Robustness of a convolutional neural network trained on dermoscopic images and challenged with close‐up images
JDDG: Journal der Deutschen Dermatologischen Gesellschaft · 2025
DOI: 10.1111/ddg.15900 ↗Kurz gefasst
Fachartikel in der Zeitschrift JDDG: Journal der Deutschen Dermatologischen Gesellschaft, 2025. Verfasst von Anastasia Sophie Vollmer und 6 weiteren Autor:innen. Beteiligte Einrichtung: Department of Dermatology, University medical Center Heidelberg, Heidelberg, Germany. Hauptthemen laut MEDLINE: Dermoscopy, Neural Networks, Computer, Skin Neoplasms, Deep Learning. Die Arbeit nennt 35 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Background and Objectives Deep learning‐convolutional neural networks (DL‐CNNs) have demonstrated high diagnostic accuracy within the domain of dermoscopy. However, many clinical settings lack dermoscopic devices, requiring reliance on close‐up images. This study evaluates the robustness of a DL‐CNN trained on dermoscopic images when challenged with close‐up images.
Methods
In this cross‐sectional study, the DL‐CNN Moleanalyzer pro, trained on 129,487 dermoscopic images, was tested on 350 skin lesions, each imaged both clinically and dermoscopically. Histopathology (89.4%) or expert consensus with a two‐year follow‐up (10.6%) served as the reference standard. Primary outcomes included sensitivity, specificity, and the receiver operating characteristic‐area under the curve (ROC‐AUC).
Results
For dermoscopic images, the DL‐CNN achieved a sensitivity of 88.2% (95% CI: 82.9%–92.0%), specificity of 69.0% (61.4%–75.8%), and a ROC‐AUC of 0.866 (0.860–0.873). For close‐up images of the same lesions, sensitivity decreased to 60.5% (53.5%–67.1%, p < 0.001), while specificity increased to 79.4% (72.3%–85.0%, p = 0.027). The ROC‐AUC for close‐up images was 0.780 (0.772–0.790, p = 0.003).
Conclusions
Our findings highlight the diagnostic limitations of applying DL‐CNNs trained on dermoscopic images to close‐up images, with reduced sensitivity but increased specificity. The results emphasize the need to improve cross‐domain adaptability. Fine‐tuning of DL‐CNNs on clinical images may enhance real‐world diagnostic accuracy.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Stichwörter der Autor:innen
Autor:innen und Institutionen
- Department of Dermatology, University medical Center Heidelberg, Heidelberg, Germany.
- Department of Dermatology, Allergology and Environmental Medicine II, Hospital Thalkirchner Street, Munich, Germany.
- Department of Genetic Epidemiology, University Medical Center, Georg-August University of Goettingen, Goettingen, Germany.
ORCID-Profile: Anastasia Sophie Vollmer, Katharina Susanne Kommoss, Holger Andreas Haenssle
Zitieren
APA
Vollmer, A. S., Winkler, J. K., Kommoss, K. S., Stolz, W., Rosenberger, A., Enk, A., & Haenssle, H. A. (2025). Robustness of a convolutional neural network trained on dermoscopic images and challenged with close‐up images. JDDG: Journal der Deutschen Dermatologischen Gesellschaft. https://doi.org/10.1111/ddg.15900
Vancouver
Vollmer AS, Winkler JK, Kommoss KS, Stolz W, Rosenberger A, Enk A, et al. Robustness of a convolutional neural network trained on dermoscopic images and challenged with close‐up images. JDDG: Journal der Deutschen Dermatologischen Gesellschaft. 2025. doi:10.1111/ddg.15900
BibTeX
@article{Vollmer2025,
author = {Vollmer, Anastasia Sophie and Winkler, Julia Katharina and Kommoss, Katharina Susanne and Stolz, Wilhelm and Rosenberger, Albert and Enk, Alexander and Haenssle, Holger Andreas},
title = {Robustness of a convolutional neural network trained on dermoscopic images and challenged with close‐up images},
journal = {JDDG: Journal der Deutschen Dermatologischen Gesellschaft},
year = {2025},
doi = {10.1111/ddg.15900}
}
EUVIMED