Improving Skin Cancer Management With Artificial Intelligence (04.17 SMARTI) (SMARTI)

A Pilot Study of an Artificial Intelligence System as a Diagnostic Aid to Improve Skin Cancer Management (04.17 SMARTI)

The study is designed to be able to prove if the Molemap Artificial Intelligence (AI) algorithm can be used as a diagnostic aid in a clinical setting. This study will determine whether the diagnostic accuracy of the Molemap AI algorithm is comparable to a specialist dermatologist, teledermatologist and registrar (as a surrogate for a general practitioner). The study patient population will be adult patients who require skin cancer assessment. The use of AI as a diagnostic aid may assist primary care physicians who have variable skill in skin cancer diagnosis and lead to more appropriate referrals (rapid referral for lesions requiring treatment and fewer referrals for benign lesions), thereby improving access and reducing waiting times for specialist care.

This is a pilot study which aims to establish whether artificial intelligence can be used as a diagnostic aid to improve diagnostic accuracy and outcomes in the specialist setting prior to conducting a much larger trial of the intervention in primary care. Objectives: To establish whether the diagnostic accuracy of an artificial intelligence system is on par with teledermatologists' clinical assessment. To establish the safety and feasibility of offering artificial intelligence as a diagnostic aid prior to conducting a large trial of the intervention in primary care. Hypotheses: The AI algorithm will have diagnostic accuracy comparable with a teledermatologists' assessment. The AI algorithm will have a diagnostic accuracy more conservative (i.e. more false positives) than dermatologists in the clinical setting. The AI algorithm will have greater diagnostic accuracy than the registrar. The AI algorithm will lead to a reduction in the number of biopsies performed by the registrar the likely impact of which will be reduced cost to patients and the healthcare system. Trial Design: The pilot study will take place in specialist dermatology and melanoma clinics in Victoria, Australia. Potential participants will be identified and screened at the general dermatology and melanoma clinics by the clinic doctors who deem the participant meet the inclusion and exclusion criteria. Intervention: Photography of lesions using a MoleMap camera device with automated artificial intelligence providing an assessment of the lesion in real time. This pilot study will be a before and after intervention trial design. For the initial 'lead-in' phase, no AI diagnosis will be provided back to the treating clinicians. This phase will be used for prospective data collection. For the intervention phase, an AI diagnosis will be provided to the dermatology registrar (who is used in this pilot study as a surrogate for the GP) and dermatologist after they have both assessed the patient clinically. Management of the lesion will be determined by the dermatologist and recorded. The safety of the device will be determined by its use in the setting of specialist dermatology clinics to ensure that patients are receiving the highest standard of care with a dermatologist providing a clinical diagnosis and management for all lesions tested. It is anticipated that the full trial will expand to include multiple sites across Australia and New Zealand.

During the lead-in phase treating clinicians will not be given the Molemap artificial intelligence diagnosis in real-time (i.e. in clinic with the patient).

During the active phase treating clinicians will be given the Molemap artificial intelligence diagnosis in real-time.

This device/software incorporates artificial intelligence to provide a diagnostic aide for clinicians of patients with potentially malignant skin lesions. The software is supported by the use of cameras for acquisition of images.

Diagnostic accuracy of the device compared to teledermatologist, dermatologist and registrar using histopathology as 'gold standard' for any lesions biopsied.

This will be assessed by comparing the lesions selected for review by the registrar with the lesions selected by the dermatologist.

Appropriateness of management by registrar compared to specialist dermatologists and impact AI might have on this.

This will be assessed by comparing the registrars clinical assessment with the dermatologists clinical assessment and if providing the AI assessment in real time has an impact.

Inclusion Criteria: Patients attending the specialist dermatology clinics for skin cancer assessment or surveillance. Patients may or may not have a lesion of concern. Patients must have at least two lesions imaged during full skin examination by a dermatologist. Age greater than 18 years. Participant is willing and able to undertake investigation of suspicious lesion (e.g. skin biopsy). Exclusion Criteria: Patient does not give informed consent. Patient is unable or unwilling to have a full skin examination Patient has a known past or current diagnosis of cognitive impairment

Felmingham C, MacNamara S, Cranwell W, Williams N, Wada M, Adler NR, Ge Z, Sharfe A, Bowling A, Haskett M, Wolfe R, Mar V. Improving Skin cancer Management with ARTificial Intelligence (SMARTI): protocol for a preintervention/postintervention trial of an artificial intelligence system used as a diagnostic aid for skin cancer management in a specialist dermatology setting. BMJ Open. 2022 Jan 4;12(1):e050203. doi: 10.1136/bmjopen-2021-050203.