Brain Stroke Detection Using Convolutional Neural Network and Deep Learning Models

Abstract

For the last few decades, machine learning is used to analyze medical dataset. Recently, deep learning technology gaining success in many domain including computer vision, image recognition, natural language processing and especially in medical field of radiology. This research attempts to diagnose brain stroke from MRI using CNN and deep learning models. The proposed methodology is to classify brain stroke MRI images into normal and abnormal images and delineate abnormal regions using semantic segmentation [4]. In particular, two types of convolutional neural network that are LeNet [2] and SegNet are used. For classification, we passed pre-processed stroke MRI for training, trained all layers of LeNet and classify normal and abnormal patient. Then this abnormal patient data stored into two dimensional array and passed this two dimensional array to SegNet which is auto encoder decoder [3] model for segmentation, trained all layers of SegNet except fully connection layer. The experimental result show that classification model achieve accuracy between 9697% and segmentation model achieve accuracy between 8587%.Through experimental results, we found that deep learning models not only used in non-medical images but also give accurate result on medical image diagnosis, especially in brain stroke detection.