Generative AI for enhanced skin cancer diagnosis, dermatologist training, and patient education
Abstract
The early detection and monitoring of suspicious skin lesions are essential for effective dermatological diagnosis and treatment, particularly in understanding the progression of nevi to melanoma. To advance this understanding, we developed a simulation framework that models the transformation of nevi into melanoma using Cycle-Consistent Generative Adversarial Networks and frame interpolation to generate a detailed dataset of simulated lesion progressions. Optical flow analysis was applied to these dermoscopic image sequences, providing quantitative insights into lesion transformations and dynamic changes. Heatmap visualizations and optical flow vectors highlighted regions of significant activity and confirmed the fidelity of the simulated transformations. Additionally, we demonstrate how this new approach can visually complement existing textual explainable AI methods in dermatology, enhancing interpretability and trust in diagnostic outcomes. These findings represent a significant step toward improving dermatological diagnostics, patient education, and the early detection of melanoma.
Details
- Organisation(s)
-
Hannover Centre for Optical Technologies (HOT)
PhoenixD: Photonics, Optics, and Engineering - Innovation Across Disciplines
- External Organisation(s)
-
Coronis Computing S.L.
University of Girona
- Type
- Conference contribution
- Publication date
- 19.03.2025
- Publication status
- Published
- Peer reviewed
- Yes
- ASJC Scopus subject areas
- Electronic, Optical and Magnetic Materials, Atomic and Molecular Physics, and Optics, Biomaterials, Radiology Nuclear Medicine and imaging
- Sustainable Development Goals
- SDG 3 - Good Health and Well-being
- Electronic version(s)
-
https://doi.org/10.1117/12.3042664 (Access:
Closed
)
