AI

Simulation-based Inference for Cardiovascular Models

author
1 minute, 11 seconds Read


This paper was accepted on the workshop Machine Studying and the Bodily Sciences at NeurIPS 2023.

Over the previous a long time, hemodynamics simulators have steadily developed and have turn into instruments of selection for finding out cardiovascular techniques in-silico. This comes naturally at the price of growing complexity since state-of-the-art fashions are non-linear partial differential equations relying on many parameters. Whereas such instruments are routinely used to simulate hemodynamics given physiological parameters, fixing the associated inverse issues — mapping waveforms to physiological parameters — has obtained comparably much less consideration. Motivated by advances in simulation-based inference (SBI), we rethink the inverse issues specified by whole-body hemodynamics as statistical inferences. In opposition to conventional analyses, SBI supplies a multi-dimensional illustration of uncertainty for particular person measurements, as encoded by posterior distributions. We carry out an uncertainty evaluation in-silico on a targeted set of physiological parameters of scientific curiosity and evaluate a number of measurement modalities. Past the corroboration of identified information, such because the feasibility of estimating coronary heart fee, our research highlights the potential of estimating new physiological parameters from standard-of-care measurements. Moreover, SBI reveals virtually related information missed by various sensitivity analyses, such because the existence of sub-populations for which parameter estimation reveals distinct uncertainty regimes. Lastly, we research the hole between in-vivo and in-silico with the MIMIC-III waveform database and critically talk about how cardiovascular simulations can inform real-world knowledge evaluation.

Source link



Source link

Similar Posts

Leave a Reply

Your email address will not be published. Required fields are marked *