An automated early ischemic stroke detection system using CNN deep learning algorithm

Abstract

Over the past few years, stroke has been among the top ten causes of death in Taiwan. Stroke symptoms belong to an emergency condition, the sooner the patient is treated, the more chance the patient recovers. However, the location of ischemic stroke in the CT image is not obvious, so the diagnosis need to rely on doctors to assess the image. The purpose of this paper is to develop an automated early ischemic stroke detection system using CNN deep learning algorithm. After entering the CT image of the brain, the system will begin image preprocessing to remove the impossible area which is not the possible of the stroke area. Then we will select the patch images and use Data Augmentation method to increase the number of patch images. Finally, we will input the patch images into the convolutional neural network for training and testing. In this paper, we used 256 patch images to train and test a CNN module that it had the ability to recognize the ischemic stroke. From the experimental results, we can find that the accuracy of the proposed method is higher than 90%. It means that the method proposed in this paper can effectively assist the doctor to diagnose.