Synthetic hematocrit-based extracellular volume quantification from photon-counting CT: Validation against MRI and systematic bias correction.

Purpose: To assess the feasibility and accuracy of synthetic hematocrit (Hct)-based myocardial extracellular volume (ECV) quantification using photon-counting detector (PCD)-CT late enhancement (LE) imaging, with cardiac MRI as the reference standard.

Methods: In this post-hoc analysis of a prospective study, patients underwent same-day cardiac MRI and PCD-CT LE imaging. MRI-based ECV (ECV) was computed using T1 mapping sequences before and after contrast administration. PCD-CT-based ECV (ECV) was calculated using synthetic Hct derived from blood pool attenuation in virtual non-contrast (VNC) images of LE scans. Agreement between ECV and ECV was evaluated using Bland-Altman analysis, Pearson correlation, and intraclass correlation coefficient (ICC) with absolute agreement. A linear regression model was applied to correct for systematic bias in ECV, generating bias-adjusted ECV values (ECV).

Results: A total of 27 patients (mean age 52.9 ± 17.2 years, 52 % female) were included. ECV systematically overestimated ECV (42.1 ± 8.7 % vs. 33.8 ± 8.1 %, p < 0.001), with a mean bias of -8.3 percentage points and wide limits of agreement (-7.3 to 23.9). Correlation between ECV and ECV was moderate (r = 0.56, p = 0.002), and ICC indicated poor reliability (ICC = 0.38). Bias correction using linear regression eliminated systematic error (ECV: 35.0 ± 4.8 %, p = 0.373), with a reduced mean bias of -1.2 percentage points, but wide limits of agreement remained (-14.4 to 12.1). ICC improved slightly to 0.51, indicating moderate reliability.

Conclusion: While synthetic Hct-based ECV estimation using PCD-CT is technically feasible, it systematically overestimates MRI-derived ECV and demonstrates considerable inter-individual variability. Bias correction effectively eliminates systematic error, but the improvement in variability remains limited.