1 Introduction

Osteoporosis is a disorder of bone mineral metabolism causing alteration in bone microarchitecture and decreased skeletal mass.^1,2 This causes a reduction in bone strength and an increased risk of fracture. Primary Osteoporosis typically affects postmenopausal women and the elderly population.³ With an increasing number of older populations, osteoporosis-related morbidity and healthcare costs associated with its diagnosis and management are escalating.⁴ Fragility fractures mainly include hip and spinal fractures, which causes substantial disability, risk of further vertebral fractures, and even early mortality.⁵ Therefore, early diagnosis can help in the prevention or intervention at an initial stage of Osteoporosis to prevent morbidity associated with these fragility fractures and may give substantial benefit to this age group.⁶

Dual-energy x-ray absorptiometry (DEXA) scan is extensively used to diagnose osteoporosis due to its high precision (error of 2 %–2.5%) and less radiation exposure making it a relatively safer modality.⁷ The world health organization (WHO) has given a standard reference based on the T score to grade osteoporosis.⁸ An increased bone formation and resorption rate leading to altered bone morphology have a higher fracture risk, independent of BMD. Surprisingly, Osteoporosis remains an under-diagnosed and under-reported condition. In an advanced country like the United States alone, roughly 80% of fragility fracture patients have failed to undergo DEXA or received concomitant treatment for Osteoporosis.⁹

There are several other ways to detect Osteoporosis, like quantitative computed topography (more sensitive than the DEXA) or quantitative ultrasound.¹⁰ Nevertheless, these tests need to be ordered by a physician as the disease is mostly asymptomatic. An alternate approach is to extract data regarding BMD while patients undergo imaging for other purposes.

Magnetic resonance imaging (MRI) of the lumbar spine is an imaging study that is often done in the aging population who complain of low back pain.¹¹ With the decrease in cortical thickness and cancellous bone volume, there are changes in the bone marrow signal intensity on the standard T1-weighted images. This correlation can be explored and exploited to detect osteoporosis using the MRI. Additionally, vertebral fractures can be accurately detected by MRI. Due to the lack of a similar study in the Indian population, the present study was conducted to investigate the reliability and feasibility of MRI in diagnosing osteoporosis with DEXA as a reference.

2 Materials and methods

2.4

2.4 Outcome measures-

Signals were measured from sagittal T1WI from the body of L1, L2, L3, and L4 vertebral bodies. Region of Interest (ROI) of size 130–180 mm2 was placed in the vertebral body in the mid-sagittal section and parasagittal sections on either side (Fig. 1). ROIs were placed away from cortical bone, posterior vertebral venous plexus, subchondral abnormality, or other vertebral lesions like haemangioma, if any. For each vertebra, four ROIs were placed, and the mean values were obtained and used for analysis. The noise was obtained in each patient by putting an ROI outside the patient's body in a region devoid of any artifact. Signal to Noise Ratio (SNR) was calculated by dividing the mean signal intensity obtained for each vertebra by the standard deviation of the noise.SNR = mean signal intensity/ SD of noise

Close

Fig. 1 Sagittal T1-weighted magnetic resonance image of the lumbar spine in a 46-year-old female with four regions of interest (ROI) manually segmented from L1 to L4, plus a region of interest placed outside the patient for the measurement of the noise; B- displays report from dual-energy X-ray absorptiometry of the same patient.

Export to PPT

SNR was obtained for each lumber vertebra from L1 to L4 level, and again mean of the four lumbar vertebrae was obtained for each patient. The mean SNR was labelled as an SNRpatient. Image analysis was conducted by two radiologists (SN, AN). Both of them were blinded to the DEXA findings.

Similarly, we measured SNR from 24 control females referred for low back pain or other reasons. Patients having trauma, contusion, fracture, malignancy, or vertebral fracture were excluded. Their age ranges from 20 to 31 years, with an average body mass index (BMI) of less than 25 kg/m2. SNR from L1 to L4 of the control group was obtained similarly to those of patients, and mean SNR of L1 to L4 was obtained for each control. The mean SNR of all the control groups was calculated and labelled as SNRcontrol.

An MRI-based scoring system named M score was obtained for each patient for diagnosis of osteoporosis, similar to the T score in DEXA. M score was calculated by getting the difference in SNR patients to SNR controls and then dividing it by the control's standard deviation (SD).M score = (SNRpatient – SNRcontrol)/SDcontrol

3 Results

A total of 33 patients participated in the study. The mean age of the study participant was 56.00 ± 10.25 years, with a minimum age of 37 years and maximum age of 74 years. Out of the 33 patients, 16 (48.5%) were osteoporotic, and 17 (51.5%) were non-osteoporotic. The other demographic details are given in Table 1.

The ROC curve was plotted to illustrate the diagnostic ability of the M score in correctly diagnosing osteoporosis (Fig. 2). The area under the curve was 0.829, with a lower and upper bound of 0.678 and 0.980, respectively. The area under the curve is significant. If the M score is greater than or equal to 2.82, the sensitivity was 87.5%, and the specificity was 76.5%. With an M score cut-off of 2.92, the sensitivity reduced to 75.0%, and the specificity increased to 83.4% (Table 2).

Close

Fig. 2 Receiver operating characteristic curve for M score.

Export to PPT

3.1

3.1 Correlations of M score and T score

M scores negatively correlated with the T score. With the increase in the T score, the M score decreased. The Spearman correlation coefficient for the spine T score was −0.651, with a p-value of <0.001, and the hip T score was −0.428, with a p-value of 0.013. The correlation coefficient suggested a moderate negative correlation between the M and T spine scores, as seen in the scatter plot (Fig. 3a and b).

