MUMBAI, India, June 30 -- Intellectual Property India has published a patent application (202641073769 A) filed by Vellore Institute Of Technology on June 14, 2026, for A Federated Artificial Intelligence System For Secure Multiclass Skin Cancer Detection.
Inventors include Suresh P; Keerthika P; A Adit Prasad; and Manjara Raj.
The application for the patent was published on June 26, 2026, under issue no. 26/2026.
Abstract: A FEDERATED ARTIFICIAL INTELLIGENCE SYSTEM FOR SECURE MULTICLASS SKIN CANCER DETECTION The present invention relates to a federated deep learning framework for privacy-preserving multiclass skin lesion diagnosis using dermoscopic images distributed across multiple healthcare institutions, as illustrated in FIG. 1. The system comprises a central coordination server, EfficientNetB0-based convolutional neural network backbone, personalized classification heads, and federated aggregation modules, as shown in FIG. 1 and FIG. 2. The invention enables collaborative training of diagnostic models without transferring raw patient images outside institutional boundaries. Participating healthcare institutions locally train the model using private dermoscopic image datasets, while only selected backbone parameter updates are shared with the central server for federated aggregation. The invention further incorporates transfer learning, personalized model adaptation, and communication-efficient aggregation mechanisms to improve diagnostic performance across heterogeneous clinical datasets. Experimental validation demonstrated improved classification accuracy, reduced communication overhead, and enhanced privacy preservation compared to centralized training methodologies.
Disclaimer: Curated by HT Syndication.
