MUMBAI, India, March 13 -- Intellectual Property India has published a patent application (202641024614 A) filed by Nri Institute Of Technology; Ms. N. Bindu Sri; Ms. M. Tejaswi; Mr. K. Vasu; Mr. G. Aditya; and Mr. G. Siva Sankara Rao, Eluru, Andhra Pradesh, on March 2, for 'thyroid disease prediction by using mbconv, bidirectional lstm with attention mechanism.'
Inventor(s) include Ms. N. Bindu Sri; Ms. M. Tejaswi; Mr. K. Vasu; Mr. G. Aditya; and Mr. G. Siva Sankara Rao.
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 proposes an advanced deep learning-based system for thyroid disease prediction utilizing a hybrid architecture that integrates Mobile Inverted Bottleneck Convolution (MBConv), Bidirectional Long Short-Term Memory (BiLSTM), and an attention mechanism. The MBConv layers are employed for efficient and high-level feature extraction from structured clinical datasets, enabling the model to capture complex nonlinear relationships among medical attributes. The BiLSTM network models bidirectional dependencies between clinical parameters, allowing the system to learn contextual interactions that influence thyroid diagnosis. An attention mechanism is incorporated to assign adaptive weights to the most relevant clinical features, thereby improving prediction accuracy and interpretability. To address class imbalance commonly found in medical datasets, the Synthetic Minority Over-sampling Technique (SMOTE) is applied. The model is trained using the Adam optimizer with cross-entropy loss to ensure stable convergence. Experimental evaluation demonstrates superior accuracy, robustness, and reliability compared to conventional machine learning approaches, supporting effective and automated thyroid disease prediction."
Disclaimer: Curated by HT Syndication.
