Early cancer detection through smartphone imaging and federated learning-enhanced convolutional neural networks
Chiang Liang Kok, Anthony Ongkojaya, Yit Yan Koh · 2025
The integration of smartphone-based imaging with artificial intelligence offers a ubiquitous and universally accessible platform for health assessment. This study leverages the use of Convolutional Neural Network (CNN) to analyze captured images of potentially malignant and benign skin lesions. CNNs are trained to extract visual key characteristics that are associated with cancerous growth, aiming to provide a risk assessment that is able to guide the user to seek timely professional diagnosis and clinical validation. By integrating a Federated Learning (FL) process in training the CNN, it provides a privacy-preserving approach towards this process, ensuring the raw data remains securely on the user’s device.