the sample bottom is reached, and no modification of PSF dimensions in lateral plane all along the depth. The main direction of the PSF shows a convergence toward a focal point that will be discussed

International audience; Modeling and inference with multivariate sequences is central in a number of signal processing applications such as acoustics, social network analysis, biomedical, and finance, to name a few. The linear-Gaussian state-space model is a common way to describe a time series through the evolution of a hidden state, with the advantage of presenting a simple inference procedure due to the celebrated Kalman filter. A fundamental question when analyzing a multivariate sequence is the search for relationships between its entries (or the entries of the modeled hidden state), especially when the inherent structure is a non-fully connected graph. In such context, graphical modeling combined with parsimony constraints allows to limit the proliferation of parameters and enables a compact data representation which is easier to interpret by the experts. In this work, we propose a novel expectation-maximization algorithm for estimating the linear matrix operator in the state equation of a linear-Gaussian state-space model. Lasso regularization is included in the M-step, that we solve using a proximal splitting Douglas-Rachford algorithm. Numerical experiments illustrate the benefits of the proposed model and inference technique, named GraphEM, over competitors relying on Granger causality.

International audience; Subspace clustering assumes that the data is separable into separate subspaces. Such a simple assumption, does not always hold. We assume that, even if the raw data is not separable into subspaces, one can learn a representation (transform coefficients) such that the learnt representation is separable into subspaces. To achieve the intended goal, we embed subspace clustering techniques (locally linear manifold clustering, sparse subspace clustering and low rank representation) into transform learning. The entire formulation is jointly learnt; giving rise to a new class of methods called transformed subspace clustering (TSC). In order to account for non-linearity, kernelized extensions of TSC are also proposed. To test the performance of the proposed techniques, benchmarking is performed on image clustering and document clustering datasets. Comparison with state-of-the-art clustering techniques shows that our formulation improves upon them.

Underdetermined or ill-posed inverse problems require additional information for sound solutions with tractable optimization algorithms. Sparsity yields consequent heuristics to that matter, with numerous applications in signal restoration, image recovery, or machine learning. Since the $\ell _0$ count measure is barely tractable, many statistical or learning approaches have invested in computable proxies, such as the $\ell _1$ norm. However, the latter does not exhibit the desirable property of scale invariance for sparse data. Extending the SOOT Euclidean/Taxicab $\ell _1$ -over- $\ell _2$ norm-ratio initially introduced for blind deconvolution, we propose SPOQ, a family of smoothed (approximately) scale-invariant penalty functions. It consists of a Lipschitz-differentiable surrogate for $\ell _p$ -over- $\ell _q$ quasi-norm/norm ratios with $p\in \,]0,2[$ and $q\geq 2$ . This surrogate is embedded into a novel majorize-minimize trust-region approach, generalizing the variable metric forward-backward algorithm. For naturally sparse mass-spectrometry signals, we show that SPOQ significantly outperforms $\ell _0$ , $\ell _1$ , Cauchy, Welsch, SCAD and $\mathrm{CEL}0$ penalties on several performance measures. Guidelines on SPOQ hyperparameters tuning are also provided, suggesting simple data-driven choices.

International audience; Variational methods are widely applied to ill-posed inverse problems for they have the ability to embed prior knowledge about the solution. However, the level of performance of these methods significantly depends on a set of parameters, which can be estimated through computationally expensive and time-consuming methods. In contrast, deep learning offers very generic and efficient architectures, at the expense of explainability, since it is often used as a black-box, without any fine control over its output. Deep unfolding provides a convenient approach to combine variational-based and deep learning approaches. Starting from a variational formulation for image restoration, we develop iRestNet, a neural network architecture obtained by unfolding a proximal interior point algorithm. Hard constraints, encoding desirable properties for the restored image, are incorporated into the network thanks to a logarithmic barrier, while the barrier parameter, the stepsize, and the penalization weight are learned by the network. We derive explicit expressions for the gradient of the proximity operator for various choices of constraints, which allows training iRestNet with gradient descent and backpropagation. In addition, we provide theoretical results regarding the stability of the network for a common inverse problem example. Numerical experiments on image deblurring problems show that the proposed approach compares favorably with both state-of-the-art variational and machine learning methods in terms of image quality.

International audience; In this article, we introduce a new proximal interior point algorithm (PIPA). This algorithm is able to handle convex optimization problems involving various constraints where the objective function is the sum of a Lipschitz differentiable term and a possibly nons-mooth one. Each iteration of PIPA involves the minimization of a merit function evaluated for decaying values of a logarithmic barrier parameter. This inner minimization is performed thanks to a finite number of subiterations of a variable metric forward-backward method employing a line search strategy. The convergence of this latter step as well as the convergence the global method itself are analyzed. The numerical efficiency of the proposed approach is demonstrated in two image processing applications.