Publication date: Jul 12, 2019
Computer vision has promise in image-based cutaneous melanoma diagnosis but clinical utility is uncertain.
To determine if computer algorithms from an international melanoma detection challenge can improve dermatologist melanoma diagnostic accuracy.
Cross-sectional study using 150 dermoscopy images (50 melanomas, 50 nevi, 50 seborrheic keratoses) from the test dataset of a melanoma detection challenge, along with algorithm results from twenty-three teams. Eight dermatologists and nine dermatology residents classified dermoscopic lesion images in an online reader study and provided their confidence level.
The top-ranked computer algorithm had a ROC area of 0.87, which was higher than the dermatologists (0.74) and the residents (0.66) (p
Marchetti, M.A., Liopyris, K., Dusza, S.W., Codella, N.C.F., Gutman, D.A., Helba, B., Kalloo, A., Halpern, A.C., and (ISIC), International Skin. Imaging. Collaboration. Computer Algorithms Show Potential for Improving Dermatologists’ Accuracy to Diagnose Cutaneous Melanoma; Results of ISIC 2017. 23371. 2019 J Am Acad Dermatol.
- Enhanced classifier training to improve precision of a convolutional neural network to identify images of skin lesions.
- Comparison of the accuracy of human readers versus machine-learning algorithms for pigmented skin lesion classification: an open, web-based, international, diagnostic study.
- Physicians trained in dermatoscopy can improve odds for early detection of melanoma
- Melanoma Detection by Means of Multiple Instance Learning.