Development of Computer Aids ASPECTS System for Acute Ischemic Stroke Patient: A Preliminary Study

Abstract

In recent years, stroke ranked within the top ten leading causes of death and the incidence is still rising. As a result of clinical interpretation of Alberta Stroke Program Early CT Score (ASPECTS), the relevant personnel to define stroke area and score range are not consistent and cause difficulty to make treatment decision. This study was to develop a computer-aided scoring system for ischemic stroke patient to help doctors effectively determine the severity of ischemic stroke. Image processing technology was used to develop the system. First, an adaptive median filter was used to filter noise in computed tomography (CT) image, and then bi-level and regional growth methods were used to obtain effective image information. After texture parameters selection through t-test and support vector machine (SVM), regions of interesting (ROI) were automatically selected. Finally, the ischemic severity were obtained based on calculated ASPECTS score (by compared the left and right sides of the brain image). The CT images of 80 sets (40 training sets and 40 test sets) were used to evaluate the system by comparing with corresponding DWI-MRI. The results showed that the area under the ROC curve of the training sets and the test sets were 0.952 and 0.938, respectively, when four parameters (autocorrelation, variance, maximum probability, and homogeneity) were chosen. Accuracy was 0.90, sensitivity was 0.76, specificity was 1, and Kappa value was 0.52 for test data respectively, and the performance was superior to the physician group.