Mimicry of whole-body blood circulation through genetic algorithm in reduced-order hemodynamic model

Physics-based reduced-order hemodynamic models have garnered significant interest because of their ability to capture whole-body cardiovascular fluctuations. However, coordinating the numerous interdependent parameters within these models remains a long-standing challenge, and the demand for the personalization of these models persists. We constructed a complex whole-body model of blood circulation (containing the heart, arterial trunk, and branches) and utilized genetic algorithms to automatically and efficiently coordinate the model parameters. Additionally, we introduced a “pseudo-distance” metric by updating the derivative dynamic time-warping algorithm to evaluate the similarity between the simulated waveforms and the target waveforms. After 40 rapid iterations, a complete match was achieved with the target in terms of the blood pressure and flow waveforms amplitude as well as the time domain, resulting in highly realistic waveform mimicry (i.e., the pseudo-distance approached zero). This model takes about 40 min, far less than the manual modeling that usually takes several months. These results indicate that GAs significantly improve the modeling efficiency of reduced-order models, thus lowering the user threshold.