HAL will be down for maintenance from Friday, June 10 at 4pm through Monday, June 13 at 9am. More information
Skip to Main content Skip to Navigation
Master Thesis

TractLearn : comparaison au modèle linéaire généralisé pour l’exploration des voies visuelles antérieures

Abstract : TractLearn was recently proposed for Diffusion-weighted MRI quantitative analyses, based on Riemannian distances between anatomical structures. It allows to detect joint quantitative variations in a group of voxels, thus in theory to decrease the number of false negative in comparison with the General Linear Model (GLM). TractLearn also takes advantage of a manifold approach to capture controls variability as standard reference. Here we aim at comparing the performance of TractLearn with the GLM in detecting optic nerve voxel alteration, using the presence of visual impairment as clinical reference. The second objective is to test whether TractLearn does require a statistical correction for multiple comparison, usually recommended in the case of individual voxel testing in Neurosciences. As TractLearn relies on a global test on a collection of voxels, applying such correction can lead to false negatives in patients with optic neuropathy. We illustrate the potential added value of this method on a case-controlled study including healthy population of single-shell diffusion MRI data, as compared with patients referred for unilateral optic neuropathy. Our contributions are to show that: 1/ TractLearn outperforms the General Linear Model in voxelwise analysis of Diffusion-weighted MRI metrics by increasing the diagnostic sensitivity. 2/ Application of statistical correction for multiple comparison in this work severely limit the number of clinically-relevant detected lesions.
Document type :
Master Thesis
Complete list of metadata

Contributor : Jean-Hugues Morneau Connect in order to contact the contributor
Submitted on : Tuesday, November 23, 2021 - 2:36:31 PM
Last modification on : Thursday, November 25, 2021 - 3:28:54 AM
Long-term archiving on: : Thursday, February 24, 2022 - 7:54:14 PM


Files produced by the author(s)


  • HAL Id : dumas-03443263, version 1


Clément Jean. TractLearn : comparaison au modèle linéaire généralisé pour l’exploration des voies visuelles antérieures. Médecine humaine et pathologie. 2021. ⟨dumas-03443263⟩



Record views


Files downloads