MUMBAI, India, Aug. 31 -- Intellectual Property India has published a patent application (202641100458 A) filed by S. Hrushikesava Raju; and Koneru Lakshmaiah Education Foundation on August 19, 2026, for An Intelligent Hybrid Deep Learning For Liver Disease Prediction Using Multimodal Fusion.
Inventors include S. Hrushikesava Raju; Srinivasa Rao Adapa; Dr. M Nagaraju; Talasila Bhanu Siva Prasad; Dr. A Vishnuvardhan Reddy; C. Lakshminatha Reddy; Kalakoti Gowrisankar; K. Selvam; S. Sherisha; Dr. Suresh Betam; and G Venkata Ramana Reddy.
The application for the patent was published on August 28, 2026, under issue no. 35/2026.
Abstract: The symptoms such as abdominal pain, fatigue, jaundice, weight loss, and fluid accumulation indicate liver cancer (hepatocellular carcinoma (HCC)). As the disease progresses, leads to liver failure, which is life-threatening if not treated early. Existing approaches, such as logistic regression, decision trees, support vector machines (SVM), and standalone ensemble models, suffer from robustness issues, are unable to handle large datasets, struggle with noise, and focus only on clinical aspects. Then, a hybrid model focuses on designing a supervised machine learning (ML) framework that leverages structured clinical variables, including demographic, biochemical, and lifestyle features, to predict the onset of liver disease at an early stage. The hybrid model consists of advanced preprocessing and feature optimization techniques, such as Mutual Information (MI) and feature selection using Boruta-based, by addressing data imbalance using SMOTE, and improves learning stability. Furthermore, combining Random Forest, XGBoost, and LightGBM classifiers, where each model captures complementary feature patterns and nonlinear interactions among key liver biomarkers (alanine aminotransferase (ALT), aspartate aminotransferase (AST), bilirubin, and albumin levels). The predictions of these learners are integrated to produce a highly optimized output. In addition, explainable artificial intelligence (XAI) techniques such as SHAP provide insights to support effective clinical decision-making. A trained hybrid ML classifier achieves high accuracy, precision, recall, and F1-score compared with other models considered, indicating its effectiveness.
Disclaimer: Curated by HT Syndication.
