MUMBAI, India, March 13 -- Intellectual Property India has published a patent application (202641024953 A) filed by Rajarajeswari College Of Engineering, Bengaluru, Karnataka, on March 3, for 'brain tumor segmentation and modified classification using enhanced deep learning algorithm.'
Inventor(s) include Dr. A. Muruganadham; and Prof. Karthikeyan Radhakrishnan.
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 present invention relates to the development of a segmentation and classification of brain tumors are still open problems in medical imaging, where accurate demarcation of tumor areas from multi-modal MRI images directly affects diagnosis and treatment strategies. To this end, this paper proposes an Enhanced Deep Learning Algorithm (EDLA), which extends conventional U-Net models by incorporating attention modules, residual learning, and a specially designed multi-scale feature fusion module to enhance segmentation performance on intricate tumor boundaries. For classification, EDLA relies on a hybrid convolutional-recurrent neural network with transfer learning from pre-trained models such as ResNet-152, supplemented with a specially designed loss function that combines the Dice loss and focal loss to handle class imbalance for glioma, meningioma, and pituitary tumor types. Tested on the BraTS 2023 challenge, EDLA yields a Dice metric of 0.94 for segmentation and 98.2% accuracy for classification, respectively 5-7% relative improvements over strong baselines such as the standard U-Net (Dice = 0.89) and nn U-Net."
Disclaimer: Curated by HT Syndication.
