Intellectual Property India Publishes Patent Application for 'Hybrid Attention Mechanism For Feature Extraction And Classification In Retinal Imaging' Filed by A. T. R. Krishna Priya; Dr. Ramya R. R.; Dhanya G S; Bindhu A; Manikandan R A; P. Kathambari; Enoch Raja DG; Mary Joycy A; Dr. N. Janaki; and Stanley Karunakaran W

MUMBAI, India, March 13 -- Intellectual Property India has published a patent application (202641025467 A) filed by A. T. R. Krishna Priya; Dr. Ramya R. R.; Dhanya G S; Bindhu A; Manikandan R A; P. Kathambari; Enoch Raja DG; Mary Joycy A; Dr. N. Janaki; and Stanley Karunakaran W, Kanyakumari, Tamil Nadu, on March 4, for 'hybrid attention mechanism for feature extraction and classification in retinal imaging.'

Inventor(s) include A. T. R. Krishna Priya; Dr. Ramya R. R.; Dhanya G S; Bindhu A; Manikandan R A; P. Kathambari; Enoch Raja DG; Mary Joycy A; Dr. N. Janaki; and Stanley Karunakaran W.

The application for the patent was published on March 13, under issue no. 11/2026.

According to the abstract released by the Intellectual Property India: "The proposed invention presents a hybrid attention mechanism for feature extraction and classification in retinal imaging, addressing the challenges of accurate disease detection and diagnosis. Retinal imaging plays a crucial role in identifying conditions such as diabetic retinopathy, glaucoma, and age-related macular degeneration. However, conventional methods often struggle with complex retinal structures and background noise, leading to misclassifications. This invention enhances retinal image analysis by incorporating spatial and channel attention mechanisms, ensuring that the most relevant pathological features are prioritized. The process begins with preprocessing using the Weiner filter, which reduces noise and improves image clarity while preserving essential retinal details. Following this, SIFT is employed for feature extraction, ensuring robust representation of structural variations. The extracted features are then processed using a hybrid attention-fused DenseNet model, where spatial attention highlights significant regions such as hemorrhages and exudates, while channel attention refines feature selection by prioritizing the most informative feature maps. By dynamically focusing on critical features, this hybrid model significantly improves classification accuracy and enhances early disease detection."

Disclaimer: Curated by HT Syndication.