Estimation of Contact PPG-Derived Vascular Indices from Reconstructed rPPG Signals

Remote photoplethysmography (rPPG) enables contactless cardiovascular monitoring using conventional cameras; however, its smooth waveform morphology limits the extraction of vascular indices that depend on fine pulse-wave characteristics. Consequently, most existing rPPG studies have focused primarily on heart rate estimation rather than morphological vascular assessment. This study investigates whether reconstructed signals can support the computation of contact-PPG (cPPG)-derived vascular indices and examines how agreement varies across timing-, amplitude-, and derivative-based indices. Fiducial and second derivative photoplethysmography (SDPPG) candidates were extracted from reconstructed signals to compute the Stiffness Index (SI), Reflection Index (RI), Augmentation Index (AIx), and an exploratory Simplified Aging Index (sAI) based on the detector-returned a, b, and e candidates. The resulting indices were compared with those derived from simultaneously acquired reference cPPG within pre-segmented, temporally paired cardiac cycles with known cycle boundaries. Reference cPPG fiducial locations were not used for representative window selection or reconstructed-signal fiducial localization. Reconstruction increased the number of cycles for which the adopted detector returned the candidates required for index computation. However, a detector audit showed that strict post-notch local maxima were present in only 5.02% of reconstructed cycles, indicating that candidate availability should not be interpreted as consistent recovery of a physiologically distinct diastolic peak. Under leave-one-subject-out cross-validation across all 28 participants, subject-level agreement under the adopted computational fiducial definition was highest for SI (r=0.963, ICC(2,1) =0.926) and for the height-independent, detector-defined systolic-to-post-notch delay (r=0.909, ICC(2,1) =0.837). An exploratory strict-local-peak sensitivity analysis retained only 37 matched cycles from four participants and descriptively showed weaker agreement than the fallback-inclusive analysis. Because this small subset contained repeated cycles from only four participants, it was considered insufficient for population-level inference and was not used to support subject-level claims regarding physiological diastolic-peak recovery. RI and AIx showed only weak associations whose 95% confidence intervals included zero, and sAI showed no meaningful agreement. Under the tested additive-noise conditions, detector-defined timing measures changed relatively little between the clean and 10 dB conditions, whereas amplitude-related measures and dicrotic notch localization degraded more markedly. Additional narrowing of the retained upper cutoff below 4 Hz produced relatively small changes, but wider-band conditions could not be evaluated using the stored signals. These findings support the use of cPPG-derived vascular indices as downstream measures for technically evaluating reconstructed rPPG morphology, but they do not constitute clinical validation or establish standalone non-contact vascular assessment.