MUMBAI, India, Jan. 23 -- Intellectual Property India has published a patent application (202641001761 A) filed by Karpagam Academy Of Higher Education; Karpagam Institute Of Technology; Dr. R Priyadharshini; and Karthikeyan T, Coimbatore, Tamil Nadu, on Jan. 7, for 'deep learning based tumour detection.'
Inventor(s) include Dr. R Priyadharshini; and Karthikeyan T.
The application for the patent was published on Jan. 23, under issue no. 04/2026.
According to the abstract released by the Intellectual Property India: "Tumor detection through medical imaging plays a vital role in early diagnosis and treatment planning, significantly influencing patient survival and clinical decision-making. Traditional diagnostic methods, which rely heavily on manual inspection of MRI, CT, and histopathological images by radiologists, are often constrained by subjectivity, time consumption, and potential human error. To address these limitations, the present invention proposes an advanced deep learning-based tumor detection and classification system designed to automate and enhance the accuracy of medical image analysis. The system leverages state-of-the-art Convolutional Neural Networks (CNNs) and modern architectures, including ResNet, VGG16, EfficientNet, and U-Net, to extract both low-level and high-level image features for precise tumor identification. The system operates through a comprehensive multi-stage pipeline beginning with data preprocessing, which includes noise reduction, normalization, and augmentation to improve image quality and ensure robust model performance. Following preprocessing, a segmentation module employing U-Net or Attention U-Net generates precise region-of-interest (ROI) masks, enabling accurate localization of tumor regions. The extracted ROI is then processed by a deep CNN model capable of learning complex patterns such as tumor shape, size, boundary irregularities, and textural variations. A classification module subsequently evaluates the extracted features to differentiate between benign and malignant tumors using fully connected neural layers and probability estimators. To improve transparency and clinical trust, the system incorporates explainable AI techniques, including Grad-CAM and saliency maps, which highlight critical regions influencing the model's predictions. Additionally, a continuous learning mechanism is integrated to refine the system's performance as more annotated data becomes available. The proposed invention provides a fast, reliable, and scalable solution for automated tumor detection, significantly reducing the diagnostic workload of radiologists while improving consistency and accuracy. Its integration within clinical workflows enables enhanced diagnostic support, timely intervention, and improved patient outcomes across diverse healthcare environments."
Disclaimer: Curated by HT Syndication.
