Skip to Main content Skip to Navigation

Segmentation automatique des anomalies de la substance blanche du sujet âgé

Thomas Samaille 1
1 ARAMIS - Algorithms, models and methods for images and signals of the human brain
Inria Paris-Rocquencourt, UPMC - Université Pierre et Marie Curie - Paris 6, ICM - Institut du Cerveau et de la Moëlle Epinière = Brain and Spine Institute
Abstract : White Matter Hyperintensities (WMH) may be observed on T2-weighted magnetic resonance imaging (MRI). Although they are commonly seen in elderly people, it has been demonstrated that a high level of WMH is a risk factor for stroke and dementia. The impact of WMH on cognition or the effect of a low level of WMH remains however controversial. Today, WMH quantification is still mainly performed by visuel rating, although a segmentation would be more sensitive and would enable volume measurement, spatial distribution analysis and lesion follow-up. This thesis introduces a new pipeline, named WHASA (White matter Hy- perintensities Automatic Segmentation Algorithm), to automatically segment WMH from a T1-weighted and a FLAIR image in multicentre studies. WHASA relies on the coupling of non linear diffusion and watershed segmentation. Regions corresponding to WMH are selected based on intensity and location character- istics. Results obtained with WHASA were evaluated with respect to a manual reference on 67 subjects, for which images have been acquired in six different centres. Performances were also compared to state-of-the-art methods on the same dataset. Finally, WHASA was applied on subjects from new studies to assess its robustness. A total of 260 subjects, for which images were acquired from eleven MRI scanners, including 1,5 T and 3 T machines, and with slice thicknesses varying between 2.5 and 6.2 mm, has been analyzed with WHASA.
Document type :
Complete list of metadata

Cited literature [149 references]  Display  Hide  Download
Contributor : Thomas Samaille Connect in order to contact the contributor
Submitted on : Wednesday, July 19, 2017 - 11:00:26 PM
Last modification on : Thursday, December 10, 2020 - 3:43:07 AM


  • HAL Id : tel-01565592, version 1


Thomas Samaille. Segmentation automatique des anomalies de la substance blanche du sujet âgé. Neurosciences. Paris 6, 2013. Français. ⟨tel-01565592⟩



Record views


Files downloads