Multi-detector computed tomography (MDCT) imaging: association of bone texture parameters with finite element analysis (FEA)-based failure load of single vertebrae and functional spinal units

Abstract

Background: Osteoporosis is a systemic skeletal disease that is characterized by low bone mass and microarchitectural deterioration, predisposing affected individuals to fragility fractures. Yet, standard measurement of areal bone mineral density (BMD) in dual-energy X-ray absorptiometry (DXA) as the current reference standard has limitations for correctly detecting osteoporosis and fracture risk, with opportunistic osteoporosis screening using computed tomography (CT) showing increasing importance. This study's objective is to compare finite element analysis (FEA)-based vertebral failure load with parameters of texture analysis (TA) derived from multi-detector CT (MDCT). Methods: MDCT data of seven subjects (mean age: 71.9±7.4 years) were included for FEA and TA. Manual segmentation was performed for the vertebral bodies T11, T12, L1, and L2 and the intervertebral discs (IVDs) T11/12, T12/L1, L1/2, and L2/3. Correlation analyses between FEA-derived failure loads and parameters of TA for the single vertebrae and two functional spinal units (FSUs) were calculated, defining FSU-1 as T11-IVD-T12-IVD-L1 and FSU-2 as T12-IVD-L1-IVD-L2. Furthermore, multivariate regressions were performed to identify the texture parameters that predicted the failure load best. Results: For single vertebrae, the strongest correlations were observed for skewnessglobal, kurtosisglobal, and gray level variance (rho =−0.7668 to −0.7362; P<0.001), while for FSUs, SumAverage, long-run emphasis, long-run low gray-level emphasis, homogeneity, and energy showed the strongest correlations (rho =−0.8187 to 0.8407; P<0.05) to failure loads. SumAverage best predicted the failure load for single vertebrae (R 2 adj =0.523, P<0.001). For the two FSUs, kurtosisglobal (FSU-1: R 2 adj =0.611, P=0.001) and skewnessglobal (FSU-2: R 2 adj =0.579, P=0.002) were the best predictors. Conclusions: TA using MDCT data of the spine was significantly associated with FEA-derived failure loads of both, single vertebrae and FSUs. Texture parameters predicted failure loads of FSUs as a more realistic in-vivo scenario equally well as compared to single vertebrae analyses. TA may reflect a less complex and time-consuming approach to accurately and non-invasively evaluate vertebral bone strength.

The value of radiomics-based CT combined with machine learning in the diagnosis of occult vertebral fractures

“…(1)Patient with fresh vertebral fractures with bone marrow edema confirmed by thoracic/lumbar MRI. Bone marrow edema was identified as significantly high signal on STIR and T2WI sequences and low signal on T1WI; (2)with chest/abdominal CT scans within 48 h of the MR examination; (3) Patients with at least one fresh fracture vertebra recognized as OVFs ,that is the fractural vertebral without visual reduction in height (Genant grade = 0) on initial CT imaging [ 12 ].…”

“…Zaworski et al [ 11 ] found that MRI-based TA of trabecular bone could evaluate vertebral bone fragility while discriminating between osteoporosis patients with or without a history of fracture. Sollmann et al [ 12 ] reported that some texture features, based on CT of the vertebral body, showed a strong correlation with finite element analysis results, indicating that radiomics parameters can be used to evaluate the vertebral load. Tabari et al [ 13 ] demonstrated that some texture parameters of vertebral trabeculae on lumbar spine CT in patients with anorexia nervosa were different than those in the healthy control group, suggesting that TA might provide information about bone health.…”

Texture Analysis Using CT and Chemical Shift Encoding-Based Water-Fat MRI Can Improve Differentiation Between Patients With and Without Osteoporotic Vertebral Fractures

“…On the basis of the distribution of gray-level values, texture analysis was used to characterize structural image properties of predefined regions by quantifying different texture features (37)(38)(39). Texture analysis was performed using the segmentation masks derived from CT and CSE-MRI enclosing single vertebral bodies, and different first-order statistical moments from global gray-level histograms, second-order features based on the graylevel co-occurrence matrix (GLCM), and higher-order features based on the gray-level run-length matrix (GLRLM) were extracted (Figure 1 and Table 1) (40,41,51,(53)(54)(55).…”