Diffusion-Weighted images analysis using regional convolutional neural network for stroke lesion classification.

Mohd Saad, Norhashimah and Abdullah, Abdul Rahim and Kandaya, Shaarmila and Mohd Noor, Nor Shahirah and Samsudin, Adam (2024) Diffusion-Weighted images analysis using regional convolutional neural network for stroke lesion classification. In: International Conference on Electrical, Computer and Energy Technologies (ICECET 2024), 25-27 July 2024.

Text
Difussion-Weighted Images Analysis Using Regional Convolutional Neural Network for Stroke Lesion Classification.pdf
Download (757kB)

Official URL: https://doi.org/10.1063/5.0228889

Abstract

Stroke is a major type of brain disorder, and the MRI sequence known as diffusion-weighted imaging (DWI) is often used to evaluate early changes related to strokes. Typically, stroke diagnosis requires manual assessment by neuroradiologists, which is subjective and time-consuming. Therefore, developing an automated stroke diagnosis system is crucial to streamline this process. This study presents a method for automatically detecting and categorizing stroke lesions using deep learning techniques applied to DWI. The dataset comprises 61 samples from Hospital Kuala Lumpur (HKL) and 69 samples from the public ISLES database, including acute hemorrhage, acute ischemic, chronic ischemic, and sub-acute ischemic strokes. The method employs a Regional Convolutional Neural Network (R-CNN) on DWI images, involving four stages: input image, region proposal network (RPN), computation of CNN features, and classification. Performance is evaluated using metrics such as accuracy, sensitivity, and specificity. The technique achieves a classification accuracy of 64.62% with an average training time of 16.3 seconds. Overall, this stroke classification method shows promise for enhancing the diagnosis and categorization of brain stroke lesions using deep learning.

Item Type:	Conference or Workshop Item (Paper)
Depositing User:	NURHASHIRAH BORHAN
Date Deposited:	28 Jul 2026 00:29
Last Modified:	28 Jul 2026 00:29
URI:	http://eprints.utem.edu.my/id/eprint/30082
Statistic Details:	View Download Statistic

Actions (login required)

View Item