Robust OCT-based thenar skin thickness quantification via a dual-path extraction framework

Abstract

Accurate skin-layer thickness quantification is needed for depth calibration in multiple μ-spatially offset Raman spectroscopy (mμSORS)-based noninvasive glucose sensing. Existing OCT-based analyses often estimate skin thickness from depth-intensity profiles generated by large-scale A-line averaging. However, this averaging can attenuate the OCT-visible lower stratum corneum (SC) boundary and make SC thickness estimation unstable in some volumes. Here, we present a dual-path SC thickness quantification framework. The framework uses the conventional profile-based pathway when the SC-related profile feature is preserved, and uses a segmentation-based recovery pathway when profile cues are degraded. In the recovery pathway, SC thickness is computed from spatially explicit SC boundaries delineated on individual B-scans, thereby reducing information loss caused by global averaging. On an independent thickness-evaluation cohort of 26 participants with 156 manually annotated B-scans, the framework showed strong agreement with manual reference measurements, with a participant-level MAE of 5.17 μm (0.6466 pixels) and RMSE of 6.58 μm (0.8225 pixels). These results suggest that OCT-derived SC thickness can provide useful geometric information for mμSORS depth calibration, especially when conventional profile-based estimation is unstable.