MUMBAI, India, Sept. 28 -- Intellectual Property India has published a patent application (202611099722 A) filed by Sachin Kumar; Dr. Sourabh Shastri; and Prof. Vibhakar Mansotra on August 18, 2026, for An Ai-Based Diagnostic Framework For Neurological Disorders Using Super-Resolution Functional Neuroimaging And Nature-Inspired Optimized Hybrid Capsule-Transformer Model.

Inventors include Mr. Sachin Kumar; Dr. Sourabh Shastri; and Prof. Vibhakar Mansotra.

The application for the patent was published on September 25, 2026, under issue no. 39/2026.

Abstract: An AI-assisted differential diagnostic system and computerized method are described for automated multi-class differential diagnosis of neurological disorders, viz. Parkinson’s Disease (PD), Schizophrenia (SZ), and Healthy Control (or Control) groups based on fMRI data sets. The diagnostic system receives bicubic super- resolution normalized fMRI slices (128×128×1) using a multi-level convolutional network to form primary capsule vectors (4096×8) by processing spatial feature activation. A dynamic routing capsule layer (8 capsules, size 8, 3 routings) is used to apply the nonlinear squashing vector function, thereby conserving spatial orientation and part-whole associations. A multi-head self-attention Transformer model (2 heads, key size 8) with layer normalization is used to capture global context dependencies from capsule layer output. The diagnostic system employs the Ant Colony Optimization (ACO) algorithm to dynamically find optimal learning-rate hyperparameters. The AI-assisted diagnostic system attains an average multi-class classification accuracy of 99.29%±0.19%, macro F1-score of 99.29%±0.19%, Matthews Correlation Coefficient of 0.9879, and Cohen’s Kappa of 0.9879 across 10-fold cross-validation with statistical significance (p 0.05). Moreover, the diagnosis process involves the use of a telemedicine interface integrated with the Internet of Things (IoT) that provides a Clinical Decision Support System (CDSS) GUI to provide super-resolution activation maps and multi-class probability for tele-neurology diagnosis.

Disclaimer: Curated by HT Syndication.