Unlocking Potential in Clinical Practice: A Comprehensive Assessment of an Artificial Intelligence Based Clinical Decision Support System for Cutaneous Melanoma Detection

Background: Artificial intelligence (AI) based clinical decision support systems (CDSS) for cutaneous melanoma detection have demonstrated high diagnostic accuracy in retrospective studies. However, the transition to prospective clinical trials in authentic primary care settings remains limited.

Objective: To assess the diagnostic performance of an AI-based CDSS, operated through a dedicated smartphone application (app) for dermoscopic imaging, for cutaneous melanoma detection in real-life primary care settings.

Methods: Two prospective, multicentre trials were conducted in Sweden. The first trial involved fifteen primary care physicians (PCPs) in near-live simulations with the app [1]. The second trial, spanning 36 primary care centres, employed the app in a real-life clinical setting and provided a dichotomous decision support text to participating physicians [2]. Lesions underwent standard diagnostic procedures, which were compared to the app outcome.

Results: From the first trial, near-live simulations highlighted trust, usability and the clinical context when PCPs assessed the app's potential in practice. The second trial assessed a total of 253 lesions, identifying 21 melanomas. The AI-based CDSS app exhibited an area under the receiver operating characteristic (AUROC) curve of 0.960 (95% CI: 0.928–0.980) for all melanomas, indicating high diagnostic accuracy. For invasive melanomas (n=11), the AUROC was 0.988 (95% CI: 0.965–0.997), with a sensitivity and specificity of 100% and 92.6%, respectively.

Conclusions: The AI-based CDSS app shows promise in enhancing cutaneous melanoma detection in primary care. A following prospective, open, multi-centre clinical trial, with a cluster randomised crossover design, will evaluate the potential of the AI-based tool in real clinical decision making, underscoring its transformative potential in primary care for skin cancer detection.

References

[1] Jonatan Helenason, Christoffer Ekström, Magnus Falk & Panagiotis Papachristou, (2024), Exploring the feasibility of an artificial intelligence based clinical decision support system for cutaneous melanoma detection in primary care – a mixed method study, Taylor & Francis Online, Scandinavian Journal of Primary Health Care, DOI: 10.1080/02813432.2023.2283190

[2] Panagiotis Papachristou, My Söderholm, Jon Pallon, Marina Taloyan, Sam Polesie, John Paoli, Chris D Anderson, Magnus Falk, (2024), Evaluation of an artificial intelligence-based decision support for detection of cutaneous melanoma in primary care – a prospective, real-life, clinical trial, Oxford Academic, British Journal of Dermatology, https://doi.org/10.1093/bjd/ljae021