MUMBAI, India, May 1 -- Intellectual Property India has published a patent application (202641051435 A) filed by Mrs. Anusha D; Sukerthi Sutraya; Chengamma Chitteti; Harshini Nallala; Srija Bandari; Anjali Berelli; and Lahari Chepuri, Hyderabad, Telangana, on April 22, for 'hybrid cnn-transformer architecture for real-time brain stroke risk prediction from mri and physiological signals.'
Inventor(s) include Mrs. Anusha D; Sukerthi Sutraya; Chengamma Chitteti; Harshini Nallala; Srija Bandari; Anjali Berelli; and Lahari Chepuri.
The application for the patent was published on May 1, under issue no. 18/2026.
According to the abstract released by the Intellectual Property India: "Brain stroke remains a critical global health challenge and a major cause of mortality and long-term disability, primarily due to delayed diagnosis, inadequate monitoring, and the absence of continuous risk assessment systems in everyday life. Early identification of stroke risk is essential to enable timely medical intervention and significantly improve survival rates and neurological recovery. However, conventional diagnostic approaches largely depend on periodic clinical evaluations and isolated imaging analysis, which may fail to capture subtle, evolving physiological changes associated with stroke onset. To overcome these limitations, the proposed invention introduces a Hybrid CNN-Transformer Architecture for Real-Time Brain Stroke Risk Prediction from MRI and Physiological Signals, designed to deliver a comprehensive, accurate, and continuous assessment of stroke risk. The system integrates advanced deep learning techniques by combining Convolutional Neural Networks (CNNs) and Transformer models, leveraging their complementary strengths. CNNs are employed for efficient extraction of spatial features from Magnetic Resonance Imaging (MRI) scans, enabling the detection of structural abnormalities, tissue damage, and early ischemic changes. In parallel, Transformer architectures are utilized to capture temporal dependencies and contextual relationships within physiological signals such as heart rate, blood pressure, and other vital parameters. The proposed framework operates on a multimodal data fusion strategy, where imaging data and physiological signals are processed simultaneously and integrated into a unified predictive model. This fusion enhances the system's ability to identify complex patterns and correlations that may not be apparent when analyzing each data source independently. The architecture incorporates attention mechanisms to prioritize critical features, thereby improving prediction accuracy and robustness. A key feature of the system is its capability for real-time monitoring and prediction, enabling continuous risk evaluation in both clinical and non-clinical environments. The model is designed to be computationally efficient, allowing deployment on modern healthcare platforms, wearable devices, and edge computing systems. This ensures that users receive timely alerts and actionable insights regarding potential stroke risk. Furthermore, the proposed approach emphasizes scalability, adaptability, and interoperability with existing medical infrastructure. It can be trained on large-scale datasets and fine-tuned for personalized risk assessment based on individual patient profiles. The system also supports integration with electronic health records (EHRs), enhancing its applicability in hospital settings. Experimental evaluations demonstrate that the hybrid architecture significantly outperforms traditional machine learning and single-modality deep learning models in terms of accuracy, sensitivity, and early detection capability. By enabling proactive healthcare management and early intervention, this invention has the potential to reduce stroke-related mortality and improve quality of life for at-risk individuals."
Disclaimer: Curated by HT Syndication.
