Osteoporosis Detection in Lumbar Spine L1-L4 Based on Trabecular Bone Texture Features

Abstract

: The proposed work can be divided into two parts: first, remove the noise from the trabecular lumbar spine L1-L4 of the X-ray images using two-stage principal component analysis with neighbourhood pixel grouping, followed by a hybrid median filter, and the texture features are enhanced with sharpening combined with range filters. Second, detect osteoporosis using texture features. This can be done by one-dimensional discrete wavelet transform, followed by a two-dimensional edge detection filter. Finally, classify the normal or osteoporotic images according to conventional classifications. Testing is conducted using X-ray images and dual-energy X-ray absorptiometry (DXA) reports from the same person. The DXA report describes a statistical analysis of normal or osteoporotic, but the proposed work is classified as osteoporotic or normal according to the texture features. Classified results are validated with the DXA and provide an average accuracy of 99.18%. The proposed method has better diagnostic accuracy than the existing method using DXA with X-ray