We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
Using deep learning for dermatologist-level detection of suspicious pigmented skin lesions from wide-field images.
- Authors
Soenksen, Luis R.; Kassis, Timothy; Conover, Susan T.; Marti-Fuster, Berta; Birkenfeld, Judith S.; Tucker-Schwartz, Jason; Naseem, Asif; Stavert, Robert R.; Kim, Caroline C.; Senna, Maryanne M.; Avilés-Izquierdo, José; Collins, James J.; Barzilay, Regina; Gray, Martha L.
- Abstract
Finding the odd one out: Early identification of skin cancer is key to improving patient outcome. Soenksen et al. built a deep convolutional neural network that examines lesions from a given patient present in wide-field images, including those taken with cell phone cameras. Rather than evaluate a single lesion at a time looking for predetermined signs of neoplasia, the algorithm identifies lesions that differ from most of the other marks on that patient's skin, flagging them for further examination and ranking them in order of suspiciousness. The algorithm performed similarly to board-certified dermatologists and could potentially be used at primary care visits to help clinicians triage suspicious lesions for follow-up. A reported 96,480 people were diagnosed with melanoma in the United States in 2019, leading to 7230 reported deaths. Early-stage identification of suspicious pigmented lesions (SPLs) in primary care settings can lead to improved melanoma prognosis and a possible 20-fold reduction in treatment cost. Despite this clinical and economic value, efficient tools for SPL detection are mostly absent. To bridge this gap, we developed an SPL analysis system for wide-field images using deep convolutional neural networks (DCNNs) and applied it to a 38,283 dermatological dataset collected from 133 patients and publicly available images. These images were obtained from a variety of consumer-grade cameras (15,244 nondermoscopy) and classified by three board-certified dermatologists. Our system achieved more than 90.3% sensitivity (95% confidence interval, 90 to 90.6) and 89.9% specificity (89.6 to 90.2%) in distinguishing SPLs from nonsuspicious lesions, skin, and complex backgrounds, avoiding the need for cumbersome individual lesion imaging. We also present a new method to extract intrapatient lesion saliency (ugly duckling criteria) on the basis of DCNN features from detected lesions. This saliency ranking was validated against three board-certified dermatologists using a set of 135 individual wide-field images from 68 dermatological patients not included in the DCNN training set, exhibiting 82.96% (67.88 to 88.26%) agreement with at least one of the top three lesions in the dermatological consensus ranking. This method could allow for rapid and accurate assessments of pigmented lesion suspiciousness within a primary care visit and could enable improved patient triaging, utilization of resources, and earlier treatment of melanoma.
- Subjects
UNITED States; SIGNAL convolution; DEEP learning; CONVOLUTIONAL neural networks; MEDICAL personnel; CAMERA phones; CELL phones; SKIN cancer
- Publication
Science Translational Medicine, 2021, Vol 13, Issue 581, p1
- ISSN
1946-6234
- Publication type
Article
- DOI
10.1126/scitranslmed.abb3652