MENU

Click here to

×

Are you sure ?

Yes, do it No, cancel

Diagnosis of White Matter Hyperintensities Using Brain Morphometry and Support Vector Machine

L Zheng, W Lu*, W Lu, L Shi, J Qiu, Shandong First Medical University & Shandong Academy of Medical Sciences, Taian, Shandong, CN,

Presentations

(Sunday, 7/12/2020)   [Eastern Time (GMT-4)]

Room: AAPM ePoster Library

Purpose: White matter hyperintensities (WMHs) represent abnormal signal intensities in white matter regions in T2 fluid attenuated inversion recovery (FLAIR) MRI images. No matter where white matter lesion is, it reflects a degeneration of neural pathways in the whole brain associated with ageing and neurological disease. Therefore, we aim to diagnose WMHs using brain morphometry from routinely T1-weighted MRI images.


Methods: Subjects were collected and underwent T1 and T2 FLAIR MRI scan. WMHs were rated based on Fazekas scale 0 to 3 by an experienced doctor using T2 FLAIR images. Finally, 38 subjects with 10 Fazekas scale-0 subjects, 17 scale-1 subjects, 6 scale-2 subjects and 5 scale-3 subjects were included. T1-weighted MRI images were processed by FSL, and were segmented according to the Harvard-Oxford atlas. Mean volume of each brain region was extracted and taken as feature for machine learning. Morphological features were normalized to a range from 0 to 1. A binary label with -1 for Fazekas scale-0 subjects and 1 for Fazekas scale 1-3 subjects was used. Sequential backward elimination was used for feature selection. A linear support vector machine (SVM) was configured for the diagnosis of WMHs based on morphological features. Leave-one-out-cross-validation was used to evaluate the performance of the configured SVM.


Results: The linear SVM classifier achieved an accuracy of 97.37%, with sensitivity of 0.9643, and specificity of 1. Area under the curve (AUC) was 0.9714 (p < 0.001, tested with 1000-times permutation test). The features contributed most to the diagnosis were the right temporal occipital fusiform cortex, left occipital fusiform gyrus, bilateral occipital pole, right thalamus, right putamen, bilateral amygdala and bilateral accumbens.


Conclusion: Brain morphometry could be used for the diagnosis of WMHs, and morphological information from certain brain regions could be used as potential biomarkers for WMHs diagnosis apart from the lesions.

Funding Support, Disclosures, and Conflict of Interest: This work was supported by the Key Research and Development Program of Shandong Province (2017GGX201010), Academic Promotion Programme of Shandong First Medical University (2019QL009), Traditional Chinese Medicine Science and Technology Development Plan of Shandong Province (2019-0359), and Taishan Scholars Program of Shandong Province (TS201712065).

Keywords

Brain, MRI, Linear Classifier

Taxonomy

IM- MRI : CAD

Contact Email