You are here : Home > Processing and analysis of multimodal biomedical images of the normal brain or of preclinical models of brain diseases

Processing and analysis of multimodal biomedical images of normal brain and preclinical models of neurodegenerative diseases

Traitement et analyse d'images biomédicales multimodales de cerveau normal ou de modèles précliniques de maladies neurodégénératives

Group leader: Thierry Delzecaux​

Published on 3 July 2018



  Thierry Delzescaux


BIOmedical Processing of Images, Computer Science Engineering Laboratory

Three-dimensional correspondence of complementary imaging data

In the field of neurodegenerative diseases (Alzheimer's, Parkinson's, Huntington, etc.), the use of anatomical and functional imaging performed in vivo or post mortem improves our understanding of pathophysiological mechanisms and provides new powerful tools to develop and validate novel therapeutic approaches. In this context, image processing plays a key role in the integration and analysis of multimodal information from microscopic to macroscopic scale.

A major contribution of our research group is the ability to perform 3D analyzes based on 2D series of histological sections reconstructed in 3D.

Exploratory analysis approaches using voxel-wise approach combined with 3D digital brain atlas can be applied to post-mortem data reconstructed in 3D. These tools are extremely powerful to study biomarkers of interest. It is thus possible to improve our knowledge of biological mechanisms through quantitative methods of analysis and also to make it possible to objectively evaluate drug efficacy.

Several types of complementary data can be obtained using in vivo imaging (MRI, PET, CT, etc.) and post mortem techniques (histology, autoradiography, microscopy, etc.). We have also developed protocols to co-register in 3D all these data.

These original methodological developments have already shown their potential to extract relevant information with superior quality compared to 2D approaches. They were successfully applied in multiple preclinical studies (in rodent and primate) performed at MIRCen. In addition, these approaches offer the unique opportunity to merge multimodal information produced at different scales for example to interpret macroscopic physical phenomena (in vivo) using biological information acquired post mortem at microscopic scale.

Processing of multimodal images and 3D reconstruction

The objectives of our team

Developments relative to post mortem imaging: 3D reconstruction methods to improve the anatomo-functional analysis of histological and autoradiographical data.

Developments relative to imaging in vivo: co-alignment methods to supervise anatomo-functional analyses of PET and MRI data.

Complementary post mortem and in vivo imaging: co-registration of multimodal imaging in the same subject using block-face photographs as an intermediate 2D / 3D modality; validation of new imaging systems, PET tracers, contrast agents for MRI; characterization of new experimental models and evaluation of new therapeutic approaches.

Creation of new integrated and automated algorithms and software to conduct reproducible and robust analyses of large image dataset: scripts for the processing of biomedical images, automatic segmentation, creation of digital brain atlases, database management, optimized exploitation of virtual microscopy data, high performance computing, etc.


Contribution to BrainVISA/Anatomist: BrainRAT, 3D HAPi, Primatologist 

Members of the laboratory associated with these projects


  • CEA-NeuroSpin, CEA-SHFJ, INRIA, AP-HP Groupe Henri Mondor Albert Chenevier (Créteil, France), Université Paris XI IMNC (Orsay, France), UPMC LIF INSERM U678, Université Paris XII, ANSTO (Australia), Université de Versailles St-Quentin-en-Yvelines, CEA-Ibitecs SIMOPRO, Université de Maastricht (Netherland), Institut Pasteur (Paris & Lille, France).


Quantification of brain metabolism and protein aggregates. Physiopathology of the brain and preclinical models of neurodegenerative diseases



Multiple stains of an AD mouse brain reconstructed and analyzed in 3D (Vandenberghe et al., SCi. Rep., 2016)

​Recent Publications

Computational optimization for fast and robust automatic segmentation in virtual microscopy using brute-force-based feature selection.
Bouvier C, Clouchoux C, Souedet N, Hérard AS, You Z, Jan C, Hantraye P, Mergoil G, Rodriguez BH, Delzescaux T.
Conf Proc ICPRAI 2018 - International Conference on Pattern Recognition and Artificial Intelligence

Primatologist: /A modular segmentation pipeline for macaque brain morphometry.
Balbastre Y, Rivière D, Souedet N, Fischer C, Hérard AS, Williams S, Vandenberghe ME, Flament J, Aron-Badin R, Hantraye P, Mangin JF, Delzescaux T.
NeuroImage (2017) Sep 9;162:306-321.

A validation dataset for Macaque brain MRI segmentation.
Balbastre Y, Rivière D, Souedet N, Fischer C, Hérard AS, Williams S, Vandenberghe ME, Flament J, Aron-Badin R, Hantraye P, Mangin JF, Delzescaux T.
Data Brief. 2017 Nov 4;16:37-42.

High-throughput 3D whole-brain quantitative histopathology in rodents. 
Vandenberghe ME, Hérard AS, Souedet N, Sadouni E, Santin MD, Briet D, Carré D, Schulz J, Hantraye P, Chabrier PE, Rooney T, Debeir T, Blanchard V, Pradier L, Dhenain M, Delzescaux T.
Sci Rep (2016) Feb 15;6:20958.

Robust supervised segmentation of neuropathology whole-slide microscopy images.
Vandenberghe ME, Balbastre Y, Souedet N, Hérard AS, Dhenain M, Frouin F, Delzescaux T.
Conf Proc IEEE Eng Med Biol Soc.;2015:3851-4.