MUMBAI, India, July 24 -- Intellectual Property India has published a patent application (202641058290 A) filed by Manipal Academy Of Higher Education on May 07, 2026, for A System And Method For Dual Attention Based Convolutional Brain Tumor Classification.
Inventors include Gouranga Mandal; Poushali Chakraborty; and Tamal Biswas.
The application for the patent was published on July 17, 2026, under issue no. 29/2026.
Abstract: The present invention discloses a lightweight self-attention based convolutional neural network system and method for multiclass brain tumor classification from magnetic resonance imaging scans. The system preprocesses input scans through contrast-limited adaptive histogram equalization and pixel intensity normalization, followed by hierarchical feature extraction using residual convolutional stages with channel-wise squeeze-and-excitation recalibration. A spatial self-attention module captures long-range spatial dependencies across the feature map, and a fully connected classification head produces a probability distribution over four diagnostic tumor classes, namely glioma, meningioma, pituitary tumor, and no tumor. The system achieves an overall classification accuracy of 98 percent with a macro F1-score of 0.98, containing about 1.5 million parameters. Intrinsic attention maps generated during inference enhance clinical interpretability, and the compact model size of about 5.73 megabytes supports deployment in resource-constrained diagnostic environments. FIG. 4
Disclaimer: Curated by HT Syndication.
