Machine learning-based prediction of motor status in glioma patients using diffusion MRI metrics along the corticospinal tract

Abstract

Along tract statistics enables white matter characterization using various diffusion MRI (dMRI) metrics. Here, we applied a machine learning (ML) method to assess the clinical utility of dMRI metrics along corticospinal tracts (CST), investigating whether motor glioma patients can be classified with respect to their motor status. The ML-based analysis included developing models based on support vector machine (SVM) using histogram-based measures of dMRI-based tract profiles (e.g., mean, standard deviation, kurtosis and skewness), following a recursive feature elimination (RFE) method based on SVM (SVM-RFE). Our model achieved high performance (74% sensitivity, 75% specificity, 74% overall accuracy and 77% AUC). Incorporating the patients demographics and clinical features such as age, tumor WHO grade, tumor location, gender and resting motor threshold (RMT) into our designed models demonstrated that these features were not as effective as microstructural measures. The results revealed that ADC, FA and RD contributed more than other features to the model.

Corpus callosum abnormalities at term-equivalent age are associated with language development at two years corrected age in infants born very preterm

“…This approach offers improved reproducibility and significantly reduces the likelihood of measurement errors. The TractSeg tool has already been utilized to evaluate white matter tracts in school-age preterm children 28 , patients with traumatic brain injuries 29 , and gliomas 30 .…”