Search

Quantifying the phagocytosis of dynamic, unstained cells is essential for evaluating neurodegenerative diseases. However, measuring rapid cell interactions and distinguishing cells from backgrounds make this task challenging when processing time-lapse phase-contrast video microscopy. In this study, we introduce a fully automated, scalable, and versatile realtime framework for quantifying and analyzing phagocytic activity. Our proposed pipeline can process large data-sets and includes a data quality verification module to counteract potential perturbations such as microscope movements and frame blurring. We also propose an explainable cell segmentation module to improve the interpretability of deep learning methods compared to black-box algorithms. This includes two interpretable deep learning capabilities: visual explanation and model simplification. We demonstrate that interpretability in deep learning is not the opposite of high performance, but rather provides essential deep learning algorithm optimization insights and solutions. Incorporating interpretable modules results in an efficient architecture design and optimized execution time. We apply this pipeline to quantify and analyze microglial cell phagocytosis in frontotemporal dementia (FTD) and obtain statistically reliable results showing that FTD mutant cells are larger and more aggressive than control cells. To stimulate translational approaches and future research, we release an open-source pipeline and a unique microglial cells phagocytosis dataset for immune system characterization in neurodegenerative diseases research. This pipeline and dataset will consistently crystallize future advances in this field, promoting the development of efficient and effective interpretable algorithms dedicated to this critical domain. https://github.com/ounissimehdi/PhagoStat

International audience; Cutaneous melanoma is an invasive cancer with a worldwide annual death toll of 57,000 (Arnold \textit{et al.}, JAMA Dermatol 2022). In a metastatic state, surgical interventions are not curative and must be coupled with targeted therapy, or immunotherapy. However, resistance appears almost systematically and late-stage prognosis can remain poor. The complexity to eradicate melanoma stems from its plasticity; these cancer cells continually adapt to the tumor microenvironment, which leads to treatment resistance. Our primary assumption is that therapeutic resistance relies in part on a series of non-genetic transitions including changes in the metabolic states of these cancer cells. The 3D spatial distribution of blood vessels that are sources of nutrition and oxygen that drive this metabolic status is an important variable for understanding zoning aspects of this adaptation process. Using Whole Slide Images (WSI) of melanoma tumors from Patient-Derived Xenograft (PDX) mouse models, we build 3D vascular models to help predict and understand the metabolic states of cancer cells within the tumor. Our 3D reconstruction pipeline was based on PDX tumor samples sectioned over 2mm depth and stained with Hematoxylin and Eosin (H\&E). The pipeline involves three primary steps, including 2D vessel segmentation using Deep Learning, intensity- and affine-based image registration, and 3D reconstruction using interpolation and 3D rendering (allowing for better interaction with biologists, pathologists, and clinicians). The originality of our computer-assisted pipeline is its capability to (a) deal with sparse data (i.e., not all tissue sections were readily available), and (b) adapt to a multitude of WSI-related challenges (e.g., epistemic uncertainty, extended processing times due to WSI scale, etc.). We posit both our 3D reconstruction pipeline, quantitative results of the major stages of the process, and a detailed illustration of the challenges faced, presenting resolutions to improve the pipeline’s efficiency.

International audience; Non-invasive brain imaging techniques allow understanding the behavior and macro changes in the brain to determine the progress of a disease. However, computational pathology provides a deeper understanding of brain disorders at cellular level, able to consolidate a diagnosis and make the bridge between the medical image and the omics analysis. In traditional histopathology, histology slides are visually inspected, under the microscope, by trained pathologists. This process is time-consuming and labor-intensive; therefore, the emergence of Computational Pathology has triggered great hope to ease this tedious task and make it more robust. This chapter focuses on understanding the state-of-the-art machine learning techniques used to analyze whole slide images within the context of brain disorders. We present a selective set of remarkable machine learning algorithms providing discriminative approaches and quality results on brain disorders. These methodologies are applied to different tasks, such as monitoring mechanisms contributing to disease progression and patient survival rates, analyzing morphological phenotypes for classification and quantitative assessment of disease, improving clinical care, diagnosing tumor specimens, and intraoperative interpretation. Thanks to the recent progress in machine learning algorithms for high-content image processing, computational pathology marks the rise of a new generation of medical discoveries and clinical protocols, including in brain disorders.

