Predicting the Development of Normal-Appearing White Matter With Radiomics in the Aging Brain: A Longitudinal Clinical Study

Abstract

Background and Purpose: Normal-appearing white matter (NAWM) means normal, but potentially microstructural changed matter around white matter hyperintensity (WMH) on conventional MR images. Radiomics is an emerging quantitative imaging technique that provides additional information compared to visual analysis. The aim of this study is to explore whether WMH could be predicted at the early stage of NAWM with texture analysis in the general elderly population. Methods: Image data was retrieved from PACS from 2012 to 2017. Subjects (≥60y) received more than two MR exams on a same machine with time intervals more than one year. By comparing baseline and follow-up images, patients with remarkable progress of WMH were included as the case group (n=51), and age-matched subjects without WMH were included as the control group (n=51). Segmentations of ROIs were done with ITK software. Two ROIs of developing NAWM (dNAWM) and non-developing NAWM (Non-dNAWM) were drawn separately on FLAIR image of each patient. dNAWM appeared normal on baseline but evolved into WMH on the follow-up image; Non-dNAWM appeared normal on both baseline and follow-up image. A third ROI of normal white matter (NWM) was then extracted from normal control groups, which displayed normal on both baseline and follow-up image. Texture features were dimensional reduced with ANOVA+MW, correlation analysis and LASSO. Three models were built basing on the optimal parameters of dimension reduction: Model1 (NWM vs. dNAWM); Model2 (Non-dNAWM vs. dNAWM) and Model 3(NWM vs. Non-dNAWM). And ROC curve was adopted to evaluate the classification validity of the models. Results: Basic characteristics between patients and controls showed no statistical difference. The AUC of Model1 in training and test group were 0.967(95%CI: 0.831 - 0.999) and 0.954 (95%CI: 0.876 - 0.989) respectively. The AUC of Model 2 were 0.939 (95%CI: 0.856 - 0.982) and 0.846 (95%CI: 0.671 - 0.950) respectively. The AUC of Model 3 were 0.713 (95%CI: 0.593 - 0.814) and 0.667 (95%CI: 0.475 - 0.825) respectively. Conclusions: Radiomics textural analysis can clearly distinguish dNAWM from non-dNAWM on FLAIR images, which could be used to predict the presence of NAWM lesions before it develops into a visible WHM.