Predicting Final Extent of Ischemic Infarction Using Artificial Neural Network Analysis of Multi-Parametric MRI in Patients with Stroke

In ischemic stroke, the extent of ischemic lesion recovery is one of the most important correlate of functional recovery in brain. Using a set of acute phase MR images (Diffusion-Weighted - DWI, T1-Weighted - T1WI, T2-Weighted T2WI, and proton density weighted - PDWI) for inputs, and the chronic T2WI at 3 months as an outcome measure, an Artificial Neural Network (ANN) was trained to predict the 3-month outcome in the form of a pixel-by-pixel forecast of the chronic T2WI. The ANN was trained and tested using 14 slices from 3 subjects using a K-Folding Cross-Validation (KFCV) method with 14 folds. The Area Under the Receiver Operator Characteristic Curve (AUROC) for 14 folds was used for training, testing and optimization of the ANN. After training and optimization, the ANN produced a map that was well correlated (r = 0.88, p ≪ 0.0001) with the T2WI at 3 months. To confirm that the trained ANN performed well against a new dataset, 13 slices from 4 other patients were shown to the trained ANN. The prediction made by the ANN had an excellent overall performance (AUROC = 0.82), and was very well correlated to the 3-month ischemic lesion on T2-Weighted image.