Computer Aided Diagnosis Model for Brain Stroke Classification in MRI Images Using Machine Learning Algorithms
Recently, brain stroke is considered as a major cause of mortality, hospitalization, and acquired disabilities. Stroke diagnosis is generally done based on the symptoms exhibited and, more specially, on the results of neuroimaging examinations. Magnetic Resonance Imaging (MRI) is a widely employed neuroimaging technique, because it seems to be faster, inexpensive and accurate. In this view, this paper presents a new computer aided diagnosis (CAD) technique using machine learning (ML) algorithms for stroke prediction in MRI images. The proposed CAD model involves preprocessing, feature extraction and classification. Initially, the input image undergoes preprocessing to improve the image quality. Followed by, a set of four feature extraction techniques, namely histogram of gradients (HOG), local binary patterns (LBP), Scale Invariant Feature Transform (SIFT) and Speeded Up Robust Features (SURF) models are applied to extract the useful feature vectors from preprocessed image. At last, two ML based classification models namely random forest (RF) and decision tree (DT) are employed for classifying the input images into appropriate classes. A detailed experimental analysis takes place on benchmark dataset. The attained experimental outcome indicated that the HOG-RF model has achieved better results over the compared methods with the maximum sensitivity of 98.6%, specificity of 99.73%, F-measure of 99.02% and accuracy of 99.62%.