Towards effective classification of brain hemorrhagic and ischemic stroke using CNN

Anjali Gautam, Balasubramanian Raman

Abstract: Highlights•Classification of brain CT scan image into hemorrhagic, ischemic and normal has been performed by our newly proposed CNN method which uses image fusion for better classification results.•The work is based on real datasets which have been collected from Himalayan Institute of Medical Sciences (HIMS), Dehradun, India. The datasets consists of ischemic, hemorrhagic and normal CT scan images.•Experiments have been carried out by categorizing images in two and three different categories. The first dataset consists of ischemic and hemorrhagic stroke images and the second dataset include one more category i.e. normal CT scan images of brain.•Experimental results show that proposed CNN approach gives better performance over AlexNet and ResNet50.•The accuracy obtained by the proposed method after 10 fold cross-validation of the CT image dataset is 98.77% on the first dataset and 93.33% on the second dataset.