MUMBAI, India, June 26 -- Intellectual Property India has published a patent application (202621052094 A) filed by Symbiosis International Deemed University on April 23, 2026, for Tri-Hybrid Attention-Fused Deep Learning System For Rapid Classification Of Eskape Pathogens From Gram-Stained Urine Microscopy Images.

Inventors include Miss. Rutuja Gumathannavar; Dr. Rupali Gangarde; Siddharth Sandeep; Pranavi Singh; Shreyas Tambe; and Vedant Ambadkar.

The application for the patent was published on June 19, 2026, under issue no. 25/2026.

Abstract: ABSTRACT TRI-HYBRID ATTENTION-FUSED DEEP LEARNING SYSTEM FOR RAPID CLASSIFICATION OF ESKAPE PATHOGENS FROM GRAM- STAINED URINE MICROSCOPY IMAGES The present invention provides a computer-implemented system (100) for automated detection and classification of antibiotic-resistant ESKAPE bacterial pathogens directly from microscopic Gram-stained urine sample images. The system comprises an image acquisition interface (110), a preprocessing module (120), and a deep learning inference engine (130) utilizing a novel tri-hybrid ensemble architecture integrating three complementary convolutional neural networks: ResNet50V2 (132) for hierarchical feature extraction, DenseNet121 (134) for dense feature connectivity, and MobileNetV2 (136) for computational efficiency. An attention-based feature fusion module (140) dynamically weights the concatenated 4352-dimensional feature tensor using learned per- feature importance, emphasizing diagnostically salient morphological patterns. The classification head (150) generates probability scores for Escherichia coli, Staphylococcus aureus, and Klebsiella pneumoniae classes. The system achieves 95.24% validation accuracy with median inference time of 156 milliseconds, enabling near-real-time clinical decision support while eliminating the 24-72-hour delay of conventional culture-based methods. A web-based deployment architecture (170, 180, 190) provides clinical interpretability through attention visualization and supports audit traceability. [

Disclaimer: Curated by HT Syndication.