Osteoporosis is a systemic skeletal metabolic disease that causes bone loss and destruction of bone microstructure, leading to reduced bone strength and an elevated risk of fracture. As of August 2020, the global prevalence of osteoporosis was approximately 18.3%, with a prevalence of approximately 23.1% in women and 11.7% in men [11]. Osteoporotic fractures have a serious impact on a patient’s quality of life and produce a heavy economic and social burden on the family and society. More than two million fragility fractures occurred in the United States in 2005, costing $17 billion. The incidence and cost of osteoporotic fractures are projected to increase by nearly 50% by 2025 [12]. The majority of these patients who suffered fragility fractures had not been screened by DXA, or DXA scanning had not revealed abnormal BMD values. In a large observational study, less than half of the women (44%) and even fewer men (21%) with osteoporotic fractures had low aBMD values [13]. Therefore, it is necessary to find an alternative to DXA as a screening tool for osteoporosis.
BMD and bone marrow fat are two important biological indicators for diagnosing osteoporosis [14]. Measurements of BMD rely mainly on DXA and QCT. DXA is a simple, low-radiation, low-cost method for measuring BMD, and it is widely used in clinical practice. However, previous literatures have confirmed that two-dimensional BMD measurements are influenced by degenerative changes in the spine, scoliosis, and arteriosclerosis in the abdominal aorta [15]. QCT, which was introduced in 1977, is capable of measuring BMD values in three dimensions, thereby providing greater sensitivity and accuracy in diagnosing osteoporosis and assessing fracture risk; however, it entails high doses of radiation [16]. Lin et al. investigated inconsistencies between DXA and QCT measurements in lumbar BMD values and found that approximately 61.1% of subjects were diagnosed with osteoporosis according to QCT measurements, and 54.1% of subjects were diagnosed with osteoporosis according to DXA, of which only 65.5% of subjects were diagnosed consistently [17]. In the present study, DXA and QCT scans were used to measure BMD values, in combination with the MRI technique for assessing BMFF. It is a novel, effective, non-invasive, and non-ionizing method for diagnosing osteoporosis indirectly using MRI examination.
Marrow adipose tissue (MAT), which accounts for 10% of total body fat, was once considered a passive fat store, but more recent studies have shown that MAT is an autocrine and paracrine tissue that can produce lipotoxic effects and modulate immune cell responses [18]. Obviously, with aging, estrogen deficiency, and glucocorticoid exposure, MSCs are more likely to differentiate into adipocytes, resulting in loss of bone mass [19]. Justesen et al. reported an increase in the percentage of adipocytes in cadaveric iliac bone, from 40% (age 30 years) to 68% (age 100 years) [20]. In addition, Griffith et al. noted that bone marrow adiposity was significantly higher in postmenopausal women compared to men of the same age [21]. It is well known that adipokines and free fatty acids released from adipocytes can directly or indirectly interfere with bone formation and bone remodeling by acting on osteoblasts and osteoclasts through different signaling pathways, resulting in an imbalance in bone metabolic activity. Saedi et al. observed a decrease in osteoblasts and an increase in osteoclasts in rats with high fat content, accompanied by a deterioration in the microarchitecture of bone trabeculae, suggesting that bone marrow fat content is closely related to changes in bone marrow cell density and bone mass [22]. Moreover, Zhu et al. found a significant increase in bone marrow fat content over time in osteoporotic rats, and it was negatively correlated with bone microstructural parameters [23]. Bone marrow fat quantification techniques mainly include MRS and CSE-MRI, which can accurately quantify bone marrow fat composition, and the results of the two methods are consistent to a large extent.
MRS is considered by many scholars to be the gold standard for quantitative analysis of bone marrow fat in vivo. Cohen et al. used MRS to measure lumbar FF values and compared them with iliac bone FF values measured by biopsy, and he ultimately found the two values were closely related (r = 0.3–0.8, P < 0.001) [24]. Mattioli et al. found a significant positive correlation between FF values based on MRS and those obtained using mDixon Quant (r = 0.86, P < 0.001) [25]. However, MRS is currently mainly used as a scientific tool to study osteoporosis. On the other hand, IDEAL-IQ sequences can obtain water phase, lipid phase, in-phase, out-phase, FF maps, and R2* maps with a single scan using the Larmor frequency difference between water and fat protons [26]. IDEAL-IQ sequence is a novel technique used to assess BMFF rapidly and accurately, through which we can not only diagnose osteoporosis but also forecast fracture risk [27].
In our study, FF maps were obtained in only 14 seconds, and they were automatically generated using the post-processing workstation to measure vertebral FF values. We found that FF values were significantly positively correlated with age and weakly correlated with visceral adiposity (r = 0.221, P = 0.003). In contrast, Baum et al. found that for postmenopausal women with or without type 2 diabetes, vertebral FF values measured by MRS were significantly correlated with visceral adiposity and total adiposity measured by QCT (r = 0.538 and 0.466, P < 0.05). In the diabetic group, FF was significantly correlated with visceral adiposity (r = 0.642, P < 0.05) [28]. The correlation between vertebral FF values and abdominal adiposity was higher in postmenopausal women compared to what we found in our study.
