MUMBAI, India, Sept. 14 -- Intellectual Property India has published a patent application (202641107162 A) filed by Hindustan Institute Of Technology And Science on September 07, 2026, for An Intelligent System And Method For Automated Multi-Region Bone Fracture Detection And Clinical Decision Support Using Deep Learning.
Inventors include Jaya S; and Sheeja Sudheer.
The application for the patent was published on September 11, 2026, under issue no. 37/2026.
Abstract: This invention develops a smart, deep learning system, to automatically detect any type of bone fracture through radiographic images taken from multiple locations. It will include: • A medical imaging acquisition module for capturing multi-region X-ray images. • An image preprocessing module, which includes resizing, normalizing and augmenting images. • A transfer learning-based deep feature extraction module. • A classification module that identifies fractured and non-fractured images. • A clinical decision support module to help healthcare professionals. This invention is a large-scale multi-region x-ray dataset which includes radiographs of upper extremities, lower extremities, hip, knee and lumbar spine. The proposed framework uses various transfer learning based deep neural network algorithms (EfficientNetB0, MobileNetV2, ResNet50, DenseNet121, InceptionV3) to automatically extract highly discriminative features from x-ray images and accurately classify them into bone fractures. In order to identify which of these five advanced architectures provides the greatest detection performance, a detailed performance evaluation was performed. This evaluation included accuracy, precision, recall, F1-score, etc. After conducting this evaluation, DenseNet121 showed the best performance for classifying images since it has a large amount of connections between the dense features and the network is efficient at propagating these features through layers. The proposed framework combines advanced image preprocessing with intelligent feature learning algorithms to enhance robustness and generalization in fracture detection, regardless of differences in image quality, orientation of the fracture, and complexity of anatomy surrounding it. The goal of this invention is to minimize errors in prediction of both false-positives and false-negatives while maximizing real-time computational efficiency which makes it well-suited for clinical application. In addition, it provides a scalable and adaptable infrastructure for expansion into additional skeletal regions; as well as the ability to integrate with HIS, telemedicine tools/applications and CAD environments providing faster, more reliable, and consistent medical decision-making.
Disclaimer: Curated by HT Syndication.
