Developing a national artificial intelligence-assisted skin cancer pathway

Colin Morton, Gillian Chin, Jacob Carse, Stephen McKenna, Rubeta Matin, Colin Fleming

Research output: Contribution to journalMeeting abstractpeer-review

4 Downloads (Pure)

Abstract

The AI Skin Cancer Consortium seeks to progress integration of artificial intelligence (AI) into the skin cancer referral pathway to promote rapid diagnosis of skin cancer and reassurance with benign lesions. In contrast to other fields of medicine, such as radiology, dermatology has lacked standardization in data acquisition, which is required for high-quality, reproducible and interoperable AI algorithm performance. Beginning in 2021, the consortium has made substantial progress towards defining a national pathway, technology architecture and data flows. We describe a standardized approach to acquisition of macroscopic and dermoscopic images, and technical and clinical metadata in a DICOM (Digital Imaging and Communications in Medicine) format, currently acquired in outpatient imaging centres and in general practice. In order to develop a transparent competitive environment for testing of this standardized National Health Service (NHS) pipeline, the consortium created a Small Business Research Initiative funded via a government agency. Three companies were awarded funding for this two-phase programme, from 15 applicants. These industry partners tested the pipeline within a trusted research environment (TRE), using real-world data. All industry partners were able to evaluate skin lesion images in combination with interoperable clinical metadata and run existing algorithms in the TRE. We report progress on this competition and also on the process of creating a national database, to permit competitive testing performance of different algorithms with specific clinical use cases. To assess the cost-effectiveness of introducing AI into the skin cancer pathway, we have also collaborated with the York Health Economics Consortium to produce a model to permit estimation of the impact of an AI triage tool to assist clinician triage. The recent vision statement from the British Association of Dermatologists encourages AI solutions intended to address a clinical unmet need and that integrates into patient pathways to enhance the service provided by healthcare professionals [https://www.bad.org.uk/clinical-services/artificial-intelligence/vision-statement-on-artificial-intelligence-ai-interventions/ (last accessed 19 March 2024)]. If the UK is to be at the forefront of skin AI, then it is imperative that the market is stimulated through use of meticulously standardized interoperable data standards with platforms that allow for transparent testing of multiple algorithms. We have successfully introduced a pipeline for safely generating standardized, high-quality images that are suitable for AI, identifying where we can achieve the greatest potential value for introducing AI into the NHS skin cancer referral pathway.
Original languageEnglish
Pages (from-to)i189
Number of pages1
JournalBritish Journal of Dermatology
Volume191
Issue number1
DOIs
Publication statusPublished - 28 Jun 2024
EventBritish Association of Dermatologists Annual Meeting: Teledermatology & Digital Dermatology Symposium - Manchester Central, Manchester, United Kingdom
Duration: 2 Jul 20244 Jul 2024
Conference number: 104

Fingerprint

Dive into the research topics of 'Developing a national artificial intelligence-assisted skin cancer pathway'. Together they form a unique fingerprint.

Cite this