Fig 1.
Overview of cardiac electrophysiological models.
Table 1.
Example specific credibility-related considerations for patient-specific models as clinical tools.
Acronyms: UQ–uncertainty quantification; QOI–quantity of interest; COU–context of use; BZ–border zone.
Table 2.
Example specific credibility-related considerations for simulation studies that use a virtual cohort of patient-specific models.
Acronyms: UQ–uncertainty quantification; QOI–quantity of interest; COU–context of use; BZ–border zone.
Fig 2.
Five example UQ paradigms in patient-specific modeling.
The N virtual patients could be N members of a virtual cohort in PSM-VC, or N new patients for a PSM-CT. In the first row, there is no UQ; the personalized parameter varies across the subjects, the non-personalized parameter is constant. In the second row, uncertainty in the personalized parameter (e.g., due to measurement error) is accounted for–represented here as a probability distribution, though other methods for accounting for uncertainty could be used. In the third row, uncertainty in the non-personalized parameter (e.g., due to population variability) is also accounted for. In the fourth row, some patient information is used to constrain the distributions for the ‘non-personalized’ parameters (e.g., if the parameter is known to differ between sexes). In the final row (PSM-VC only) the non-personalized parameter is randomly sampled in the cohort members, rather than taking the population average value for all patients. Other options are possible.
Fig 3.
A): left ventricular meshes for 24 patients (red–border zone; black–scar). B) base timing location for one patient. C) simulated V1 ECG for all patients (note: apical pacing not sinus rhythm).
Table 3.
Details on meshes used in mesh resolution study.
Fig 4.
Discretization errors (defined as the difference between very low/low/medium results and the high-resolution results), for each patient, in normalized apex-to-base activation time (top row) and simulated ECG (middle and bottom rows). Note: Log scale used for top and middle rows. ECG lag errors (bottom row) take integer values, and since most results are exactly zero for the medium resolution meshes, a log scale is not used for ECG lag.
Table 4.
Discretization error coefficients of variation (COV), across patients, using different resolution meshes (columns) for normalized apex-to-base activation time and simulated ECG.
Fig 5.
A) repolarization gradient map for sample patient, pacing 0.2cm from scar. Colors represent repolarization gradient except scar is colored black. B) original mesh and meshes with expanded and contracted border zone for the same patient. Tissue is blue, BZ is red, scar is pink.
Fig 6.
Impact of uncertainty in border zone extent on high repolarization gradient volumes (HRGV).
A): boxplots comparing pacing at 0.2cm from scar with pacing 4.5cm from scar for each case. B): Range of HRGV values across each of the three cases (contracted BZ (stars), baseline (squares), expanded BZ (diamonds)) for each patient.
Fig 7.
Impact of uncertainty in tissue conductivity on high repolarization gradient volumes (HRGV).
A): boxplots comparing pacing at 0.2cm from scar with pacing 4.5cm from scar for each case. B): Range of HRGV values across each of the three cases (low (diamonds), baseline (squares), high conductivity (stars)) for each patient.
Table 5.
Summary of observations on considerations for PSM credibility assessment in relation to V&V40 credibility factors and gradations.
PSMs as clinical tools (PSM-CT) or PSM virtual cohort studies (PSM-VC) are considered separately.
Table 6.
Original V&V40 gradation for the ‘Discretization Error’ credibility factor, and example possible gradations for PSM-CT and PSM-VC.
Changes highlighted in bold. [Original V&V40 example gradations reprinted from ASME V&V 40–2018, by permission of The American Society of Mechanical Engineers. All rights reserved].