A Deep Learning Approach for Knee Osteoporosis Screening Using Convolutional and Transfer Learning Architectures on Radiographic Images

Inchara Y; Supreetha Gowda H D

doi:10.70528/IJLRP.v7.i7.2285

A Deep Learning Approach for Knee Osteoporosis Screening Using Convolutional and Transfer Learning Architectures on Radiographic Images

Author(s)	Ms. Inchara Y, Dr. Supreetha Gowda H D
Country	India
Abstract	Osteoporosis is a progressive bone disorder that reduces bone mineral density and substantially increases fracture risk, particularly at weight-bearing joints such as the knee. Manual radiographic assessment of osteoporosis is reliable but depends on expert availability and can be time-consuming when large volumes of images require review, and subtle early-stage density loss (osteopenia) is often difficult to identify by visual inspection alone. This study presents an automated, image-based screening approach that classifies knee radiographs into three diagnostic categories — Normal, Osteopenia, and Osteoporosis — using a custom Convolutional Neural Network (CNN) alongside three transfer-learning architectures (VGG16, DenseNet121, and ResNet50). Images from a publicly available, class-labeled knee X-ray dataset are resized, normalized, and augmented (rotation, shifting, zooming, flipping) before training to improve generalization and reduce overfitting. The trained models are evaluated and compared using accuracy, precision, recall, F1-score, and AUC, alongside per-class confusion matrices. Among the four architectures, the custom CNN and ResNet50 achieved the strongest performance, with overall accuracies of 90.4% and 92.2% respectively, while VGG16 and DenseNet121 achieved comparatively lower accuracies of 72.4% and 71.7%. The best-performing model is deployed through a Flask-based web application that accepts an uploaded knee radiograph and returns the predicted class, confidence score, class-wise probabilities, and a supportive suggestion. The results indicate that deep convolutional and residual architectures can provide a practical, reproducible first-pass screening aid for knee osteoporosis, intended to support — rather than replace — clinical diagnosis.
Keywords	Knee osteoporosis; osteopenia; convolutional neural network; transfer learning; ResNet50; VGG16; DenseNet121; medical image classification.
Field	Computer > Artificial Intelligence / Simulation / Virtual Reality
Published In	Volume 7, Issue 7, July 2026
Published On	2026-07-11
DOI	https://doi.org/10.70528/IJLRP.v7.i7.2285

View / Download PDF File

About IJLRP Fees & Payment Current Issue Publication Archive	Submit Research Paper Track Submission Status Publication Guidelines Publication Ethics	Join as a Reviewer Editors & Reviewers Get Reviewer Membership Certi.	Website/Journal Policies Usage Policy Content Policies Privacy Policy

Contact Us		+91-9687-828-838	editor@ijlrp.com

International Journal of Leading Research Publication

A Widely Indexed Open Access Peer Reviewed Multidisciplinary Monthly Scholarly International Journal

A Deep Learning Approach for Knee Osteoporosis Screening Using Convolutional and Transfer Learning Architectures on Radiographic Images

Share this