Rapid thrombogenesis prediction in COVID-19 patients using DNN with data labeling

Area of Science:

• Medical Imaging and Computational Biology
• Artificial Intelligence in Healthcare

Background:

• COVID-19 surge led to increased blood clotting incidence.
• Blood clots pose risks like cerebral hemorrhage, necessitating early detection.
• Current detection methods like CFD simulations are time-intensive.

Purpose of the Study:

• To develop a framework for early detection of blood clotting events in COVID-19 patients.
• To expedite the diagnostic process using AI and computational modeling.
• To optimize machine learning models for accurate thrombogenesis prediction.

Main Methods:

• Utilized patient-specific data from Computational Fluid Dynamics (CFD) simulations.
• Developed and optimized a deep neural network (DNN) model.
• Performed hyperparameter tuning and comparative simulations across machine learning algorithms.

Main Results:

• The DNN model demonstrated potential for expedited prediction of clotting events.
• Identified optimal data sizes for achieving high predictive accuracy.
• The framework integrates CFD simulations with machine learning for enhanced diagnostics.

Conclusions:

• The proposed framework offers a promising approach for early detection of thrombogenesis in COVID-19 patients.
• Combining computational modeling with AI accelerates diagnosis and management.
• This facilitates timely intervention for secondary complications.