Osteoporosis Diagnosis through Visual Segmentation and Classification: Extensive Review

Abstract

Osteoporosis, a prevalent and progressive bone disorder, is characterized by diminished bone density and quality, resulting in heightened fragility and susceptibility to fractures. Various factors such as age, gender, hormonal fluctuations, and genetics contribute to the development of osteoporosis. Given its widespread prevalence, accurate and timely diagnosis is essential, prompting the exploration of advanced diagnostic methodologies. This survey aims to review and analyze the utilization of advanced technologies in bone health assessment, with a focus on image segmentation and deep learning models. Specifically, Convolutional Neural Networks (CNNs) are employed for feature extraction, followed by the application of transfer learning algorithms such as VGG-19, U-Net, and RESNET. The integration of pre-trained models and transfer learning techniques enhances the models' capability to discern variations in osteoporosis severity. Evaluation metrics including F1score, recall, and confusion matrix are utilized to assess the performance of the reviewed approaches.