Osteoporosis Prediction Using VGG16 and ResNet50

Abstract

Low bone mass and structural degradation are the hallmarks of osteoporosis, a disorder that increases the risk of fractures, especially in the elderly. For prompt intervention and fracture prevention, early identification is essential. However, osteoporosis is frequently not detected until advanced stages by existing diagnostic techniques. In order to overcome this difficulty, scientists suggest using machine learning to automatically identify osteoporosis early in X-ray pictures. Utilizing two cutting- edge convolutional neural network architectures, ResNet50 and VGG16, their system was pretrained on extensive datasets and refined on a carefully selected dataset of X-ray pictures. When identifying images as suggestive of osteoporosis or normal bone density, the ResNet50 model showed an accuracy of 98%, whereas the VGG16 model achieved 78% accuracy. By combining these models and using sophisticated image segmentation methods, the system detects early osteoporosis indications with an overall accuracy of 96%. This automated method has the potential to decrease the incidence of fractures linked to osteoporosis, enable early treatment initiation, and increase the rate of early diagnosis.

Ähnliche Arbeiten