Medical Imaging Technology

Permanent URI for this collection

http://dspace.alquds.edu:4000/handle/20.500.12213/6128

Browse

Browsing Medical Imaging Technology by Author "Ibraheem Bassam Ibraheem Qdaih"

Now showing 1 - 1 of 1
  • Thumbnail Image
    Item
    Hybrid Artificial Intelligence Model for Prediction and Classification Across all Stages of Brain Ischemic Stroke in Non-enhanced Computerized Tomography Images
    (Al-Quds University, 2024-05-26) Ibraheem Bassam Ibraheem Qdaih; إبراهيم بسام إبراهيم قديح
    Stroke is a major global health issue, resulting in significant mortality and disability among approximately 16 million people annually. Rapid response is crucial to mitigate brain damage and improve patient outcomes. Strokes, which are primarily categorized into ischemic and hemorrhagic types, vary in presentation and can be influenced by modifiable risk factors. In regions like Palestine, with limited economic resources, stroke is a prevalent cause of death. Diagnostic challenges are heightened by the limitations of brain non-enhanced CT (B-NECT) scans, which vary in effectiveness based on the stroke's stage. This study introduces a novel artificial intelligence-based framework, the Stroke Precision Enhancement Model (SPEM), which employs image processing, deep learning, and machine learning techniques to enhance the classification of ischemic stroke stages in B-NECT images. This real-time hybrid model integrates Contrast Limited Adaptive Histogram Equalization (CLAHE) for preprocessing, with feature extraction conducted through Densely Connected Convolutional Networks-121 (DenseNet-121). Classification is performed using Support Vector Machine (SVM), Random Forest (RF), and Logistic Regression (LR), with a focus on determining the most effective method based on various performance metrics. The results indicate exceptional performance of the SPEM, especially when combining DenseNet-121 with the LR classifier. Notably, in the hyper-acute stage, the model achieved an accuracy of 0.9957, a precision of 0.9914, and a remarkable Area Under the Receiver Operating Characteristic Curve (AUC) of 0.9999, with a processing time of just 0.04 seconds. Similar high performance was maintained across other stroke stages. These findings highlight the potential of this AI-enhanced model in facilitating faster and more accurate clinical decisions for early-stage stroke treatment. The hybrid model shows promise in predicting and classifying ischemic strokes and could significantly impact clinical practice upon further research, validation on larger datasets, enhanced interpretability, and integration into clinical workflows