Mining data from hemodynamic simulations for generating prediction and explanation models

Area of Science:

• Biomedical Engineering
• Medical Imaging
• Computational Fluid Dynamics

Background:

• Stroke is a leading cause of death, often linked to carotid bifurcation stenosis.
• Artery wall shear stress is a critical hemodynamic factor in stroke risk.
• Accurate detection of hemodynamic abnormalities is crucial for stroke prevention.

Purpose of the Study:

• To develop a prototype medical expert system for detecting increased artery wall shear stress.
• To aid medical experts in identifying hemodynamic abnormalities indicative of stroke risk.
• To enhance the diagnostic capabilities for carotid bifurcation stenosis.

Main Methods:

• Utilizing simulated data to train and validate predictive models.
• Applying methodologies for predicting wall shear stress magnitudes and locations.
• Implementing techniques for estimating the reliability of model predictions.
• Developing a user-friendly explanation module for model decisions.

Main Results:

• The system successfully predicted magnitudes and locations of maximum wall shear stress.
• Reliability estimation of the computed predictions was effectively implemented.
• The developed methodologies demonstrated utility in the target problem domain.
• The prototype showed potential for aiding medical experts in diagnosis.

Conclusions:

• The proposed medical expert system prototype is a valuable tool for detecting hemodynamic abnormalities.
• The methodologies employed can significantly aid in identifying stroke risk factors.
• Further development could enhance clinical application in stroke prevention and management.