Empirical Evaluation of Artificial Intelligence (AI) Powered Brain Stroke Detection Methodology by Using Learning Based Image Processing Techniques

Stroke is a leading cause of death and disability worldwide, making early detection and accurate diagnosis critical for patient outcomes. This study presents a Hybrid Learning Assisted Tumor Detection Scheme (HLTDS) for stroke detection from brain CT images. The proposed model integrates Convolutional Neural Networks (CNNs) with advanced feature extraction techniques, including AutoEncoders and edge detection, to enhance accuracy and computational efficiency. HLTDS was evaluated against nine existing models: CNN, ResNet50, VGG16, MobileNet, DenseNet, Inception, Xception, and EfficientNet, based on key performance metrics such as accuracy, precision, recall, and F1-score. The results demonstrate that HLTDS outperforms all compared models with an accuracy of 96.45%, achieving the highest precision (95.82%) and recall (94.67<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup>), making it the most reliable model for identifying stroke-related anomalies. The model's efficiency in terms of training time (58 minutes) and inference time (145 milliseconds) positions it as an optimal choice for real-time clinical applications. Furthermore, HLTDS achieves a balance between model size and memory usage, making it suitable for deployment in resource-constrained environments. These results highlight the importance of hybrid learning approaches in improving stroke detection while ensuring scalability. Future work will focus on expanding the model's capability to handle multiclass classification and further reducing computational costs.