A Physics-Integrated Deep Learning Approach for Patient-Specific Non-Newtonian Blood Viscosity Assessment using PPG

A physics-integrated deep learning approach using a hybrid 1D CNN-LSTM neural network to estimate patient-specific non-Newtonian blood viscosity from photoplethysmography (PPG) signals. The optimized model achieves 81.1% overall accuracy with 84.0% in the physiological shear range. IRB-approved study conducted at Severance Hospital, Yonsei University.

Computer Methods and Programs in Biomedicine, Vol. 265, 108740