Ischemic Brain Stroke Detection from MRI Image using Logistic Regression Classifier

In this paper an efficient model for ischemic brain stroke detection from magnetic resonance imaging (MRI) using machine learning approach namely logistic regression classifier is proposed. The MRI images are pre-processed to reduce noise and converted into gray images. Then the stroke portions of the MRI gray images are segmented by using hue, saturation, and value (HSV) color threshold and the segmented stroke images are converted into binary images to reduce computational complexity. The stroke features namely mean hue, standard deviation, mean variance and area of affected lesion i.e. stroke portion have been extracted. Finally, logistic regression classifier is used to identify the classes of test image. The proposed model shows an accuracy of 96%, sensitivity of 92.3% and specificity of 100% for testing datasets.