Deep‐learning‐based, computer‐aided classifier developed with a small dataset of clinical images surpasses board‐certified dermatologists in skin tumour diagnosis. (19th September 2018)

• Record Type: Journal Article
• Authors: Fujisawa, Y.; Otomo, Y.; Ogata, Y.; Nakamura, Y.; Fujita, R.; Ishitsuka, Y.; Watanabe, R.; Okiyama, N.; Ohara, K.; Fujimoto, M.
• Is Part Of: British journal of dermatology. Volume 180:Number 2(2019)
• Page Start: 373
• Page End: 381
• Publication Date: 2018-09-19
• Subjects: Dermatology -- Periodicals; Skin -- Diseases -- Periodicals; 616.5
• DOI: 10.1111/bjd.16924
• Languages: English
• ISSNs: 0007-0963

Abstract

Summary: Background: Application of deep‐learning technology to skin cancer classification can potentially improve the sensitivity and specificity of skin cancer screening, but the number of training images required for such a system is thought to be extremely large. Objectives: To determine whether deep‐learning technology could be used to develop an efficient skin cancer classification system with a relatively small dataset of clinical images. Methods: A deep convolutional neural network (DCNN) was trained using a dataset of 4867 clinical images obtained from 1842 patients diagnosed with skin tumours at the University of Tsukuba Hospital from 2003 to 2016. The images consisted of 14 diagnoses, including both malignant and benign conditions. Its performance was tested against 13 board‐certified dermatologists and nine dermatology trainees. Results: The overall classification accuracy of the trained DCNN was 76·5%. The DCNN achieved 96·3% sensitivity (correctly classified malignant as malignant) and 89·5% specificity (correctly classified benign as benign). Although the accuracy of malignant or benign classification by the board‐certified dermatologists was statistically higher than that of the dermatology trainees (85·3% ± 3·7% and 74·4% ± 6·8%, P < 0·01), the DCNN achieved even greater accuracy, as high as 92·4% ± 2·1% ( P < 0·001). Conclusions: We have developed an efficient skin tumour classifier using a DCNN trained on a relatively small dataset. The DCNN classified images of skin tumours more accurately than board‐certified dermatologists. Collectively, the current system may have capabilities for screening purposes in general medical practice, particularly because it requires only a single clinical image for classification. Abstract : What's already known about this topic? Several computer‐aided classification systems have been introduced that achieve high sensitivity for melanoma detection; however, low specificity was a trade‐off for high sensitivity. The application of deep‐learning technology to skin cancer classification could potentially improve the sensitivity and specificity of skin cancer screening. The number of training images required for such a system is thought to be extremely large, and compiling large datasets for rare skin conditions is difficult. What does this study add? Our deep convolutional neural network (DCNN), trained on only 4867 images from 1842 patients, classified images of skin tumours into 14 different diagnoses more accurately than board‐certified dermatologists. The fluctuation range was only ± 3·2% by fivefold cross‐validation, showing the robustness of the system. Our DCNN system requires only a single image and provides 96·3% sensitivity and 89·5% specificity in the detection of skin cancer; however, it should be validated in a prospective clinical study before use for screening purposes in general medical practice. Linked Editorial: Janda and Soyer. Br J Dermatol 2019;180 :247–248 . Plain language summary available online Respond to this article