Description
Date depot: 18 mars 2021
Titre: Graph-based Mathematical Morphology for the Characterization of the Spatial Organization of Histological Structures in High-Content Images: Appl. Tumor Microenvironment in Breast Cancer
Directeur de thèse:
Daniel RACOCEANU (ICM)
Domaine scientifique: Sciences et technologies de l'information et de la communication
Thématique CNRS : Non defini
Resumé: Besides its contribution to computer-aided diagnosis, Digital Pathology has opened up a new dimension in the investigation of complex diseases, by adopting innovative image analysis tools able to provide new features exposing different facets of the disease, impossible to analyze using classical microscopy by the human eye. One of the most challenging problems in histological image analysis is the evaluation of the spatial organizations of histo- logical structures in the tissue. In fact, histological sections may contain a very large number of cells of different types and irregularly distributed, which makes their spatial content indescribable in a simple manner.Graph-based methods have been widely explored in this direction, as they are effective representation tools having the expressive ability to describe spatial characteristics and neighborhood relationships that are visually interpreted by the pathologist. We can distinguish three main families of graph-based methods used for this purpose: syntactic structure analysis, network analysis and spectral analysis. However, another distinctive set of methods based on mathematical morphology on graphs can be developed for this issue. The main goal of this dissertation is the development of a framework able to provide quantitative evaluation of the spatial arrangements of histological structures using graph-based mathematical morphology. The main contributions of this work are five-fold.First, we propose a theoretical framework dedicated to the evaluation of the spatial arrangement of points based on features derived from mathematical morphology on graphs. Two morphological transforms are established and they have shown the ability to describe the spatial distribution of a point set and how distinct point sets are located relative to each other. The framework suggests the integration of graph-based mathematical morphology in spatial point pattern analysis. The new methodology would find an effective use in a wide range of practical problems.Second, we present a comprehensive review on graph-based methods explored in histopatho- logical image analysis. We propose a classification of the state-of-the-art methodologies based on the type of the graph, the category of the feature extraction technique, and the histological structures considered.Third, we develop a method for nuclei detection and classification from histopathological images of HES (Haematoxylin, Eosin, Saffron)-stained breast cancer tissues, using supervised machine learning algorithms based on color and texture information, and methods for collagen and adipose tissue segmentation.Fourth, we study the spatial distribution of the different tissue components and the spatial interactions between them using the proposed graph-based mathematical morphology approach. The analysis is conducted on a dataset composed of 55 whole slide images (WSI) of tissue sections obtained from surgical resections of invasive breast carcinomas from dis- tinct patients. The approach offers a comprehensive visual interpretation of different spatial aspects encountered in histopathology, and has shown the ability to distinguish between different tumor architectures and tumor microenvironment configurations.Fifth, we propose an original tissue simulation approach that gives access to a large spectrum of tumor-like architectures based on tools of classical mathematical morphology and statistics. The method can be used to generate large synthetic datasets in order to evaluate the effectiveness of graph-based methods in the quantification of the spatial organization of histological structures.
Doctorant.e: Ben Cheikh Bassem