Melanoma is one of the deadliest forms of skin cancer, but it is highly treatable when detected early. Traditional screening relies heavily on visual inspection by clinicians, which can be time-consuming and challenging, especially given the large number of benign lesions compared to malignant ones. In recent years, 3D Total Body Photography (TBP) has emerged as a transformative technology for skin surveillance, enabling comprehensive imaging of the entire skin surface and allowing clinicians to track changes in lesions over time. However, the volume of data generated by TBP is immense, with each dataset containing thousands of lesion and non-lesion images, making manual review impractical.

This project develops an AI screening system using 3D TBP to automatically identify and prioritize lesions that are most suspicious for melanoma. The integration of AI is essential to manage the scale and complexity of TBP data, helping to highlight high-risk lesions that warrant closer examination. The system integrates two key components:

1. AI Screening Algorithm – identifies high-risk lesions from TBP image tiles using lesion detection, enhancement, and 'ugly duckling' contextual analysis.
2. Diagnostic Algorithm – analyses flagged lesions with dermoscopic images, highlighting the dermoscopic patterns and providing "reasoning process" that aligns with clinical assessment criteria.

By combining TBP and dermoscopy, the tool prioritises high-risk lesions for clinician review, improving efficiency and reducing missed melanomas. Validation against clinical experts ensures the system enhances, rather than replaces, medical judgment.

Research Objectives

• Develop a robust and validated AI system for automated lesion screening using 3D TBP data.
• Evaluate the system's performance against expert dermatologists in a prospective clinical trial to ensure it enhances, rather than replaces, clinical judgment.
• Translate the validated system into an accessible and equitable clinical tool to improve melanoma screening across Australia.

The Outcome

The project will deliver a comprehensive AI-powered skin cancer screening tool for use in clinical settings. Outcomes include:

• A validated AI system for identifying high-risk lesions from TBP and dermoscopy.
• A new benchmark for AI-assisted melanoma screening, with potential to improve early diagnosis, patient survival, and healthcare efficiency.