National audience; Essential for a good adoption, as well as for a wise and unbiased use, explicability is a real technology lock to the evolution of Artificial Intelligence (AI), in particular concerning Machine and Deep Learning. Without an effective explicability of the proposed algorithms, these techniques will remain a black box for health (and not only) professionals, researchers, engineers and technicians - who assume (and will continue to assume) the full responsibility of their actions. Increasingly, engineers and designers of AI tools will have to demonstrate their responsibility by providing algorithms that guarantee the explicability of the proposed models. This article presents the motivations of an explainable AI, the main characteristics of the conceptual landscape of explainability in AI, the major families of explainability methods - with a focus on some of the most common methods, to finally present some of the opportunities, challenges and perspectives of this exciting field of human-machine interaction. Indeed, only through a good understanding of the challenges associated with this technological revolution that we will be able to transform AI into assets for our companies as well as for our human actors, partners and customers.; Essentielle pour une adoption efficace comme pour une utilisation avisée et objective de l'Intelligence Artificielle (IA), l'explicabilité est un véritable verrou de l'évolution de ces technologies, en particulier concernant l'apprentissage automatique et profond. Sans une réelle explicabilité des algorithmes proposés, ces technologies resteront une boîte noire pour les professionnels de santé (et pas seulement), chercheurs, ingénieurs, techniciens - qui assument (et vont continuer à assumer) la pleine responsabilité de leurs actes.De plus en plus, les ingénieurs exploitants et concepteurs d'outils d'IA devront donc faire preuve de responsabilité, en fournissant des algorithmes permettant de garantir l'explicabilité des modèles proposés.Cet article présente les motivations d'une IA explicable, les principales caractéristiques du paysage conceptuel de l'explicabilité en IA, les grandes familles de méthodes pour l'explicabilité - avec un focus sur quelques méthodes parmi les plus courantes, pour finir sur un aperçu des opportunités, challenges et perspectives de ce domaine passionnant de l'interaction homme-machine.En effet, c'est uniquement par une bonne compréhension des challenges associés à cette révolution technologique que nous pourrons la transformer en atout pour nos entreprises ainsi que pour l'ensemble de nos acteurs, partenaires et clients humains.

International audience; Being a hallmark of several solid tumors, hypoxia - a state of reduced level of tissue oxygen tension and a result of aberrant vasculature - leads to several alterations in the tumor microenvironment. Hypoxic regions of neoplasm are prone to be more resistant towards radiation therapy than compared to well oxygenated ones (A. L. Harris 2002). Furthermore, hypoxia and its mediators influence multiple signaling pathways and gene regulation to promote neovascularization, invasion, migration, adhesion, metastasis, and phenotypic switches (D. S. Widmer et al. 2013, A. Tameemi et al., 2019). Hence hypoxia is one of the leading factors which contributes towards intratumor heterogeneity and resistance against treatments, these two features being particularly important and common in many invasive tumors including melanoma (B. Bedogni et al. 2009, D'Aguanno et al. 2021). Estimation of accurate hypoxia profile would be key for better prognosis and design of more efficient treatment approaches. Mathematical modeling has been proven a useful tool to understand and predict such complex dynamics. Several computational and mathematical models have been proposed to describe tissue oxygenation, however the majority of them are restricted to synthetic data and qualitative results, lacking application to and connection with real tumor tissues and experimental results.We propose mechanistic modeling frameworks, which are driven by experimental data, to explain and mimic oxygen-hypoxia dynamics. The data is in the form of tissue scans of Patient Derived Xenograft (PDX) of breast, ovarian and pancreatic as well as human melanoma tumors. These scans of tumor tissue slices are immunohistochemical stained with CD31 -cluster of differentiation 31, marking the presence of endothelial cells- and CAIX- carbonic anhydrase IX, regulated by the hypoxia-inducible factor (HIF) 1, is an intrinsic marker of tumor hypoxia - markers. Keeping the data availability in mind, the distribution of oxygen is described by a reaction-diffusion partial differential equation with the source term incorporating the contribution from blood vessel density (obtained from CD31 staining) for the 2D model and from the vasculature architecture and the geometry of each blood vessel (reconstructed from several 2D tissue slices) for the 3D model. Next, hypoxia is modeled from the obtained oxygen distribution using an algebraic equation. The further steps include estimation of parameters and validation. The obtained parameters demonstrate biological relevance. 3D reconstruction, which is underway, is required for obtaining 3D profiles of oxygen and hypoxia. This requirement leads to another aspect of this work consisting in quantification of the error made when 2D models are used instead of more realistic 3D models. This is important since the 3D reconstruction is not always feasible, especially for patient tissue samples. A framework to quantify this approximation error would be essential for evaluating the hypoxia profile for clinical applications. Future work involves development of a general framework, applicable to most of the solid tumors, to estimate oxygen and hypoxia distribution based on the 3D reconstruction of blood vessels as well as for the 2D case with an error bound due to the approximation.

