MUMBAI, India, Aug. 12 -- Intellectual Property India has published a patent application (202641094169 A) filed by Karpagam College Of Engineering; Karpagam Academy Of Higher Education; Ms. Aarthi D; and Monisha S on August 04, 2026, for A Deep Learning-Based Multimodal System For Early Detection Of Autism Spectrum Disorder (asd).
Inventors include Ms. Aarthi D; and Monisha S.
The application for the patent was published on August 07, 2026, under issue no. 32/2026.
Abstract: Autism Spectrum Disorder (ASD) affects children’s communication and behavior, but early detection remains challenging due to varied symptoms and reliance on observational diagnosis. This project proposes a deep learning–based multimodal system for early ASD prediction using brain MRI, eye-tracking data, voice and speech patterns, and body movement. Specialized neural networks are used for each data type, including 3D- CNNs for MRI scans, 2D-CNNs for eye tracking, RNNs/LSTMs for speech and voice, and Temporal Convolutional Networks for body motion. The models are trained independently and combined using late fusion to produce a final prediction. Data preprocessing includes normalization, scaling, and augmentation, with training performed using the Adam optimizer and cross-entropy loss. Stratified k-fold validation ensures reliable results. Experiments on datasets such as ABIDE, Figshare, and MMASD show that multimodal learning significantly improves prediction accuracy. Explainability tools like Grad-CAM and SHAP enhance model transparency and clinical trust.
Disclaimer: Curated by HT Syndication.
