Enhancing Stroke Identification with Computational Analysis of Neuroimages using Machine Learning

Abstract

The early and accurate diagnosis of stroke is critical for effective treatment and improved patient outcomes. Traditional diagnostic methods often face challenges in achieving high accuracy and efficiency. In this study, we propose an innovative machine learning-based diagnostic model utilizing ResNet and MobileNet architectures to classify neuroimages into normal and stroke categories. We combined the best parts of two AI systems – ResNet's powerful image analysis and MobileNet's efficiency – to build our brain scan diagnostic tool. We trained it on a wide variety of brain images, using special preprocessing techniques to help it work well across different situations and deliver reliable results. Initial experiments demonstrate that ResNet achieves a training accuracy of 94% with normal images, while MobileNet achieves an impressive 92% training accuracy with normal images. These results highlight the potential of our proposed model to significantly improve the accuracy and speed of stroke diagnosis, providing a valuable tool for clinicians and healthcare providers. Future work will focus on further validation with larger datasets and real-world clinical trials to establish the model's efficacy and reliability in clinical settings. This study underscores the transformative potential of deep learning models in advancing stroke diagnosis and enhancing patient care.