Precision measurement of stratum corneum thickness in OCT images using variational autoencoders and advanced DSP techniques

TL;DR

This paper reviews how OCT imaging of skin has evolved, using deep learning and signal processing to better measure the outer skin layer for clinical use.

Contribution

The paper introduces hybrid frameworks combining physics-based signal processing and deep learning for more accurate and interpretable OCT analysis.

Findings

Hybrid models improve boundary detection and reduce annotation needs in OCT imaging.

These methods enhance clinical monitoring and evaluation of skin treatments.

Future directions include real-time analysis and explainable algorithms for personalized care.

Abstract

Optical coherence tomography (OCT) has emerged as a cornerstone technique for in vivo skin imaging; however, reliable and clinically meaningful quantification of stratum corneum (SC) thickness remains challenging. This review summarizes 2 decades of methodological evolution, highlighting the transition from early manual and rule-based approaches to modern deep learning–driven segmentation strategies. Particular emphasis is placed on recent hybrid frameworks that integrate physics-informed digital signal processing with generative deep learning models, which collectively improve boundary detection robustness, reduce annotation dependency, and enhance model interpretability. These advances have significantly expanded the clinical utility of OCT-based SC assessment, enabling more sensitive disease monitoring, improved evaluation of therapeutic and cosmetic interventions, and broader applications in dermatologic diagnostics. Finally, we outline emerging opportunities for real-time, marker-free analysis, multimodal data fusion, and the development of explainable and generalizable algorithms to support precision and personalized dermatologic care.

Figures