Automated Detection of Psoriasis and Aging Biomarkers Using Machine Learning on Optoacoustic Mesoscopy Images
Year: 2025
Type: article
Abstract: Ultra-wideband raster-scan optoacoustic mesoscopy (RSOM) is an advanced imaging modality that has shownexceptional capability in visualizing in-vivo epidermal and dermal structures with high resolution. Despite its promise, theautomatic and quantitative analysis of three-dimensional RSOM datasets remains largely unaddressed. In this study, weintroduce DeepRAP (Deep Learning RSOM Analysis Pipeline), a novel framework designed to analyze and quantifymorphological skin features from RSOM images and extract clinically relevant imaging biomarkers for diseasecharacterization. DeepRAP employs a multi-network segmentation strategy based on convolutional neural networks (CNNs)enhanced through transfer learning. This architecture facilitates the automatic identification of skin layers and precisesegmentation of the dermal microvasculature, achieving performance on par with expert human annotation. The frameworkwas validated against manual segmentation using RSOM data from 25 psoriasis patients undergoing treatment. The extractedbiomarkers successfully characterized disease severity and progression, showing a strong correlation with physicianassessments and histological data. In a distinct validation experiment, DeepRAP was applied to a timeseries dataset capturingocclusion-induced hyperemia in 10 healthy volunteers. The framework effectively tracked changes in microvascularbiomarkers during occlusion and subsequent reperfusion, demonstrating both high accuracy and reproducibility.Additionally, analysis of a cohort of 75 individuals revealed a significant association between microvascular features in thedermal layer and age, with fine vascular patterns showing the strongest age-related correlation. These findings highlightDeepRAP’s potential to automate and accelerate in-vivo skin analysis, offering a non-invasive alternative to traditionalbiopsybased methods. The framework enhances the clinical and translational relevance of RSOM by enabling highthroughput, quantitative assessment of skin morphology and vascular health.
Authors: Jemimah, Jeyasutha, Associate Professor Dr. M., Dhanaseelan, Professor Dr. F. Ramesh
Language: English
Subfield: Biomedical Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W6949658552
FWCI:
Cites:
Cited by:
Related to: