Prediction of Ischemic Stroke Lesion and Clinical Outcome in Multi-modal MRI Images Using Random Forests

Abstract

Herein, we present an automated segmentation method for ischemic stroke lesion segmentation in multi-modal MRI images. The method is based on an ensemble learning technique called random forest (RF), which generates several classifiers and combines their results in order to make decisions. In RF, we employ several meaningful features such as intensities, entropy, gradient etc. to classify the voxels in multi-modal MRI images. The segmentation method is validated on both training and testing data, obtained from MICCAI ISLES-2016 challenge dataset. The evaluation of the method is done by performing two tasks: ischemic stroke lesion outcome prediction (Task I) and clinical outcome prediction (Task II). For Task I, the performance of the method is evaluated relative to the manual segmentation, done by the clinical experts. For Task II, the performance of the method is evaluated relative to the 90 days mRS score, provided as ground truths by ISLES-2016 challenge organizers. The experimental results show the robustness of the segmentation method, and that it achieves reasonable accuracy for the prediction of both ischemic stroke lesion and clinical outcome in multi-modal MRI images.