Intellectual Property India Publishes Patent Application for A Hybrid Dual-Modality Machine Learning System For Early Detection Of Breast And Lung Cancer From Medical Images And Pdf-Based Clinical Reports

MUMBAI, India, July 13 -- Intellectual Property India has published a patent application (202641083791 A) filed by Nitte Meenakshi Institute Of Technology, Nitte Deemed To Be University, Bengaluru; Dr. Sujata Joshi; Abinash Yadav; Aman Malik; and Vrishab Magothra on July 08, 2026, for A Hybrid Dual-Modality Machine Learning System For Early Detection Of Breast And Lung Cancer From Medical Images And Pdf-Based Clinical Reports.

Inventors include Nitte Meenakshi Institute Of Technology, Nitte Deemed To; Dr. Sujata Joshi; Abinash Yadav; Aman Malik; and Vrishab Magothra.

The application for the patent was published on July 10, 2026, under issue no. 28/2026.

Abstract: ABSTRACT OF THE INVENTION Title: A Hybrid Dual-Modality Machine Learning System for Early Detection of Breast and Lung Cancer from Medical Images and PDF-Based Clinical Reports The invention relates to a hybrid dual-modality machine learning system for the early detection of breast and lung cancer that integrates a Convolutional Neural Network (CNN) for medical image classification with a Logistic Regression model for PDF clinical report classification, deployed as a unified Django web application accessible to clinical staff without specialized computing infrastructure. The system comprises: a Web Interface (Block 100) built with Django and Tailwind CSS providing role-based Doctor and Administrator access with image (JPG/PNG) and PDF upload modules; a Preprocessing Unit (Block 200) implementing a parallel image pipeline (OpenCV resize to 224×224, Gaussian Blur, pixel normalisation [0,1]) and PDF pipeline (pdfplumber extraction, regex parsing of 30 features, mean imputation, z-score normalisation); a CNN Model (Block 300) with TensorFlow/Keras architecture comprising 3×Conv2D(32,64,128), MaxPool, Dropout(0.25), Dense(128), Sigmoid achieving 93.5% accuracy/94.8% recall for breast cancer and 91.2% accuracy/92.6% recall for lung cancer image classification; a Logistic Regression Model (Block 400) with scikit-learn liblinear/L2 solver on 30 Wisconsin Breast Cancer Dataset features achieving 95.6% accuracy/0.99 recall; and a Prediction Output Module (Block 500) delivering Malignant/Benign classification with confidence scores in 3–5 seconds per prediction. The present invention provides a flexible, scalable, and interpretable AI-driven cancer screening platform applicable to diverse clinical settings including primary healthcare centres and telemedicine platforms lacking expert radiological resources. Trained on open-source datasets (Wisconsin Breast Cancer UCI, MIAS, DDSM, LUNA16), it bridges the gap between machine learning research and practical clinical deployment by unifying image-based deep learning and report-based classical ML within a single accessible web application, substantially advancing early cancer detection capability in resource-limited healthcare environments.

Disclaimer: Curated by HT Syndication.