MUMBAI, India, Aug. 31 -- Intellectual Property India has published a patent application (202641098982 A) filed by Adhiparasakthi Engineering College, Melmaruvathur on August 16, 2026, for Enhanced Breast Cancer Classification Using Grad-Cam And Shap With An Adaptive Learning Rate And Feature Fusion Approach.
Inventors include Dr. C. Dhaya; Ms. A. Jayanthi; S. Divyadharshini; Dr. R. Srivel; Ms. P. Gajalakshmi; and Ms. Tm. Padmapriya.
The application for the patent was published on August 28, 2026, under issue no. 35/2026.
Abstract: Abstract The present invention relates to an explainable deep-learning system and method for the classification of breast cancer histopathological and mammographic images into benign and malignant categories. The invention discloses a Convolutional Neural Network (CNN) based classifier trained using an adaptive learning-rate mechanism, implemented through EarlyStopping and ReduceLROnPlateau callbacks, which dynamically regulates the learning rate during training to accelerate convergence, prevent overfitting and improve prediction stability. The invention further discloses a feature-fusion module that combines low-level and high- level image representations extracted from multiple convolutional layers or pre-trained backbone networks, thereby enhancing discriminatory power beyond what is achievable from a single feature source. To overcome the black-box limitation of conventional CNN-based diagnostic systems, the invention integrates two complementary Explainable Artificial Intelligence (XAI) modules — Gradient-weighted Class Activation Mapping (Grad-CAM), which generates a visual heat-map identifying the image regions responsible for a given prediction, and SHapley Additive exPlanations (SHAP), which quantifies the contribution of individual input features to the classification decision. The resulting system achieves a classification accuracy of 95.2% while simultaneously producing spatial and feature-level explanations that can be verified by a radiologist or pathologist, thereby improving diagnostic reliability, clinical trust and adoption of AI-assisted breast-cancer screening. The invention may be embodied as a software module, a cloud-hosted inference service, or a web-based clinical decision-support portal. Keywords Breast Cancer Classification, Convolutional Neural Network, Explainable Artificial Intelligence, Grad-CAM, SHAP, Adaptive Learning Rate, ReduceLROnPlateau, EarlyStopping, Feature Fusion, Computer-Aided Diagnosis.
Disclaimer: Curated by HT Syndication.
