An educational machine learning demonstration framework for plastic surgeons using open datasets
Kian Daneshi1, Yasona Neocleous2, Abigail G-Medhin3
1School of Population Health and Medicine, University of Sheffield, Sheffield, UK; Department of Bioengineering, Imperial College London, London, UK.
Journal of Plastic, Reconstructive & Aesthetic Surgery : JPRAS
|February 28, 2026
Summary
Plastic surgeons can now build and deploy AI tools for melanoma detection using open datasets and machine learning. This framework, DermAI-Melanoma, empowers clinicians with practical data science skills for improved diagnostic capabilities.
Area of Science:
- Dermatology and Plastic Surgery
- Artificial Intelligence
- Machine Learning
- Data Science
Background:
- Limited practical engagement with data science in dermatology and plastic surgery.
- Open datasets enable clinicians to explore machine learning without specialized infrastructure.
- DermAI-Melanoma is an open data demonstration framework for plastic surgeons using melanoma classification.
Purpose of the Study:
- Demonstrate reproducible training and deployment of deep learning models for melanoma classification.
- Provide a didactic case study for plastic surgeons to engage with data science.
- Showcase the use of public datasets for AI development in surgery.
Main Methods:
- Utilized the SIIM-ISIC 2020 melanoma dataset with patient-level stratification to prevent data leakage.
- Applied image preprocessing including resizing, color balancing, and augmentation.
- Trained two convolutional neural networks: EfficientNet-B3 and MobileNetV3-Small, deploying them via TensorFlow.js.
Main Results:
- EfficientNet-B3 achieved 97% test accuracy, detecting 92% of melanomas.
- MobileNetV3-Small achieved 94% accuracy, with <5 MB storage and <2s smartphone deployment.
- Model performance is comparable to dermatologist benchmarks in melanoma detection literature.
Conclusions:
- DermAI-Melanoma empowers plastic surgeons to build transparent, deployable AI tools using open data.
- Similar frameworks can foster education, research, and innovation in plastic surgery.
- Embracing open data sharing and cross-disciplinary collaboration is crucial for advancing AI in surgery.