Brain Lesion Segmentation And Classification Using Diffusion-Weighted Imaging (DWI)

Research and development of brain detection and diagnosis system for brain disorder based on Magnetic Resonance Imaging (MRI) have become one of the most common interest in the past few years. Out of various MRI techniques, Diffusion-Weighted Imaging (DWI) remains the most accurate technique for early detection and discrimination of several brain lesions such as stroke. This study proposed the image analysis technique for automatically segmenting and classifying abnormal lesion structures from DWI. Four lesions namely acute stroke, chronic stroke, solid tumor and necrosis were analyzed. The proposed analysis framework were pre-processing, segmentation, features extraction and classification. Four different segmentation techniques were proposed based on Thresholding with Morphological Operation (TMO), Fuzzy C-Means (FCM), Fuzzy C-Means with Active Contour (FCMAC) and Fuzzy C-Means with Correlation Template (FCMCT) to segment the lesion’s region. Next, the statistical parameters from spatial and wavelet transforms were extracted from the Region of Interest (ROI) as features. These features were classified using a rule-based classifier for automatic classification. The results indicate that FCMCT offered the best performance for Jaccard Index, Dice Index, False Positive Rate and False Negative Rate which were 0.6, 0.73, 0.19 and 0.2 respectively. The overall accuracy, sensitivity and specificity for the classification were 89 %, 86 % and 96 %. In conclusion, the proposed hybrid analysis has the potential to be explored as a computer-aided tool to detect and diagnose of human brain lesion