MUMBAI, India, July 30 -- Intellectual Property India has published a patent application (202631083809 A) filed by Nagaland University on July 08, 2026, for Privacy-Preserving Multimodal Continual Learning System For Incremental Cancer Detection.
Inventors include Sujata Dash; Bhupesh Deka; H. Rohen Singh; Shanchamo Yanthan; Teisovi Angami; Sourav Hazarika; and Sudipta Patowary.
The application for the patent was published on July 24, 2026, under issue no. 30/2026.
Abstract: The present disclosure relates to a privacy-preserving multimodal continual learning system for incremental cancer detection. The system comprises modality-specific feature extractors processing histopathological images (11), photographic images (12), and clinical data (13), a multimodal feature fusion module (14), and a shared frozen backbone (15) for stable feature representation. An adaptive replay memory module (18) selects representative samples using uncertainty estimation (22) and diversity selection (23) for efficient knowledge retention. A task-aware knowledge distillation mechanism including teacher model (31) and student model (35) enables adaptive knowledge transfer using soft targets (33) and distillation loss (34). The system further supports cross-domain learning across cancer types using modules (41– 45). The invention enables incremental learning without access to complete historical datasets while preserving performance.
Disclaimer: Curated by HT Syndication.
