A Skin Image Reference Tool to Aid Healthcare Providers' Diagnosis
- Conditions
- Dermatologic Disease
- Registration Number
- NCT07033169
- Lead Sponsor
- Wake Forest University Health Sciences
- Brief Summary
Consented patients will have three images taken of their dermatologic conditions within the Belle.ai software. These images will be uploaded and saved within the Belle software system where a single AI-generated differential list will be generated based on the three photos. All photos uploaded will be de-identified. The software will not have any unique identifiers of participants saved in the system. The photos will be named based on participant enrollment numbers or unique code numbers and no unique identifiers will be attached to the photos. There will be no data collection form necessary for this study
- Detailed Description
Belle.ai provides a differential diagnosis from more than 2,000 different skin conditions leveraging a database trained on over 500,000 images. The image referencing technology deploys deep learning to analyze an uploaded clinical image and then matches its geometric pattern characteristics to Belle.ai's database of images to provide reference differentials. The purpose is to determine the validity of the Belle.ai software in diagnosing common dermatologic diseases across a range of skin tones.
Consented patients will have three images taken of their dermatologic disease within the Belle.ai software. These images will be uploaded and saved within the Belle system where a single AI-generated differential list will be generated based on the three photos. The study coordinator will review uploaded patient "cases" and assign the cases for review and adjudication to designated Dermatologic Review Committee (DRC) members within the Belle web portal. Successful validation will require \>80% concordance between Belle.ai's primary working diagnosis (#1 on the differential) and our dermatology experts. A team of dermatology experts will then secondarily assess the concordance among the remaining diagnoses.
Recruitment & Eligibility
- Status
- NOT_YET_RECRUITING
- Sex
- All
- Target Recruitment
- 400
- Patient must present to an Advocate Health dermatology clinic
- Patient must have the ability and willingness to provide informed consent and comply with study procedures and visits
- Participant dermatologists must have access to the required technology (e.g., smartphone with internet access) and be capable of using it for the required image capture
- Patients who are unable to comply with study procedures due to physical or mental health limitations (as assessed by study coordinator)
- Pediatric, adolescent, and teen patients who present with dermatological conditions on their genitalia will not be included in the study (in support of patient privacy concerns).
Study & Design
- Study Type
- OBSERVATIONAL
- Study Design
- Not specified
- Primary Outcome Measures
Name Time Method concordance of Belle.ai diagnoses with physician diagnoses. Day 1 The study coordinator will review uploaded patient "cases" and assign them for review and adjudication to designated Dermatologic Review Committee (DRC) members within the Belle web portal. The DRC will be comprised of 1-2 Advocate Health board-certified dermatologists from each of the Winston, Charlotte, and Midwest dermatology practices.
- Secondary Outcome Measures
Name Time Method
Trial Locations
- Locations (1)
Wake Forest University Health Sciences Department of Dermatology
🇺🇸Winston-Salem, North Carolina, United States