Publication: Automated detection and evaluation of ischemic stroke on ct brain imaging using machine learning techniques
Loading...
Date
2025-06
Authors
Sharuddin, Nur Amirah Atikah
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
This study investigates the application of machine learning algorithms for the detection of ischemic stroke using CT brain images. Stroke, particularly ischemic stroke, remains a leading cause of death and disability globally. The early detection and diagnosis of ischemic stroke are crucial for minimizing long-term damage and improving patient outcomes. Traditional methods of diagnosis rely on the expertise of radiologists, which can be time-consuming and prone to inter-observer variability. This research aims to develop an automated system for ischemic stroke detection by leveraging machine learning techniques such as Support Vector Machine (SVM), K-Nearest Neighbours (KNN), and Random Forest (RF), applied to CT brain images. The study uses a dataset consisting of 397 ischemic stroke CT images and 25 normal brain scans. A series of preprocessing steps, including resizing, normalization, and noise reduction, were performed on the CT images to ensure they were suitable for machine learning analysis. Relevant features were extracted from the images, such as intensity, texture, and shape, which were then used to train the machine learning models. The performance of the models was evaluated using metrics such as accuracy, precision, recall, F1-score, and AUC. The Random Forest model achieved the highest accuracy at 92.76%, with an AUC of 0.973, outperforming both the KNN and SVM models. The KNN model achieved an accuracy of 93.93% with an AUC of 0.940, while the SVM model achieved an accuracy of 87.87% with an AUC of 0.984. Additionally, the training time for each model was recorded: SVM took 0.0152 seconds, KNN took 0.0114 seconds, and Random Forest took 0.2083 seconds. The results demonstrate that machine learning models, particularly Random Forest and KNN, can provide accurate and consistent stroke detection, offering potential for rapid and reliable clinical application, with KNN being the fastest in training time.
Description
Keywords
-