A reinforcement learning model for AI-based decision support in skin cancer.
Journal
Nature medicine
ISSN: 1546-170X
Titre abrégé: Nat Med
Pays: United States
ID NLM: 9502015
Informations de publication
Date de publication:
08 2023
08 2023
Historique:
received:
31
08
2022
accepted:
28
06
2023
medline:
17
8
2023
pubmed:
28
7
2023
entrez:
27
7
2023
Statut:
ppublish
Résumé
We investigated whether human preferences hold the potential to improve diagnostic artificial intelligence (AI)-based decision support using skin cancer diagnosis as a use case. We utilized nonuniform rewards and penalties based on expert-generated tables, balancing the benefits and harms of various diagnostic errors, which were applied using reinforcement learning. Compared with supervised learning, the reinforcement learning model improved the sensitivity for melanoma from 61.4% to 79.5% (95% confidence interval (CI): 73.5-85.6%) and for basal cell carcinoma from 79.4% to 87.1% (95% CI: 80.3-93.9%). AI overconfidence was also reduced while simultaneously maintaining accuracy. Reinforcement learning increased the rate of correct diagnoses made by dermatologists by 12.0% (95% CI: 8.8-15.1%) and improved the rate of optimal management decisions from 57.4% to 65.3% (95% CI: 61.7-68.9%). We further demonstrated that the reward-adjusted reinforcement learning model and a threshold-based model outperformed naïve supervised learning in various clinical scenarios. Our findings suggest the potential for incorporating human preferences into image-based diagnostic algorithms.
Identifiants
pubmed: 37501017
doi: 10.1038/s41591-023-02475-5
pii: 10.1038/s41591-023-02475-5
pmc: PMC10427421
doi:
Types de publication
Journal Article
Research Support, Non-U.S. Gov't
Research Support, N.I.H., Extramural
Langues
eng
Sous-ensembles de citation
IM
Pagination
1941-1946Subventions
Organisme : NCI NIH HHS
ID : P30 CA008748
Pays : United States
Organisme : NCI NIH HHS
ID : U24 CA264369
Pays : United States
Informations de copyright
© 2023. The Author(s).
Références
Esteva, A. et al. Dermatologist-level classification of skin cancer with deep neural networks. Nature 542, 115–118 (2017).
doi: 10.1038/nature21056
pubmed: 28117445
pmcid: 8382232
Tschandl, P. et al. Human-computer collaboration for skin cancer recognition. Nat. Med. https://doi.org/10.1038/s41591-020-0942-0 (2020).
doi: 10.1038/s41591-020-0942-0
pubmed: 32572267
Tschandl, P. et al. Comparison of the accuracy of human readers versus machine-learning algorithms for pigmented skin lesion classification: an open, web-based, international, diagnostic study. Lancet Oncol. 20, 938–947 (2019).
doi: 10.1016/S1470-2045(19)30333-X
pubmed: 31201137
pmcid: 8237239
Haenssle, H. A. et al. Man against machine: diagnostic performance of a deep learning convolutional neural network for dermoscopic melanoma recognition in comparison to 58 dermatologists. Ann. Oncol. 29, 1836–1842 (2018).
doi: 10.1093/annonc/mdy166
pubmed: 29846502
McKinney, S. M. et al. International evaluation of an AI system for breast cancer screening. Nature 577, 89–94 (2020).
doi: 10.1038/s41586-019-1799-6
pubmed: 31894144
Haggenmüller, S. et al. Skin cancer classification via convolutional neural networks: systematic review of studies involving human experts. Eur. J. Cancer 156, 202–216 (2021).
doi: 10.1016/j.ejca.2021.06.049
pubmed: 34509059
Birch, J., Creel, K. A., Jha, A. K. & Plutynski, A. Clinical decisions using AI must consider patient values. Nat. Med. 28, 229–232 (2022).
doi: 10.1038/s41591-021-01624-y
pubmed: 35102337
pmcid: 9825143
Song, C. & Li, X. Cost-Sensitive KNN algorithm for cancer prediction based on entropy analysis. Entropy 24, 253 (2022).
doi: 10.3390/e24020253
pubmed: 35205547
pmcid: 8871087
Collell, G., Prelec, D. & Patil, K. R. A simple plug-in bagging ensemble based on threshold-moving for classifying binary and multiclass imbalanced data. Neurocomputing 275, 330–340 (2018).
doi: 10.1016/j.neucom.2017.08.035
pubmed: 29398782
pmcid: 5750819
Yala, A. et al. Optimizing risk-based breast cancer screening policies with reinforcement learning. Nat. Med. 28, 136–143 (2022).
doi: 10.1038/s41591-021-01599-w
pubmed: 35027757
Mnih, V. et al. Human-level control through deep reinforcement learning. Nature 518, 529–533 (2015).
doi: 10.1038/nature14236
pubmed: 25719670
Combalia, M. et al. Validation of artificial intelligence prediction models for skin cancer diagnosis using dermoscopy images: the 2019 International Skin Imaging Collaboration Grand Challenge. Lancet Digit Health 4, e330–e339 (2022).
doi: 10.1016/S2589-7500(22)00021-8
pubmed: 35461690
pmcid: 9295694
Miller, K. D. et al. Cancer treatment and survivorship statistics, 2022. CA Cancer J. Clin. https://doi.org/10.3322/caac.21731 (2022).
doi: 10.3322/caac.21731
pubmed: 36346402
Tschandl, P., Rosendahl, C. & Kittler, H. The HAM10000 dataset, a large collection of multi-source dermatoscopic images of common pigmented skin lesions. Sci. Data 5, 180161 (2018).
doi: 10.1038/sdata.2018.161
pubmed: 30106392
pmcid: 6091241
Fraenkel, L. & Fried, T. R. Individualized medical decision making: necessary, achievable, but not yet attainable. Arch Intern Med. 170, 566–569 (2010).
Rinner, C., Kittler, H., Rosendahl, C. & Tschandl, P. Analysis of collective human intelligence for diagnosis of pigmented skin lesions harnessed by gamification via a web-based training platform: simulation reader study. J. Med. Internet Res. 22, e15597 (2020).
doi: 10.2196/15597
pubmed: 32012058
pmcid: 7007585
R Core Team. R: A Language and Environment for Statistical Computing (R Foundation for Statistical Computing, 2022).
Tschandl, P. The HAM10000 dataset, a large collection of multi-source dermatoscopic images of common pigmented skin lesions. Harvard Dataverse, V4 https://doi.org/10.7910/DVN/DBW86T (2018).
Harald, K. A reinforcement learning model for AI based decision support in skin cancer. Harvard Dataverse https://doi.org/10.7910/DVN/PWQMQ7 (2023).