Close

Fig. 3 Scatter plot of M score and T score between spine (a) and hip (b).

Export to PPT

4 Discussion

DEXA scan is the best available tool in the present times to measure bone strength in terms of areal density (gm/cm2). This is the choice investigation for osteoporosis.¹² However, the irony is that DEXA examinations are performed fewer in subjects who do not have fragility fractures but are at high risk of so, and sometimes patients with fragility fractures are not evaluated and treated for osteoporosis.¹³ This can be attributed to the lack of a screening policy; the disease is mainly silent. The advantage of a DEXA is a high specificity but a low detection rate.¹² BMD measured through DEXA is now known to be a poor predictor of fracture as bone's mechanical property and quality depends on bone microarchitecture, mineral content, etc.¹⁴ Therefore, other parameters indicative of bone quality not amenable to be detected by DEXA need to be measured.

With an increase in age, bone marrow fat progressively increases, and there is a proportionate decline in bone strength, though a direct causal relationship has not been found.¹⁵ Rosen CJ et al. proposed that trabecular microarchitecture gets altered with increased bone marrow fat in osteoporosis.¹⁶ The adipocytes may replace spaces previously occupied by trabecular bone. This has also been seen in several other histomorphometry studies.^17–19 This is the core principle on which the MRI assessment is based. Shah LM found that the T1-weighted spin-echo images help evaluate the cellular content in the bone marrow. The fat content, because of its hydrophobic groups, results in a short T1 relaxation time and an increased fat content also causes a diffusely increased T1-weighted hyperintensity.¹⁷ Bandirali et al. also worked on the same principle, and we also followed in their footsteps.⁵

Alternatively, Chang et al. studied the fat fraction of the vertebral body and derived a score that negatively correlated with osteoporosis.²⁰ Studies have suggested that T1WI had high diagnostic capabilities in osteoporosis as T1WI can better reflect the signal from fat in the bone marrow.^5,20–22 In a study by Koyama H et al., they concluded that routine MRI, especially SNR of T1WI has a potential for the assessment of osteoporosis and can be a reliable diagnostic indicator.²³ Other sequences like MR spectroscopy have been studied to detect the fat component of the vertebra in osteoporosis, however this requires a skilled operator, time consuming and reflects the fat content of part of a single vertebra.²⁴ Bandirali developed a scoring system based on this T1-hyperintensity and proposed that this is a better predictor of fracture risk.⁵

MRI of the lumbar spine is more routinely performed for elderly patients who are investigated for low back pain.¹¹ Hence, this opportunity can also provide a screening tool for osteoporosis. We had an 87% sensitivity in comparison to DEXA as a reference standard. Several other researchers have found a similarly high sensitivity.^5,10,25,26 This may eliminate the use of another investigating tool (DEXA) for diagnosis and can be helpful in follow-up patients who need MRI for some ailment and thus reduce the need for radiation. In the setting of a surgical indication, an M score may give an insight into the bone quality before instrumentation so that the surgeon can gear up accordingly.

The cut-off value we found was an M score of 2.82, which had 77% specificity and 87% sensitivity. Other researchers, such as Azin et al., found a cut-off of 2.05 with a sensitivity of 90% and specificity of 87% in detecting osteoporosis while Saad et al. found a cut-off of 3.5 with a sensitivity of 93.3% and specificity of 83.5%.^21,22 Bandirali et al. found a higher threshold M score value of 5.5 for a specificity of 90%.⁵ This is much higher than what we or Azin et al. or Saad et al. reported. The authors themselves admit that M-score values are valid for the same single MRI machine they are built on. Azin et al. also reported that besides the machine, it is also influenced by set-up, exterior and the harmony of magnetic field.²¹ Therefore, each system in each center needs separate calibrations. i.e., build its thresholds based on different SNR values provided by each MR system for proper adaption by the clinicians. Bandirali et al. also found its higher predictive value in the degenerative spine, obese patients, and fragility fractures.⁵ We did not do a subgroup analysis of our cohort. Schellinger D et al. found that among the lumbar vertebrae, L3 vertebra was least affected by degenerative changes which is commonly found in the other lumbar vertebra.¹³

This study has limitations. Firstly, our cohort group was small, and studies in more comprehensive and multicentric centers with similar machines (MRI and DEXA) should be undertaken. We have not differentiated osteopenia in our group but have taken a dichotomous variable of osteoporosis (yes or no). Only T1W images were used. T2W and STIR, or chemical shift sequence, are not utilized in our study that can provide further information. Future studies can be directed with the multiparameter study. The females in our reference group were young and presumed to be healthy and therefore did not undergo DEXA study. Regardless of these shortcomings, this study does find MRI as a useful test in detecting osteoporosis especially in the high-risk patients.

5 Conclusion

Our study indicates that MRI investigations are helpful in Osteoporosis assessments. Even though MRI may not replace DEXA, the gold standard for osteoporosis diagnosis, it can give insight into elderly patients who get an MRI routinely for back pain. It may also have a prognostic value.

Funding/sponsorship

This research did not receive any specific grant from funding agencies in the public, commercial or not-for-profit sectors.

Institutional ethical committee approval

T/IM-NF/Ortho/20/111.

Authors contribution

Mantu Jain, Conceptualization, methodology, writing the original draft.

Suprava Naik, Conceptualization, methodology, writing the original draft, project administration.

Narayan Prasad Mishra, Methodology, Resources.

Sujit K Tripathy, Conceptualization, writing (review and editing), Resources.

Aishwarya Neha, Methodology, writing (review and editing).

Dinesh Prasad Sahu Data analysis.

Lubaib KP, methodology, Resources, writing (review and editing).