A Novel Approach for Predicting Osteoporosis Using Texture Analysis on Magnetic Resonance Images

Abstract

Osteoporosis (OP), a particularly common disease among people above 65 years, makes bones weak and brittle. This predisposes an individual to fragility fractures that could result in fatalities. Currently, Dual Energy X-Ray Absorptiometry (DXA) is the state-of-the-art technique for diagnosing OP. But people often postpone DXA scans due to limited compliance resulting in OP remaining undiagnosed until it becomes problematic. The elderly, on the other hand, often get Magnetic Resonance Imaging (MRI) scans for other indications. The purpose of this research is to demonstrate that these MRI scans could be simultaneously used as a method to predict OP. We used MRI and DXA scans of 13 female and 13 male individuals obtained from the NIH Osteoarthritis Initiative. We considered three types of MRI scans: TSE, 2D MESE, and 3D DESS. Next, we performed texture analysis using a software tool called MaZda on the MRI scans at three different regions of interest (ROIs): one on the distal femoral head and two on the proximal tibial head. We then used Pearson and Spearman correlation to measure the strength of associations between the bone texture obtained from MRI scans via the texture analysis, and bone mineral density (BMD) obtained from DXA scans. Our results showed that several texture features of bone MRI, including wavelet parameters, contrast, and mean, correlate significantly with BMD values. This suggests that data collected from MRI scans can be used as a predictor for OP.