Dermatologist versus artificial intelligence confidence in dermoscopy diagnosis: Complementary information that may affect decision‐making
Experimental Dermatology · 2023
DOI: 10.1111/exd.14892 ↗Kurz gefasst
Vergleichsstudie in der Zeitschrift Experimental Dermatology, 2023. Verfasst von Pieter Van Molle und 7 weiteren Autor:innen. Beteiligte Einrichtung: IDLab, Department of Information Technology, Ghent University-IMEC, Ghent, Belgium. Hauptthemen laut MEDLINE: Artificial Intelligence, Dermatologists, Dermoscopy, Skin Diseases. Die Arbeit wurde bisher 6-mal zitiert (Crossref) und nennt 29 Literaturangaben. Gefördert von: Agentschap Innoveren en Ondernemen.
Abstract (Originalsprache)
In dermatology, deep learning may be applied for skin lesion classification. However, for a given input image, a neural network only outputs a label, obtained using the class probabilities, which do not model uncertainty. Our group developed a novel method to quantify uncertainty in stochastic neural networks. In this study, we aimed to train such network for skin lesion classification and evaluate its diagnostic performance and uncertainty, and compare the results to the assessments by a group of dermatologists. By passing duplicates of an image through such a stochastic neural network, we obtained distributions per class, rather than a single probability value. We interpreted the overlap between these distributions as the output uncertainty, where a high overlap indicated a high uncertainty, and vice versa. We had 29 dermatologists diagnose a series of skin lesions and rate their confidence. We compared these results to those of the network. The network achieved a sensitivity and specificity of 50% and 88%, comparable to the average dermatologist (respectively 68% and 73%). Higher confidence/less uncertainty was associated with better diagnostic performance both in the neural network and in dermatologists. We found no correlation between the uncertainty of the neural network and the confidence of dermatologists ( R = −0.06, p = 0.77). Dermatologists should not blindly trust the output of a neural network, especially when its uncertainty is high. The addition of an uncertainty score may stimulate the human‐computer interaction.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Stichwörter der Autor:innen
Autor:innen und Institutionen
- IDLab, Department of Information Technology, Ghent University-IMEC, Ghent, Belgium.
- Department of Dermatology, Ghent University Hospital, Ghent, Belgium.
- Cancer Research Institute Ghent (CRIG), Ghent, Belgium.
ORCID-Profile: Pieter Van Molle, Sofie Mylle, Tim Verbelen, Cedric De Boom, Bert Vankeirsbilck, Evelien Verhaeghe, Bart Dhoedt, Lieve Brochez
Förderung
- Agentschap Innoveren en Ondernemen
Zitieren
APA
Van Molle, P., Mylle, S., Verbelen, T., De Boom, C., Vankeirsbilck, B., Verhaeghe, E., Dhoedt, B., & Brochez, L. (2023). Dermatologist versus artificial intelligence confidence in dermoscopy diagnosis: Complementary information that may affect decision‐making. Experimental Dermatology. https://doi.org/10.1111/exd.14892
Vancouver
Van Molle P, Mylle S, Verbelen T, De Boom C, Vankeirsbilck B, Verhaeghe E, et al. Dermatologist versus artificial intelligence confidence in dermoscopy diagnosis: Complementary information that may affect decision‐making. Experimental Dermatology. 2023. doi:10.1111/exd.14892
BibTeX
@article{VanMolle2023,
author = {Van Molle, Pieter and Mylle, Sofie and Verbelen, Tim and De Boom, Cedric and Vankeirsbilck, Bert and Verhaeghe, Evelien and Dhoedt, Bart and Brochez, Lieve},
title = {Dermatologist versus artificial intelligence confidence in dermoscopy diagnosis: Complementary information that may affect decision‐making},
journal = {Experimental Dermatology},
year = {2023},
doi = {10.1111/exd.14892}
}
EUVIMED