MUMBAI, India, July 7 -- Intellectual Property India has published a patent application (202641081098 A) filed by Sasi Institute Of Technology And Engineering; M V V A L Sunitha; M. Anantha Lakshmi; Rambabu Pasumarthy; and Yasoda Khumbham on July 01, 2026, for A Hybrid Multimodal Framework For Early Parkinson'S Disease Detection Using Voice Feature Extraction And Cnn-Based Spiral Image Analysis Networks Applicants.

Inventors include M V V A L Sunitha; M. Anantha Lakshmi; Rambabu Pasumarthy; and Yasoda Khumbham.

The application for the patent was published on July 03, 2026, under issue no. 27/2026.

Abstract: Parkinson’s disease (PD) is a progressive neurodegenerative disorder that predominantly affects older adults, causing impairments in motor functions and speech. Early diagnosis is essential for effective disease management but remains challenging due to subtle initial symptoms and the limitations of conventional diagnostic techniques. Most existing detection methods rely on a single data modality, such as voice recordings or handwriting analysis, which often results in lower accuracy and limited practical applicability. To overcome these limitations, this study presents a multimodal Parkinson’s disease detection framework that combines spiral handwriting analysis with voice feature extraction. Convolutional Neural Networks (CNNs) are employed to capture discriminative spatial features from spiral handwriting images, while the Light Gradient Boosting Machine (LightGBM) classifier is utilized for efficient analysis of voice-based features. Feature-level fusion integrates information from both modalities, and Principal Component Analysis (PCA) is applied to reduce dimensionality and improve computational efficiency. The proposed framework achieves an accuracy of 95.27%, outperforming traditional single-modality approaches. By integrating visual and acoustic information, the system provides enhanced robustness and more reliable identification of both motor and speech-related symptoms associated with Parkinson’s disease. These findings demonstrate the potential of multimodal learning for improving early diagnosis and support the development of a non-invasive, cost-effective, and scalable solution suitable for real-world healthcare applications and remote diagnostic systems.

Disclaimer: Curated by HT Syndication.