Multimodal Analysis of Bone Mineral Density Classification

Osteoporosis is a skeletal disorder marked by diminished bone mineral density and mass, impacting bone health, mass, or alterations in bone structure and strength. This decline in bone strength heightens the susceptibility to fractures. Notably, osteoporosis is often asymptomatic, making it a "silent" ailment. The assessment of deficiencies in joint images is a prolonged procedure conducted manually, demanding the expertise of professionals to periodically analyze the scans. This manual approach results in time delays and heightened costs. To address this issue efficiently, this work aims to pinpoint bone mineral density, a crucial factor in identifying bone-related diseases and assessing the risk of fractures in an automated fashion by comparing Dual Energy X-ray Absorptiometry and X-ray images. The workflow is initiated with image pre-processing followed by Image enhancement, Data Augmentation, Data partition, feature extraction and classification. The images are enhanced using a median filter, normalization techniques and CLAHE - Contrast Limited adaptive histogram equalization algorithm. VGG-16, ResNet, DenseNet, InceptionV3 and Inception-ResNet are the different transfer learning models employed in this work to analyze and extract the best algorithm that provides the desired results.