Physics-Informed Neural Network-Based Pulsatile Flow Modeling and Targeted Drug Delivery Optimization in Computational Hemodynamics.

Background: Targeted drug delivery in vascular diseases requires accurate modeling of blood flow dynamics. This study utilizes physics-informed neural networks (PINNs) to simulate pulsatile hemodynamics and optimize delivery timing for maximum therapeutic efficacy. Methods: A PINN framework was developed to solve the Navier-Stokes and convection-diffusion equations in a reconstructed arterial domain. Pulsatile inlet conditions were imposed to replicate physiological blood flow. Drug bolus injections were simulated at varying phases of the cardiac cycle to evaluate optimal timing. Results: The model achieved high accuracy with under 2% relative error compared to finite element benchmarks. Maximum drug accumulation (78%) at the target site occurred when injected 0.2 seconds post-systole, with minimal off-target dispersion (8%). Hemodynamic parameters, such as peak velocity (0.65 m/s) and wall shear stress (2.5 Pa), were consistent with physiological norms. Conclusion: PINNs offer a robust, data-efficient approach for simulating vascular dynamics and optimizing personalized drug delivery strategies.