• Politics
  • Diversity, equity and inclusion
  • Financial Decision Making
  • Telehealth
  • Patient Experience
  • Leadership
  • Point of Care Tools
  • Product Solutions
  • Management
  • Technology
  • Healthcare Transformation
  • Data + Technology
  • Safer Hospitals
  • Business
  • Providers in Practice
  • Mergers and Acquisitions
  • AI & Data Analytics
  • Cybersecurity
  • Interoperability & EHRs
  • Medical Devices
  • Pop Health Tech
  • Precision Medicine
  • Virtual Care
  • Health equity

AI Performs Similarly to Humans in Identifying Cancerous Lesions


AI can play a role in accurately identifying skin cancer, researchers found.


Photo/Thumb have been modified. Courtesy of Jakub Jirsák - stock.adobe.com.

As more researchers turn to artificial intelligence (AI) to explore the accuracy of the technology compared to human experts, a recent study found that an algorithm performed similarly to specialists when identifying melanoma from dermoscopic images of selected lesions.

The AI algorithm achieved an area under the receiver operator characteristic curve (AUROC) of 91.8% and a 100% sensitivity, the study, published in JAMA Network Open, revealed. When compared to clinicians, the AI had a specificity of 64.8%, while human experts achieved a specificity of 69.9%.

“As the burden of skin cancer increases, artificial intelligence technology could play a role in identifying lesions with a high likelihood of melanoma,” the researchers reported.

AI-based services could transform patient diagnosis pathways and enable greater efficiencies throughout the healthcare service, the researchers added.

The researchers set out to determine the accuracy of an AI algorithm in identifying melanoma in dermoscopic images taken with a smartphone — iPhone 6s or Galaxy S6 — and DSLR cameras.

Patients were included in the study if they went to a dermatology or plastic surgery clinic with at least one skin lesion referred for histological evaluation. The study included 514 patients in total.

The research team collected the histopathology results on biopsied excised lesions and categorized them as melanoma, dysplastic nevi or other.

Before the study, the AI algorithm was trained with published dermoscopic images. Nearly 300 images from this study were extracted from the data and used to further train the AI. The subset included 36 confirmed melanoma lesions, 67 randomly selected nonmelanoma lesions and 186 control lesions.

Photos from each camera helped train a version of the algorithm, which was used to assess the leftover images from that camera type.

The AI generated a number from zero to one to reflect its confidence that the lesion was melanoma. Zero indicated that the lesion was certainly benign, and one indicated that the lesion was certainly malignant melanoma.

Overall, 1,550 images of skin lesions were included in the analysis, of which 551 were biopsied lesions and 999 were control lesions.

Training data included 858 images of 286 lesions from 92 patients.

The algorithm achieved an AUROC of 90.1% for biopsied lesions and 95.8% for all lesions using the iPhone 6s. For the Galaxy S6, it achieved an AUROC of 85.8% for biopsied lesions and 93.8% for all lesions. The DSLR camera had an AUROC of 86.9% for biopsied lesions and 91.8% for all lesions.

Specialists achieved an AUROC of 77.8% and a specificity of 69.9%.

“The results of the study showed that the algorithm and specialists identified melanoma in selected suspicious pigmented skin lesions at a similar level of accuracy,” the researchers wrote.

Researchers are now investigating the ability of the AI to identify nonmelanoma skin cancer and benign conditions.

Get the best insights inside digital health directly to your inbox.


AI Could Boost Identification of Chron's Disease

Machine Learning Analyzes MRI Scans 186-Times Faster than Humans

AI Accurately Detects Key Findings in Chest X-Rays in 10 Seconds

Related Videos
Image: Ron Southwick, Chief Healthcare Executive
George Van Antwerp, MBA
Edmondo Robinson, MD
Craig Newman
Related Content
© 2024 MJH Life Sciences

All rights reserved.