Deep Learning Approaches for Stroke Detection Using CNN and Transfer Learning Techniques

Abstract

to enhance the diagnosis of stroke disease through medical image classification. Leveraging the power of neural networks, we conducted a comparative study involving popular transfer learning techniques such as VGG16, VGG19, and ResNet-50. While these complex architectures demonstrated their potential, we also recognized the pitfalls of deep models, leading us to introduce a simpler Convolutional Neural Network (CNN) model. Through rigorous evaluation, our proposed CNN model achieved an exceptional accuracy of 99.60%, underscoring the efficacy of a streamlined approach. Our findings emphasize the balance between sophisticated methodologies and pragmatic solutions, showcasing how AI can significantly impact medical diagnoses. As a practical application, we developed an intuitive application that enables users to classify medical images, bridging the gap between AI advancements and real-world medical practices. This project contributes to the advancement of AI in healthcare, showcasing the potential for accurate and efficient stroke disease diagnosis.