MUMBAI, India, Aug. 31 -- Intellectual Property India has published a patent application (202631101488 A) filed by Dr. Subrata Sinha; Saurav Mali; Suvomoy Datta; Dr. Utpala Borgohain; Dr. Ujjal Saikia; Dr. Surabhi Johari; and Tanushri Chatterjii on August 22, 2026, for A System For Cbam-Driven Attention-Based Cnn Architecture For Lung Cancer Detection.
Inventors include Dr. Subrata Sinha; Saurav Mali; Suvomoy Datta; Dr. Utpala Borgohain; Dr. Ujjal Saikia; Dr. Surabhi Johari; and Tanushri Chatterjii.
The application for the patent was published on August 28, 2026, under issue no. 35/2026.
Abstract: The present invention discloses a CBAM-driven attention-based convolutional neural network architecture for lung cancer detection from histopathological images. The system comprises three sequential convolutional blocks each containing two convolutional layers with 7 by 7 filters at filter depths of sixteen, thirty-two, and sixty-four channels, one max-pooling layer, and one Convolutional Block Attention Module applying sequential channel and spatial attention mechanisms. A flatten layer converts extracted feature maps into one-dimensional vectors processed through three fully connected dense layers. A softmax output layer produces probability distributions across Lung Adenocarcinoma, Lung Squamous Cell Carcinoma, and Non-Malignant categories. The model is trained on 11,580 lung histopathological images at 512 by 512 pixel resolution using Adam optimizer at learning rate 0.0003, batch size sixty- four, fifty epochs with early stopping at patience five. The system achieves 89.57 percent accuracy and area under the curve of 0.94, providing improved discriminative feature representation, enhanced classification reliability, and robust lung cancer subtype detection.
Disclaimer: Curated by HT Syndication.
