HemoPIC digital twin automates brain perfusion mapping without manual AIF

HemoPIC tackles a key bottleneck in perfusion imaging: the reliance on manual arterial input function (AIF) selection followed by deconvolution. This conventional approach produces summary maps but lacks an executable temporal model for simulation or mechanistic insight. The authors—Yi-Chen Lee and Peirong Liu—introduce a physics-informed cerebral hemodynamics digital twin that explains perfusion time series through tracer mass conservation and a lumped parameter hemodynamic model. HemoPIC jointly estimates digital twin parameters and latent states from perfusion imaging data, eliminating manual AIF selection and deconvolution from routine quantification. The output includes clinically actionable perfusion indices such as CBF (cerebral blood flow), CBV (cerebral blood volume), and MTT (mean transit time), which are directly used in stroke and tumor diagnosis.

Experiments show HemoPIC reconstructs tracer dynamics, generates physiologically consistent perfusion maps with lesion hypoperfusion patterns, and satisfies central volume consistency. Critically, the approach yields a mechanistic hemodynamic digital twin that enables forward simulation and counterfactual intervention analysis—useful for exploring how changes in blood flow affect tissue before clinical trials. The code is publicly available on GitHub, making the method reproducible and extensible. This work bridges routine perfusion quantification and deeper hemodynamic modeling, offering a practical upgrade for both clinical workflows and neuroscience research.