How accurate are smartphone apps that use artificial intelligence to detect skin cancer in Australian adults?

Study of the efficacy of use of 4 commercially available smartphone applications in the self diagnosis of melanoma from images of naevi and suspicious lesions. While the first three apps are directed for consumer use and uses just the phone camera, the fourth app is directed to clinician users, and requires a dermoscope attachment to the mobile camera. Photos of naevi/suspicious lesions will be taken by smartphones. A dermatology consultation will determine if lesions are suspicious or benign. For patients with suspicious lesions, photos will be taken prior to the traditional dermatological intervention involving excision of the lesion. The smartphone photo's will be submitted to each of the smartphone applications for evaluation. The smartphone application evaluations will then be compared to the histo-pathological report and dermatologist diagnosis for suspect lesions, and to the dermatologist diagnosis for benign lesions (as no histo-pathological diagnosis will be available for benign lesions). Patients will not be required to use their phones, or download the apps for this study. We will use research owned iphones, and will pay for annual subscriptions to all the tested apps, so we have access to all app features. Research staff will take the photos using the research iphones within each app tested. Later photos will be stored on the University's secured research drive, and deleted from the phone. Most apps require a photo to be taken while using the app, and therefore it will not be the exact same photo used by each app for assessment. Up to five attempts will be made to take a photo of a selected lesion for each app. If the app does not successfully assess the image within those five attempts, it will be recorded as 'unevaluable for this app'. The three applications used in this study are; 1. SkinVision, developed by SkinVision B.V., requiring iOS 12.0 or later. This app offers risk profiling and storing skin images to a body map. For individual lesion photos the app first uses an AI algorithm to assess risk, then the lesion is also reviewed by a dermatologist through telehealth platform within 48 hours. This app is a regulated medical device with European CE marking. More information on this app can be found here: https://apps.apple.com/au/app/skinvision/id545293136 This app is directed for consumer use. 2. Nevus AI: developed by FatCube Limited, requiring iOS 15.1 or later. This app allows uploads of lesion images to provide an AI risk assessment for melanoma. The app also facilitates a ‘second opinion’ tool to provide tele-dermatologist review by 2 independent dermatologists. This app is registered as a Class I medical device with MHRA (UK). More information of this app can be found here: https://apps.apple.com/us/app/nevus-ai/id6502666699 This app is directed for consumer use. 3. AI Dermatologist (Skin Scanner): developed by Ashwood BK ChP, requiring iOS 15.6 or later. This app provides a broad recognition of 58 different skin conditions, including skin cancers, using AI technology. This app has no regulatory approval to be used as a diagnostic tool. More information on this app can be found here: https://apps.apple.com/us/app/ai-dermatologist-skin-scanner/id1511472597 This app is directed for consumer use. 4. VEOS DermaGraphix: developed by Canfield Scientific Inc. This application requires iOS 17.0 or later, and a dermatoscope attachment to be used with a smartphone camera. Unlike the above three apps that are consumer-directed, This app is intended for use by clinicians. More information of this app can be found here: https://apps.apple.com/us/app/veos-dermagraphix/id6446220269 This app is intended for use by clinicians.