The utility and reliability of a deep learning algorithm as a diagnosis support tool in head & neck non-melanoma skin malignancies.

Autor: Medela A; Medical Data Science, Legit.Health, Bilbao, Spain., Sabater A; Medical Data Science, Legit.Health, Bilbao, Spain., Montilla IH; Medical Data Science, Legit.Health, Bilbao, Spain., MacCarthy T; Clinical Endpoint Innovation, Legit.Health, Bilbao, Spain., Aguilar A; Clinical Endpoint Innovation, Legit.Health, Bilbao, Spain., Chiesa-Estomba CM; Department of Otorhinolaryngology-Head & Neck Surgery, Osakidetza, Donostia University Hospital, San Sebastian, 20014, Spain. chiesaestomba86@gmail.com.; Bioguipuzkoa Health Research Institute, San Sebastian, 20014, Spain. chiesaestomba86@gmail.com.; Faculty of Medicine, Deusto University, Bilbao, Spain. chiesaestomba86@gmail.com.; Servicio de Otorrinolaringología - Cirugía de Cabeza y Cuello, Hospital Universitario Donostia, Paseo Dr. Begiristain #1. CP, San Sebastian- Donosti, 20014, España. chiesaestomba86@gmail.com.
Jazyk: angličtina
Zdroj: European archives of oto-rhino-laryngology : official journal of the European Federation of Oto-Rhino-Laryngological Societies (EUFOS) : affiliated with the German Society for Oto-Rhino-Laryngology - Head and Neck Surgery [Eur Arch Otorhinolaryngol] 2024 Sep 06. Date of Electronic Publication: 2024 Sep 06.
DOI: 10.1007/s00405-024-08951-z
Abstrakt: Objective: The incidence of non-melanoma skin cancers, encompassing basal cell carcinoma (BCC) and cutaneous squamous cell carcinoma (cSCC), is on the rise globally and new methods to improve skin malignancy diagnosis are necessary. This study aims to assess the performance of a CE-certified medical device as a diagnosis support tool in a head & neck (H&N) outpatient clinic, specifically focusing on the classification of three key diagnostics: BCC, cSCC, and non-malignant lesions (such as Actinic Cheilitis, Actinic Keratosis, and Seborrheic Keratosis).
Methods: a prospective, longitudinal, non-randomized study was designed to evaluate the performance of a deep learning-based method as a diagnosis tool in a group of patients referred to the head & neck clinic for suspicious skin lesions.
Results: 135 patients were included, 92 (68.1%) were male and 43 (31.9%) were female. The median age was 71 years +/- 9 (Min: 56/Max: 91). Of those, 108 were malignant pathologies (54 basal cell carcinoma and 54 squamous cell carcinoma) and 27 benign pathologies (14 seborrheic keratoses, 2 actinic keratoses, and 11 actinic cheilitis). Of special significance is the remarkable performance of the medical device in identifying malignant lesions (basal cell carcinoma and squamous cell carcinoma) within the top-5 most likely diagnoses in above 90% of cases, underscoring its potential utility for early diagnosis and treatment.
Conclusion: In this study, the effectiveness of deep learning methods, with a particular focus on vision transformers, as a diagnostic aid for H&N cutaneous non-melanoma skin cancers was demonstrated, highlighting its potential value for early detection and treatment of non-melanoma skin cancers. In this vein, further research is needed in the future to elucidate the role of this technology, because of its potential in the primary care clinic, dermatology, and head & neck surgery clinic as well as in patients with suspicious lesions, as a self-exploration tool.
(© 2024. The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature.)
Databáze: MEDLINE
Externí odkaz: