Automated Image Classification of Skin Lesions
Convolutional neural network classification of dermatological photographs captured on a mobile device, evaluated for triage of skin neglected tropical diseases before confirmatory testing.
- Technology
- Image Analysis · Convolutional network classification
- What it measures
- Dermatological image features
- Developer
- Academic and digital health groups, including Fiocruz (Brazil), Tulane University (USA) and the World Health Organization
- Specimen
- None (clinical/imaging)
- Development status
- Assay in development
- Regulatory status
- None / RUO
- Intended use
- Triage of skin neglected tropical diseases before confirmatory testing
- Main limitation
- Cannot identify the organism, and training images under-represent darker skin.
- Availability
- The WHO Skin NTDs app, which uses two AI algorithms to classify skin images, is free on Android and iOS (WHO, December 2024). Its performance was still being evaluated.
- Linked publications
- 11
- Collaborating institutions
- Microsoft (Redmond, WA, USA), Novartis Foundation (Basel, Switzerland), UniversalDoctor (Barcelona, Spain)
- TPP-1, confirming leprosy
- 62.0 of 100, partial alignment
Open this test in the landscapeAll 37 tests · Home