Efficient Cascade Model for Pathological Brain Image Detection by Magnetic Resonance Imaging

Early diagnosis plays an important role in preventing the progression of and treating pathological brain disease. Although exceptional progress has been achieved in the areas of pattern recognition and biomedical sciences, early diagnosis remains a challenging problem. The development of automatic and accurate computer-aided diagnosis (CAD) systems for distinguishing brain disease based on magnetic resonance imaging (MRI) has gained tremendous importance in recent years. These systems assist radiologists in the accurate analysis of brain MRI images and they can substantially reduce the time required for analysis. Classifying brain images as normal or pathological is essential for understanding and analyzing medical images. In this study, we suggest a novel approach for distinguishing pathological and normal brain images. The proposed system employs contrast limited adaptive histogram equalization (CLAHE) for image preprocessing, the curvelet transform for feature extraction, probabilistic principal component analysis (PPCA) for feature reduction, and a linear support vector machine (L-SVM) for classification. The proposed scheme was validated using a dataset of 90 images (18 normal and 72 abnormal). According to the results of the 5x5-fold cross-validation, the proposed method outperformed seven state-of-the-art algorithms in terms of its sensitivity, specificity, precision, and accuracy.