Modal complexity and power loss in patient-specific Fontan cavopulmonary connection flow

Power loss (PL) across the total cavopulmonary connection (TCPC) varies among Fontan patients, but scalar energetic metrics cannot describe how unsteady flow structures organize within the connection. We evaluated whether volume-weighted proper orthogonal decomposition (POD) provides structure-resolved descriptors of TCPC flow organization and whether truncated POD reconstructions preserve PL. Fifteen patient-specific TCPC anatomies were reconstructed from cardiac magnetic resonance imaging and simulated using unsteady computational fluid dynamics with pulsatile caval inflow and clinically measured pulmonary flow splits. POD was applied to velocity snapshots from the final cardiac cycle. We quantified modal organization using the energy-capture mode count Nε and normalized Shannon entropy H. Statistical analyses enabled comparison of PL, indexed power loss (iPL), and POD metrics. PL was associated with broader POD spectra: PL correlated with N0.995 (ρs = 0.812, p-value = 0.0002) and H(ρs = 0.650, p-value = 0.0108). In contrast, iPL associations were weaker and less precise, including iPL–N0.995 (ρs = 0.479, p-value = 0.075) and iPL–H (ρs = 0.218, p-value = 0.428). PL reconstruction error decreased from a median of 1.59% at N0.995 to below 0.01% at N0.999. These results show that volume-weighted POD identifies differences in TCPC fluctuation organization that are more clearly associated with PL than iPL, and that PL preservation provides a useful metric-level test for POD truncation.