Please login or create an account. If you do not have access to this content, you will be shown a 30 second preview and licensing options.
- Presentation
Australia’s National Skin Cancer Screening Program and AI Integration in Dermatology
Description
Victoria Maher, a dermatologist from Melbourne, described Australia’s challenge of rapidly increasing demand for skin checks amid major workforce shortages, confusing consumer messaging, and poor access in regional areas. In response, the federal government funded a roadmap toward a national targeted melanoma screening program, and her team used a major infrastructure grant to deploy 3D imaging systems across multiple states, including rural and remote sites. Their cohort study, started in 2020, enrolls participants through self-referral or clinician referral, uses validated risk assessment, nurse-acquired dermoscopic imaging, biospecimen collection, and dermatologist review, with follow-up intervals based on risk. The study has enrolled over 9,000 participants and captured more than 700 melanomas, with interim results showing high reviewer concordance and a focus on detecting early, thin disease. She emphasized using AI to automate risk stratification, help staff choose lesions for dermoscopy, reduce unnecessary biopsies, and identify patients needing face-to-face assessment. Her team is also developing foundation and vision-language models, improving explainability and zero-shot performance, and building image-text datasets to address the lack of annotation in dermatology. The next phase is a step-wedge randomized trial across 15 sites, integrating AI into screening workflows and adding tailored prevention interventions in partnership with SunSmart.
View moreConclusions
- Australia’s skin-cancer burden and workforce shortages make universal screening impractical, so a targeted, risk-based approach is needed.
- A national melanoma screening program is being planned to replace confusing, ad hoc consumer-driven skin-check pathways with an organized system.
- The ACEMID cohort has shown that large-scale 3D total-body imaging screening is feasible across metropolitan and regional sites, including very remote areas.
- Participants can be successfully stratified into risk groups, with higher-risk people reviewed more often and low-risk people seen less frequently.
- The screening program is detecting many melanomas, mostly in situ and thin invasive lesions, suggesting it is finding disease earlier.
- Pathology review within the study shows high agreement between expert reviewers, but lower concordance with initial community diagnoses, indicating room for quality improvement.
- AI is viewed primarily as a triage tool to automate risk assessment, prioritize lesions for dermoscopy, and reduce unnecessary face-to-face visits and biopsies.
- Foundation-model and vision-language approaches are being developed to improve automated dermatology diagnosis, zero-shot performance, and explainability.
- Building better dermatology AI requires larger, better-annotated multimodal datasets, since dermatology lacks the image-text pairing common in pathology and radiology.
- The next phase is a stepped-wedge randomized rollout to test whether AI-assisted screening can safely improve efficiency and access while maintaining clinical effectiveness.
- Preventive interventions tailored to individual risk can be integrated into screening and appear acceptable and effective.
- Overall, the work suggests that combining coordinated screening, expert pathology, multimodal AI, and tailored prevention could improve melanoma detection and reduce unnecessary clinical workload in Australia.
- Reyes-Marcelino, Cancer Epi 2021
- Glasziou et al. MJA 2020
- Yan, S et. Al. Nature Medicine, 2025.
- Yan et al. Derm1M, 2025
- Yan, S et. al. Under Review.
- Table 1. The Melanocytic Pathology Assessment Tool and Hierarchy for Diagnosis Version 2.0