Home // HEALTHINFO 2025, The Tenth International Conference on Informatics and Assistive Technologies for Health-Care, Medical Support and Wellbeing // View article
Authors:
Debora Beneduce
Guido Pagana
Fabrizio Bertone
Giuseppe Caragnano
Keywords: machine learning; convolutional neural network; teledermatology; skin ulcers monitoring.
Abstract:
This paper presents an automated pipeline for the detection, segmentation, and severity classification of cutaneous ulcers, addressing the clinical need for objective and remote wound monitoring. Despite increasing interest, real-time and interpretable Machine Learning tools in this domain remain scarce. We propose a hybrid solution combining classical image processing and Machine Learning techniques. Exploiting the guaranteed Convolutional Neural Network performance in binary segmentation tasks, a modified U-Net architecture, trained on grayscale digital images enhanced via Contrast Limited Adaptive Histogram Equalization, achieved high segmentation performance with an Intersection over Union of 0.82, Precision of 0.93, Recall of 0.89, and Dice coefficient of 0.88, using fewer than 2 million parameters. For severity classification, superpixel-wise brightness histograms were used to extract six discriminative features. A logistic regression model trained on these features reached a classification accuracy of 94%, effectively distinguishing between ulcer classes despite intra-class variability. The system offers robust performance with fast inference of 100 milliseconds per image and skin phototype-independence.
Pages: 35 to 44
Copyright: Copyright (c) IARIA, 2025
Publication date: October 26, 2025
Published in: conference
ISSN: 2519-8491
ISBN: 978-1-68558-312-5
Location: Barcelona, Spain
Dates: from October 26, 2025 to October 30, 2025