Boosting Diagnostic Accuracy of Osteoporosis in Knee Radiograph Through Fine-Tuning CNN

Osteoporosis is a serious worldwide medical problem that might be challenging to identify promptly owing to the absence of indicators. At the moment, DEXA scans, CT scans, and other techniques with expensive devices and payroll expenses are the mainstays of osteoporosis evaluation. Consequently, an improved, accurate and affordable approach is essential for osteoporosis diagnosis. With the advancement of deep learning, systems for the automated identification of illnesses are regularly presented. Leveraging datasets from chest X-rays accessible for free, the present research assesses the efficacy of several convolutional neural network (CNN) models with the best extreme parameters for osteoporosis detection. Both custom CNN designs and already trained CNN structures for VGG-16 have been incorporated into the assessed system. According to the research results, the VGG-16 with fine-tuning outperformed the one without fine-tuning with an 86.36% accuracy, 86.67% precision, 86.36% recall and 86.34% f1-score, which makes it a potential and reliable model for osteoporosis prediction. The automated diagnosis approach built on CNN can help practitioners promptly, correctly, and reliably identify osteoporosis. This development results from enhanced patient outcomes and increased system productivity.