EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC 4.0

Overcoming measurement challenges in clinical practice: a deep learning-based approach to monocular surface area measurement

Alfonso Medela, Alberto Sabater, Gerardo Fernández, Taig Mac Carthy, Andy Aguilar, Diego Herrera, Meritxell Falqués, Antonio Martorell

Skin Health and Disease · 31.07.2026 · 6 (4) · 519-527

DOI: 10.1093/skinhd/vzag064 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Skin Health and Disease, online erschienen am 04.05.2026. Verfasst von Alfonso Medela und 7 weiteren Autor:innen. Beteiligte Einrichtung: Department of Medical Data Science, Legit.Health, Bilbao, Spain. Die Arbeit wurde bisher 2-mal zitiert (Crossref) und nennt 19 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC 4.0). Gefördert von: AI Labs Group, S.L.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Background

Precise area measurement of body surfaces is essential in clinical trials and clinical practice, particularly for the clinician, to monitor lesion size, assess skin condition severity and determine precise treatment dosage. The inherent difficulty of this task lies in the analysis of irregular surfaces or lesions with complex outlines. Consequently, existing solutions often lack accuracy or rely on expensive and complex technology, limiting their practicality.

Objectives

To introduce a cost-effective and accessible artificial intelligence (AI)-driven methodology that accurately measures body surface areas by effectively accounting for surface relief, outline and orientation.

Methods

The proposed approach is based on the analysis of monocular smartphone images. By using a set of specific markers, a state-of-the-art depth estimation model and surface triangulation, the area of delimited body surfaces can be measured with high precision.

Results

We tested this area measurement technology in controlled and simulated dermatological scenarios where a precise gold standard can be established. The proposed methodology achieves high precision and robustness across different camera configurations and body surfaces of different anatomical complexity. Notably, the proposed approach significantly outperforms traditional methods with a mean absolute error of 20 mm2 vs. 47 mm2.

Conclusions

The results consistently demonstrate high accuracy while being an accessible and time-efficient solution. When integrated into an AI-based medical device, this framework improves its usability and broadens its applications to tasks such as lesion severity assessment and clinical trial standardization, making it a valuable tool for clinicians and patients.

Autor:innen und Institutionen

  • Department of Medical Data Science, Legit.Health, Bilbao, Spain.
  • Department of Clinical Endpoint Innovation, Legit.Health, Bilbao, Spain.
  • Global Clinical Development, Almirall R&D, Sant Feliu de Llobregat, Spain.
  • Department of Dermatology, Hospital de Manises, Valencia, Spain.

Förderung

  • AI Labs Group, S.L

Zitieren

APA

Medela, A., Sabater, A., Fernández, G., Mac Carthy, T., Aguilar, A., Herrera, D., Falqués, M., & Martorell, A. (2026). Overcoming measurement challenges in clinical practice: a deep learning-based approach to monocular surface area measurement. Skin Health and Disease, 6(4), 519-527. https://doi.org/10.1093/skinhd/vzag064

Vancouver

Medela A, Sabater A, Fernández G, Mac Carthy T, Aguilar A, Herrera D, et al. Overcoming measurement challenges in clinical practice: a deep learning-based approach to monocular surface area measurement. Skin Health and Disease. 2026;6(4):519-527. doi:10.1093/skinhd/vzag064

BibTeX
@article{Medela2026,
  author = {Medela, Alfonso and Sabater, Alberto and Fernández, Gerardo and Mac Carthy, Taig and Aguilar, Andy and Herrera, Diego and Falqués, Meritxell and Martorell, Antonio},
  title = {Overcoming measurement challenges in clinical practice: a deep learning-based approach to monocular surface area measurement},
  journal = {Skin Health and Disease},
  year = {2026},
  volume = {6},
  number = {4},
  pages = {519--527},
  doi = {10.1093/skinhd/vzag064}
}
Kostenlos starten →