International audience

Accurate hepatic vessel segmentation and registration using ultrasound (US) can contribute to beneficial navigation during hepatic surgery. However, it is challenging due to noise and speckle in US imaging and liver deformation. Therefore, a workflow is developed using a reduced 3D U-Net for segmentation, followed by non-rigid coherent point drift (CPD) registration. By means of electromagnetically tracked US, 61 3D volumes were acquired during surgery. Dice scores of 0.77, 0.65 and 0.66 were achieved for segmentation of all vasculature, hepatic vein and portal vein respectively. This compares to inter-observer variabilities of 0.85, 0.88 and 0.74 respectively. Target registration error at a tumor lesion of interest was lower (7.1 mm) when registration is performed either on the hepatic or the portal vein, compared to using all vasculature (8.9 mm). Using clinical data, we developed a workflow consisting of multi-class segmentation combined with selective non-rigid registration that leads to sufficient accuracy for integration in computer assisted liver surgery.

International audience; A once rare disease, malignant cutaneous melanoma has evolved to become one of the most fatal forms of cancers, accounting for a global death toll of 50,000 every year. While surgical interventions have shown to increase survival rates for patients with early-stage melanoma, these survival rates decrease as the disease progresses to metastases. Surgical interventions alone are no longer curative, and interventions must include chemotherapies to improve line of defense. However, while therapeutic interventions are available, the efficacy of these treatments can be greatly impacted by the complexity of the disease. When introduced to new metastatic sites, the cancer cells are faced with new environmental conditions, in which sometimes nutrients and oxygen are scarcely available. Through a biological phenomenon referred to as metabolic rewiring, the cancer cells adapt to their new, metabolic changes, promoting its own survival and proliferation under all stressful conditions. While the impact of metabolic rewiring and its influence on treatment efficacy has been established, the direct correlation between the two is not completely understood. The MALMO project (Mathematical Approaches to Modelling Metabolic Plasticity and Heterogeneity in Melanoma) aims to bridge the gap in knowledge within this domain by combining mathematical approaches with state-of-the-art artificial intelligence (AI) methods to automate pathology grading systems, conduct quantitative image analysis (including feature extraction of relevant biomarkers), and develop a model to predict the development of heterogeneity in melanoma tumors. The mathematical model – based on partial differential equations (PDE) – is being designed to account for time, space, and biological scales, such as molecular, cellular, and tissular, in order to understand how melanoma behaves in oxygen and nutrient-deprived tumor microenvironments. The mathematical model is augmented with the help of AI, in which features from hematoxylin and eosin (H&E) and cluster of differentiation 31 (CD31)-stained whole slide images (WSI), as well as Hyperion images, are being extracted and evaluated to assist with the modeling component. In this paper, we introduce the MALMO project and demonstrate some preliminary findings using AI. Understanding this rewiring and developing a prediction model will help to further our understanding of melanoma progression. In doing so, we can begin creating tailored therapies which could better target metastatic melanoma and improve the survival rate for these late stage affected patients.

International audience; While the advent of artificial intelligence (AI) has seen an increase in computational power and development of novel technologies to expedite current human-run processes, with this breakneck speed some concerns have risen around the technology's social, legal, and ethical implications. Responsible AI (RAI) has emerged as a new discipline designed to ensure AI technologies produce equitable outcomes and that AI is developed for the betterment of humanity. In this review poster, some of the AI-related concerns are highlighted, various aspects of RAI are defined (as there current remains no real consensus to what RAI involves), some attempts at creating quantifiable AI are illustrated, and some future directions are proposed.