Generalizability and usefulness of artificial intelligence for skin cancer diagnostics: An algorithm validation study
Publikation: Bidrag til tidsskrift › Tidsskriftartikel › Forskning › fagfællebedømt
Standard
Generalizability and usefulness of artificial intelligence for skin cancer diagnostics: An algorithm validation study. / Ternov, Niels K.; Christensen, Anders N.; Kampen, Peter J. T.; Als, Gustav; Vestergaard, Tine; Konge, Lars; Tolsgaard, Martin; Hölmich, Lisbet r.; Guitera, Pascale; Chakera, Annette H.; Hannemose, Morten R.
I: JEADV Clinical Practice, Bind 1, Nr. 4, 2022, s. 344-354.Publikation: Bidrag til tidsskrift › Tidsskriftartikel › Forskning › fagfællebedømt
Harvard
APA
Vancouver
Author
Bibtex
}
RIS
TY - JOUR
T1 - Generalizability and usefulness of artificial intelligence for skin cancer diagnostics: An algorithm validation study
AU - Ternov, Niels K.
AU - Christensen, Anders N.
AU - Kampen, Peter J. T.
AU - Als, Gustav
AU - Vestergaard, Tine
AU - Konge, Lars
AU - Tolsgaard, Martin
AU - Hölmich, Lisbet r.
AU - Guitera, Pascale
AU - Chakera, Annette H.
AU - Hannemose, Morten R.
PY - 2022
Y1 - 2022
N2 - BackgroundArtificial intelligence can be trained to outperform dermatologists in image-based skin cancer diagnostics. However, the networks' sensitivity to biases and overfitting may hamper their clinical applicability.ObjectivesThe aim of this study was to explain the potential consequences of implementing convolutional neural networks for stand-alone melanoma diagnostics and skin lesion triage.MethodsIn this algorithm validation study on retrospective data, we reproduced and evaluated the performance of state-of-the-art artificial intelligence (convolutional neural networks) for skin cancer diagnostics. The networks were trained on 25,331 annotated dermoscopic skin lesion images from an open-source data set (ISIC-2019) and tested using a novel data set (AISC-2021) consisting of 26,591 annotated dermoscopic skin lesion images. We tested the trained algorithms' ability to generalize to new data and their diagnostic performance in two simulations (melanoma diagnostics and skin lesion triage).ResultsThe trained algorithms performed significantly less accurate diagnostics on images of nevi, melanomas and actinic keratoses from the AISC-2021 data set than the ISIC-2019 data set (p < 0.003). Almost one-third (31.1%) of the melanomas were misclassified during the melanoma diagnostics simulation, irrespective of their Breslow thickness. Furthermore, the algorithms marked 92.7% of the lesions ‘suspicious’ during the triage simulation, which yielded a triage sensitivity and specificity of 99.7% and 8.2%, respectively.ConclusionsAlthough state-of-the-art artificial intelligence outperforms dermatologists on image-based skin lesion classification within an artificial setting, additional data and technological advances are needed before clinical implementation
AB - BackgroundArtificial intelligence can be trained to outperform dermatologists in image-based skin cancer diagnostics. However, the networks' sensitivity to biases and overfitting may hamper their clinical applicability.ObjectivesThe aim of this study was to explain the potential consequences of implementing convolutional neural networks for stand-alone melanoma diagnostics and skin lesion triage.MethodsIn this algorithm validation study on retrospective data, we reproduced and evaluated the performance of state-of-the-art artificial intelligence (convolutional neural networks) for skin cancer diagnostics. The networks were trained on 25,331 annotated dermoscopic skin lesion images from an open-source data set (ISIC-2019) and tested using a novel data set (AISC-2021) consisting of 26,591 annotated dermoscopic skin lesion images. We tested the trained algorithms' ability to generalize to new data and their diagnostic performance in two simulations (melanoma diagnostics and skin lesion triage).ResultsThe trained algorithms performed significantly less accurate diagnostics on images of nevi, melanomas and actinic keratoses from the AISC-2021 data set than the ISIC-2019 data set (p < 0.003). Almost one-third (31.1%) of the melanomas were misclassified during the melanoma diagnostics simulation, irrespective of their Breslow thickness. Furthermore, the algorithms marked 92.7% of the lesions ‘suspicious’ during the triage simulation, which yielded a triage sensitivity and specificity of 99.7% and 8.2%, respectively.ConclusionsAlthough state-of-the-art artificial intelligence outperforms dermatologists on image-based skin lesion classification within an artificial setting, additional data and technological advances are needed before clinical implementation
U2 - 10.1002/jvc2.59
DO - 10.1002/jvc2.59
M3 - Journal article
VL - 1
SP - 344
EP - 354
JO - JEADV Clinical Practice
JF - JEADV Clinical Practice
SN - 2768-6566
IS - 4
ER -
ID: 346603187