MENU

Click here to

×

Are you sure ?

Yes, do it No, cancel

Predicting the Severity of White Matter Hyperintensities Using Structural MRI and Machine Learning

W Lu1*, H Li2, L Zheng1, L Shi1, W Lu1, J Qiu1, (1) Shandong First Medical University & Shandong Academy of Medical Sciences, Taian, Shandong, CN, (2) Shandong University Of Science And Technilogy, Qingdao, Shandong, CN

Presentations

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

Room: AAPM ePoster Library

Purpose: The Fazekas scale is widely used clinically to quantify the severity of white matter hyperintensities (WMHs) by the amount of white matter T2 hyperintense lesions. Although imaging features of WMHs are hyperintense lesions, WMHs have an influence on the entire brain. It is supposed that structural measures from routine T1-weight images could predict the severity of WMHs.


Methods: Thirty-eight subjects with WMHs were collected and rated on Fazekas scale by an experienced doctor. The enrolled subjects included 10 scale-0 subjects, 17 scale-1 subjects, 6 scale-2 subjects and 5 scale-3 subjects. T1-weighted structural images were obtained from the enrolled subjects. FSL was used for the processing of T1-weighted images. Structural features were extracted from the processed T1-weighted images via the Harvard-Oxford atlas. The extracted structural features were then normalized to a range from 0 to 1. A quaternary label for Fazekas scale 0-3 was used. Recursive feature elimination was used for feature selection and a linear support vector machine was configured to predict the severity of WMHs via the selected structural features. Leave-one-out-cross-validation was used for performance evaluation with accuracy, sensitivity, specificity and receiver operating characteristics curve as evaluating metrics.


Results: The configured SVM classifier achieved an average accuracy of 86.8421% for the prediction of WMHs severity. The prediction accuracy for Fazekas scale 0-3 subjects were 84.2105%, 79.8474%, 84.2105% and 89.4737%, respectively. The brain regions contributed most to the prediction were the left superior temporal gyrus, right inferior temporal gyrus, right frontal medial cortex, right central opercular cortex, right parietal operculum cortex, left caudate and bilateral accumbens.


Conclusion: Structural MRI features and machine learning could accurately predict the severity of WMHs. The results also demonstrated that certain brain regions could be used as potential biomarkers for clinical diagnosis of WMHs.

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, CAD, MRI

Taxonomy

IM- MRI : Machine learning, computer vision

Contact Email