<![CDATA[The paper presents an automated and computationally lightweight system that aims at diagnosing the presence of the ischemic stroke in the brain based on the MRI images and prioritizes mathematical clarity over the overpowering complexity. Workflow starts with AWCES algorithm to improve the quality of images, which is applied based on adaptive windowing determined by the entropy, and then proceeds to Watershed segmentation based on markers to word out the areas that are suspected of being vascularly blocked. The texture descriptors are then obtained using the Local Binary Patterns (LBP) and sent to a Random Forest classifier which differentiates between the damaged and the healthy brain tissue. The training stage has been integrated with the SMOTE technique to address the problem of class imbalance that is severe in the dataset. With a stratified five-fold cross-validation, the system demonstrated an AUC of 0.99, a specificity of 94% and a recall of 70%. These results indicate that the classical methods based on open mathematics can be competitive with deep learning networks, providing a high-quality and quick diagnostic methodology that could be used as a primary solution in the diagnosis of a stroke and as a complement to the exploratory three-dimensional visualization.]]>
Copyrights © 2026
