A new artificial intelligence (AI)-powered tool is set to revolutionize the diagnosis of skin diseases, including melanoma, by providing faster and more accurate results. Developed by a team led by Monash University and involving researchers from the University of Queensland, the tool, called **PanDerm**, analyzes multiple imaging types simultaneously, including close-up photos, dermoscopic images, pathology slides, and total body photographs.
In a recent study, PanDerm enhanced the accuracy of skin cancer diagnosis by 11% when used by dermatologists. It also improved diagnostic accuracy for other skin conditions by 16.5% among non-dermatologist healthcare professionals. PanDerm has the ability to detect skin cancer early, identifying potential problems before they are visible to clinicians.
Professor H. Peter Soyer from the University of Queensland, a key figure in the development of the tool, highlighted that PanDerm could be especially valuable in settings with limited resources or where access to dermatologists is scarce. The tool could bridge gaps between urban, regional, and rural healthcare settings by supporting existing clinical workflows.
The strength of PanDerm lies in its ability to operate even with limited annotated data, a common challenge in medical environments that often lack standard labeled data. Trained on more than two million skin images, the tool was developed using data sourced from 11 medical institutions worldwide.
First author and PhD student Siyuan Yan from Monash University emphasized the importance of PanDerm’s multimodal approach. By synthesizing data from different imaging techniques, the system can interpret skin conditions as dermatologists do, considering various visual sources to make accurate diagnoses.
PanDerm’s capabilities go beyond skin cancer diagnosis. It was evaluated across a broad range of clinical tasks, such as predicting the likelihood of cancer recurrence or spread, assessing skin type, counting moles, tracking lesion changes, and diagnosing various skin conditions. This makes it a more versatile tool compared to existing models, which are typically trained for single tasks.
Professor Victoria Mar, Director of the Victorian Melanoma Service at Alfred Health, said PanDerm is particularly useful for detecting subtle changes in lesions over time. This ability could help provide clues about lesion biology and future metastatic potential, supporting earlier diagnosis and more consistent monitoring, especially for patients at high risk of melanoma.
While PanDerm has shown promising results in research, it is still in the evaluation phase before it can be implemented widely in healthcare settings. The team plans to further test the tool, establishing standardized protocols for cross-demographic assessments and ensuring that it performs equitably across different patient populations and healthcare environments.
The research, which involves experts from Monash University and various international institutions, aims to improve the early detection and treatment of skin diseases, ultimately leading to better patient outcomes.
Related Topics:
