Assessment Of Bone Quality In Osteoporotic Women Through 3t MRI Radiomics Of The Proximal Femur

Abstract

Background: Osteoporosis is a systemic skeletal condition characterized by low bone strength, which increases the risk for fragility fractures in particular after menopause. Although traditional assessment through dual-energy X-ray absorptiometry (DXA) provides information about bone mineral density (BMD), it does not assess the microarchitectural integrity that affects true bone quality. The use of high-field magnetic resonance imaging (MRI), especially at 3 Tesla (3T), and radiomics permits non-invasive assessment of trabecular bone and cortical microstructure by extracting quantitative image-based biomarkers. The objective of the current investigation was to assess bone quality in osteoporotic women using the proximal femur in 3T MRI radiomics to correlate radiomic features to BMD based on DXA.

Materials and Methods: An observational study was carried out involving 80 postmenopausal women aged between 50 and 75. The participants were classified into two groups, one with 40 women with osteoporosis (T-score ≤ −2.5) and another with 40 age-matched healthy controls. MRI of the proximal femur was obtained on a 3T system using T1-weighted and proton density sequences. The region of interest (ROI) was manually segmented and radiomic features were obtained that included first order statistics, gray-level co-occurrence matrix (GLCM), and gray-level run-length matrix (GLRLM) parameters. Statistical analysis included Pearson correlation between radiomic features and BMD as well as ROC analysis for diagnostic discrimination.

Results: Participants with osteoporosis showed significantly increased values for entropy, skewness, and run-length nonuniformity compared to controls (p < 0.01), which indicate greater spatial textural heterogeneity. There was a notable decrease in homogeneity and uniformity. Strong negative correlations were noted between BMD and GLCM derived entropy (r = −0.71, p < 0.001), and positive correlations with uniformity (r = 0.68, p < 0.001). The best combination of radiomic features was able to classify osteoporotic bone from normal bone with an accuracy of 88%.

Conclusion: Using 3T MRI radiomics of the proximal femur, we have developed a non-invasive biomarker-based method to quantitatively measure bone quality in women with osteoporosis. The texture features extracted reflect the constituents of microarchitectural deterioration and have the potential to enhance DXA in the early detection, risk stratification, and monitoring of osteoporosis.