Intellectual Property India Publishes Patent Application for An Explainable Artificial Intelligence Based Deep Learning System For Automated Breast Cancer Detection And Multi-Class Classification Of Histopathological Images

MUMBAI, India, Sept. 7 -- Intellectual Property India has published a patent application (202641104117 A) filed by Sankar Krishnamurthi; and Vairachilai Shenbagavel on August 29, 2026, for An Explainable Artificial Intelligence Based Deep Learning System For Automated Breast Cancer Detection And Multi-Class Classification Of Histopathological Images.

Inventors include Sankar Krishnamurthi; and Vairachilai Shenbagavel.

The application for the patent was published on September 04, 2026, under issue no. 36/2026.

Abstract: The present invention supports a Deep Learning System based on Explainable Artificial Intelligence (XAI) for Automated Detection and Multi-Class Classification of Breast Cancer with Histopathological Images. The system provides input for digitized histopathological image of the breast, and then processes it to do pre-processing, stain normalization, identification, and extraction of tissue region images. A convolutional neural network and a Transformer/attention-based network extracts the complementary local morphological and contextual features respectively, in a hybrid deep learning architecture. Based on extracted representations, an adaptive feature fusion module fuses the extracted features to produce a single representation of features, which is fed into a multi-class classification module to give values for the class probabilities for the pathologies and the prediction of a class. A module providing a confidence estimation, generates a confidence value related to the prediction. A spatial importance map, one or several explainability techniques, is calculated by an explainable AI module with respect to a classification decision. The second module used for generating a heat map spatially overlaps the importance map on the image used as a histological sample. The system also handles whole slide image analysis by making predictions with a patch level and aggregating these predictions in space. The diagnostic visualization consists of the predicted pathological class, probability distribution, confidence value, and the explanation heatmap to give a computer-assisted and interpretable approach for automated breast histopathological image analysis.

Disclaimer: Curated by HT Syndication.