MUMBAI, India, June 30 -- Intellectual Property India has published a patent application (202641076174 A) filed by Vellore Institute Of Technology on June 19, 2026, for Brain Tumor Classification Using Vision Transformer And Kolmogorov-Arnold Networks With Explainable Artificial Intelligence.
Inventors include Iyapparaja M; Arun Kathires Pl; and Yogeshwaran K.
The application for the patent was published on June 26, 2026, under issue no. 26/2026.
Abstract: The present invention relates to an artificial intelligence–based system (100) for brain tumor classification using a hybrid Vision Transformer (ViT) and Kolmogorov- Arnold Network (KAN) architecture. Unlike conventional methods employing MLP-based classification heads, the proposed system (100) utilizes learnable B-spline functions within the KAN framework to enhance model expressiveness and interpretability. The system (100) performs fine-grained classification of 44 tumor types using multi-modal MRI data comprising T1, T1C+, and T2 images. Further, dual explainability techniques, including Grad-CAM and Attention Rollout, are integrated to generate interpretable and clinically relevant outputs. The invention addresses limitations of existing deep learning models by improving classification accuracy, scalability, and transparency, thereby enabling reliable and deployable diagnostic support in medical imaging applications.
Disclaimer: Curated by HT Syndication.
