StratifIAD

Refining Alzheimer’s disease patient stratification using explainable artificial intelligence in computational histopathology.

Big Brain Theory (BBT3)
PIs: Daniel Racoceanu, Benoît Delatour

Collaborators: Lev Stimmer, Susana Boluda, Anuradha Kar, Gabriel Jimenez, Mehdi Ounissi

StratifIAD develops explainable artificial intelligence approaches for computational histopathology, focusing on the localization, characterization and interpretation of tau and amyloid-beta aggregates in Alzheimer’s disease whole-slide histopathology.

StratifIAD project overview

Research Themes

Alzheimer’s Disease
Computational Histopathology
Explainable AI
Whole-Slide Imaging
Deep Learning
Graph-Based Tissue Modeling

Project Overview

The project aims to develop algorithms for fully automated localization and characterization of tau and amyloid-beta aggregates in histological images of the human brain.

The morphology and spatial organization of these aggregates are highly heterogeneous. Amyloid-beta accumulation may appear as focal deposits or diffuse plaques, whereas tau lesions can form neurofibrillary tangles and neuritic structures with complex topological organization.

Tau Pathology Amyloid-beta Deep Learning Explainability

Scientific Objectives

  • Develop deep learning models with explainability mechanisms for segmentation and interpretation of brain tissue histopathology.
  • Build cloud-based platforms for visualization, annotation refinement and collaborative analysis of whole-slide images.
  • Explore graph-based models for representing morphology and topology of tau-related structures.
  • Improve patient stratification through interpretable computational biomarkers.

Scientific Vision

StratifIAD investigates how explainable AI can support computational pathology not only as a prediction tool, but also as a scientific instrument for understanding disease heterogeneity and tissue organization.

The project places strong emphasis on traceability, interpretability and reproducibility, with the objective of developing clinically meaningful AI systems that remain compatible with expert-driven histopathological analysis.

Resources

GitHub Organization

Repositories, code and experimental resources.

Open GitHub

MICCAI 2022 Study

Neuritic plaque segmentation with explainable deep learning.

Open Project

Selected Publications

  • A meta-graph approach for analyzing whole-slide histopathological images of human brain tissue with Alzheimer’s disease biomarkers
    Gabriel Jimenez, Pablo Mas, Anuradha Kar, Julien Peyrache, Lea Ingrassia, Susana Boluda, Benoît Delatour, Lev Stimmer, Daniel Racoceanu — SPIE Medical Imaging 2023.
  • Visual Deep Learning-Based Explanation for Neuritic Plaques Segmentation in Alzheimer’s Disease Using Weakly Annotated Whole Slide Histopathological Images
    G. Jimenez et al. — MICCAI 2022.
  • Improving segmentation quality with fusion based post-processing: an assessment with 3D deep learning and classical segmentation pipelines
    A. Kar — CVPR-CVMI 2022.
  • Interpretable Deep Learning in Computational Histopathology for refined identification of Alzheimer’s disease biomarkers
    Garay, Kar, Ounissi, Stimmer, Delatour and Racoceanu — AAIC 2022.
  • A deep learning framework for stratification of Alzheimer’s disease patients using whole-slide histopathological brain tissue images
    Kar, Jimenez, Ounissi, Stimmer, Delatour and Racoceanu — ECDP 2022.
  • Empowering Researchers for Understanding Alzheimer’s Disease using Explainable AI
    Jimenez, Kar, Ounissi and Racoceanu — AI4Health Winter School 2022.