A U-Net model for epidermal segmentation in optical coherence tomography images of actinic keratosis

Sep 22, 2026

Upload by Miles

Actinic keratosis (AK) is a pre-cancerous skin lesion driven by excessive ultraviolet exposure. Optical coherence tomography (OCT) offers sub-surface structural cues for classification, yet manual interpretation is slow and varies across observers. This study develops and optimizes a U-Net deep-learning model for automated epidermal segmentation of AK in OCT images, balancing accuracy with computational efficiency. Multiple hyperparameter configurations are tested using Dice and Jaccard comparisons against expert annotations, with the best result at 256×256 resolution, batch size 2, and 50 epochs.

LanguageEnglish

CategoryHealthcare

FormatPDF

Pages15

Reading Time38 min

AccessFree to download

Abstract

Introduction

OPEN ACCESS Citation: Angelis T, Philipsen PA, Ortner VK, Fredman G, Haedersdal M, Untracht GR (2026) A U-Net model for epidermal segmentation in optical coherence tomography images of actinic keratosis. PLoS One 21(6): e0346059 . [https://doi.org/10.1371/journal.pone.0346059](https://doi.org/10.1371/journal.pone.0346059) Editor: Xiaohui Zhang, Bayer Crop Science United States: Bayer CropScience LP, UNITED STATES OF AMERICA Received: October 17, 2025 Accepted: March 15, 2026 Published: June 5, 2026 Copyright: © 2026 Angelis et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Data availability statement: All data and software underlying the manuscript has now been published online and is available at this DOI: [https://doi.org/10.11583/DTU.31282534](https://doi.org/10.11583/DTU.31282534) . Funding: MH received a research grant from Studies & Me for the COAKS study. The funders RESEARCH ARTICLE A U-Net model for epidermal segmentation in optical coherence tomography images of actinic keratosis Theofanis Angelis1,2*, Peter A. Philipsen1, Vinzent K. Ortner1, Gabriella Fredman1, Merete Haedersdal1,3, Gavrielle R. Untracht1,2 1 Department of Dermatology, Copenhagen University Hospital, Bispebjerg and Frederiksberg, Copenhagen, Denmark, 2 Department of Health Technology, Technical University of Denmark, Kongens Lyngby, Denmark, 3 Department of Clinical Medicine, Faculty of Health and Medical Science, University of Copenhagen, Copenhagen, Denmark * [tangelis@outlook.com](tangelis@outlook.com) Abstract Actinic keratosis (AK) is a pre-cancerous skin lesion typically caused by excessive exposure to ultraviolet light. Optical coherence tomography (OCT) can provide sub-surface information relevant for lesion classification, but manual analysis of images is time-consuming and subject to inter-observer variability; automated segmentation based on deep learning models can provide faster and more consistent results. However, structural changes, such as thickening and hyperkeratosis, complicate epidermal segmentation tasks. For this reason, we aimed to develop and optimize a U-Net model for the automated epidermal segmentation of AK lesions in OCT images, combining both accuracy and computational efficiency. Multiple configurations were evaluated by varying hyperparameters, including image sizes (256 × 256, 512 × 512, 1024 × 1024, and 464 × 1356 pixels), batch sizes (2, 4, 8, and 16), and training durations (50, 100, and 150 epochs) . Model performance was evaluated quantitatively using several metrics including the Dice coefficient and Jaccard index, comparing automated epidermal segmentations against expert annotations. The optimal configuration, with an image resolution of 256 × 256 pixels and a batch size of 2 over 50 epochs, had a Dice score of 0.86 and a Jaccard index of 0.76. Higher resolutions and longer training increased computation without significantly enhancing the accuracy values and sometimes caused overfitting. These findings show that a simple U-Net architecture could achieve efficient and accurate epidermal segmentation in AK lesions, provided it is fine-tuned to enhance performance. 1. Introduction Actinic keratosis (AK) is a common pre-cancerous skin lesion that may develop into squamous cell carcinoma (SCC) if it remains untreated [ 1] . It appears as solitary or PLOS One | [https://doi.org/10.1371/journal.pone.0346059](https://doi.org/10.1371/journal.pone.0346059) June 5, 2026 1 / 15 had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. The study was performed as a part of the Danish Research Center for Skin Cancer ([www.researchcenterforskin](www.researchcenterforskin)[cancer.org](cancer.org)) and the Skin Cancer Innovation Clinical Aca

FAQs

What problem does the U-Net model address in actinic keratosis OCT images?

It enables automated epidermal segmentation in OCT images to reduce the time burden and inter-observer variability of manual image analysis.

How were different model configurations evaluated?

The study varied image size, batch size, and training duration, then assessed performance quantitatively using Dice coefficient and Jaccard index against expert annotations.

What configuration produced the best segmentation performance?

The optimal setup used an image resolution of 256×256 pixels, batch size 2, and training over 50 epochs, achieving a Dice score of 0.86 and a Jaccard index of 0.76.

Sep 22, 2026

Upload by Miles