Skip to Main content Skip to Navigation
Master Thesis

Comparison of automatic segmentation versus visual assessment of white matter hyperintensities and parenchymal atrophy in Alzheimer’s disease

Abstract : Introduction: White matter hyperintensities (WMH) and parenchymal atrophy are common find-ings in elderly and in dementia. Automated segmentation could be helpful for quantification in patients referred for amnestic Mild Cognitive Impairment (aMCI) and probable Alzheimer’s dis-ease (AD). Methods: 123 patients referred for probable AD or aMCI had a MRI (Axial Flair, 3D T1-FFE). Normalized WMH and cortical volumes were measured using Pixyl.neuro and FreeSurfer and compared to usual visual scores. Results: 110 out of 123 patients (89%) were analyzed after exclusion because of technical issues or inaccurate segmentation. WMH volumes were correlated with Fazekas’ visual scores (r=0.78). Hippocampal volumes were correlated with Scheltens’ visual scores (r>0.63). However frontal and parietal cortical volumes were not correlated with visual scores of parietal atrophy and the gradient of frontoparietal atrophy. Conclusion: Automated segmentations are reliable to estimate WMH and hippocampal volumes, as visually assessed. However, automated cortical segmentation does not allow to classify fron-toparietal atrophy as routinely performed in clinical practice.
Document type :
Master Thesis
Complete list of metadatas

Cited literature [62 references]  Display  Hide  Download

https://dumas.ccsd.cnrs.fr/dumas-01887452
Contributor : Jean-Hugues Morneau <>
Submitted on : Monday, October 8, 2018 - 10:39:34 AM
Last modification on : Wednesday, July 15, 2020 - 8:58:04 AM
Long-term archiving on: : Wednesday, January 9, 2019 - 12:22:32 PM

File

2018GREA5086_malaureille_lison...
Files produced by the author(s)

Identifiers

  • HAL Id : dumas-01887452, version 1

Citation

Lison Malaureille. Comparison of automatic segmentation versus visual assessment of white matter hyperintensities and parenchymal atrophy in Alzheimer’s disease. Human health and pathology. 2018. ⟨dumas-01887452⟩

Share

Metrics

Record views

79

Files downloads

140