In addition, we found that vertebral FF values were significantly lower in subjects in the normal control group than in the osteopenia and osteoporosis groups. They were moderately correlated with aBMD values and highly negatively correlated with vBMD values. After controlling for gender and age factors, the correlation between FF and BMD values remained significant and was generally consistent with previous studies. Liu et al. showed that BMFF values were moderately–highly correlated with BMD values and also showed good performance in differentiating populations with different levels of bone mass [29]. Chang et al. studied the correlation between FF values measured by MRI and BMD values measured by DXA and found that FF values were highly negatively correlated with aBMD values (r = −0.93, P < 0.001), which is significantly higher than our results (r = −0.515, P < 0.001). This is probably because they included patients over 50 years of age, and the sample size was relatively small [30]. The results of ROC analysis showed that FF values were more effective in assessing low bone mass measured by QCT than by DXA, so QCT scans combined with IDEAL-IQ sequences would be superior to complete a comprehensive assessment of osteoporosis.
Later on, another parameter of IDEAL-IQ sequence, the R2* value, was gradually investigated for its application in osteoporosis. Li et al. used the FF and R2* values obtained by m-Dixon Quant to assess osteoporosis and found that FF and R2* were both negatively correlated with vBMD (r = −0.747, P < 0.001; r = −0.498, P = 0.007); both of these parameters can effectively distinguish normal and abnormal BMD values [31]. Kim et al. found a moderate negative correlation between BMD and FF values (r = −0.436, P < 0.001). R2* correlated more significantly with BMD in women than in men, especially in postmenopausal women (r = 0.644, P < 0.001). Therefore, R2* may help predict osteoporosis [32]. Measurements of R2* values were also performed in our study, but the results showed that R2* was only weakly correlated with aBMD (r = 0.192, P = 0.011) and did not correlate with vBMD values, which may be due to numerous reasons, such as variability in the included populations, inconsistency of scanning parameters, and poor resolution of R2* maps.
No studies have investigated the potential of combining DXA, QCT scans, and IDEAL-IQ sequences to diagnose osteoporosis. In previous studies, the results of the correlation between DXA and QCT measurements were unquestionable. Numerous studies have shown that QCT is more reproducible and accurate in diagnosing osteoporosis than DXA. The detection rate of osteoporosis in 313 elderly men was 10.9% for DXA and 45.1% for QCT, and Xu et al. concluded that QCT was more sensitive in identifying osteoporosis in elderly men [33]. In our study, the detection rate of osteoporosis was higher in DXA compared to QCT, with distributions of 24.7% and 21.3%, respectively, which may be due to abdominal fat accumulation, based on a review of the previous literature. Yu et al. simulated the effect of increased body fat on DXA and QCT BMD measurements and found that adding a fat layer decreased spinal aBMD values, while it increased vBMD values. They determined that abdominal fat accumulation decreased the accuracy of DXA by a factor of 1–2, while it had a smaller effect on QCT [34].
With the rapid development of artificial intelligence (AI) and machine learning, the performance of many AI algorithms has been proven to be comparable to that of expert physicians. The application of AI makes the diagnosis of osteoporosis efficient and automated and simultaneously makes osteoporosis management more standardized and integrated [35]. Zhao et al. used radiomics to predict osteoporosis using FF maps for the first time and adopted deep learning-based segmentation to detect osteoporosis and osteopenia, which was beneficial for early clinical screening of osteoporosis and prevention of fragility fractures [36]. The sensitivity of AI-based osteoporosis diagnosis was as high as 0.96, with a specificity of 0.95 [37]. However, its application in clinical practice requires more prospective studies in multiple centers.
Of course, this study has some limitations. First, the study sample was relatively small because of the small osteoporosis population. Second, the AUC analysis used low bone mass as a positive criterion instead of osteoporosis. Third, there is a significant difference in osteoporosis prevalence between men and women, but this report does not provide an analysis according to gender and age differences. Finally, this study was based on a single center only, while the FF cut-off values for diagnosing osteopenia and osteoporosis require multicenter studies.
In summary, the IDEAL-IQ sequence is a quick, easy, and non-invasive technique that accurately quantifies bone marrow fat and overcomes the shortcomings of diagnosing osteoporosis with a single BMD value. The FF values obtained from this modality are more effective in identifying low bone mass when combined with QCT than with DXA, so QCT together with IDEAL-IQ sequence is superior for performing a comprehensive assessment of bone strength. IDEAL-IQ sequences can help clinicians understand the pathophysiological changes that occur in patients with osteoporosis, and they permit early screening of patients at risk for osteoporosis. Combined with BMD, it is ideal to assess osteoporosis and reduce the incidence of fragility fractures.