Improved hypertensive stroke classification based on multi-scale feature fusion of head axial CT angiogram and...

Area of Science:

• Neurology
• Medical Imaging
• Artificial Intelligence

Background:

• Strokes, including ischemic and hemorrhagic types, are critical cardiovascular diseases.
• Computed tomography (CT) and computed tomography angiography (CTA) are standard for stroke diagnosis.
• Integrating imaging with clinical data, especially considering hypertension as an etiology, remains underexplored.

Purpose of the Study:

• To develop an automated classification model for distinguishing stroke types.
• To incorporate hypertension as an independent etiological factor in stroke classification.
• To leverage multimodal learning by combining brain imaging and clinical data for improved diagnostic accuracy.

Main Methods:

• A preprocessing workflow for head axial CT angiograms, including noise reduction and feature enhancement.
• Extraction of regions of interest from medical images.
• A multi-scale feature fusion model integrating location and deep features.
• A multimodal learning framework combining imaging and clinical data.

Main Results:

• The proposed models demonstrated superior performance compared to state-of-the-art methods on real-world data.
• The results highlight the significant potential of multimodal learning in diagnosing brain diseases.
• The model successfully classified stroke types using integrated imaging and clinical information.

Conclusions:

• The developed methodologies can form the basis for AI-driven diagnostic assistance technology for stroke categorization.
• This approach offers a pathway to more accurate and efficient stroke diagnosis.
• Further extension of these techniques can enhance clinical decision-making in neurology.