This is a case-control single-center study approved by the local institutional ethics committee; written informed consent was obtained from all patients.
Study population
Our case-control study was done over a period of time of 1 year from June 2016 to June 2017 and involved 50 patients and 50 normal volunteers who underwent lumbar MRI. Inclusion criteria included post-menopausal females > 50 years who had undergone lumbar MRI for low back pain, and the reference control group of healthy females of matched age with normal body mass index (BMI = 19–25 kg/m2) to calculate the MRI-based score equivalent to the T-score used in DEXA. Exclusion criteria were (a) MRI of the lumbar spine with contrast, (b) known patients with neoplastic disease, (c) traumatic dorso-lumbar vertebral injuries, (d) time elapsed between MRI examination and DEXA longer than 6 months, (e) absolute contraindication to MR imaging (claustrophobia, implanted defibrillator or pacemaker, cochlear implant, some types of clips used on brain aneurysms), and (f) patients refusing participation in the study.
All patients were subjected to history taking (age, menstrual, medical), DEXA scan, and conventional MRI assessment of the lumbar spine. Then the calculated quantitative MRI-based score (M-score) to detect osteoporosis was done and correlated with BMD measured by DEXA.
DEXA scan
DEXA scan for the lumbar spine was done for all participants by using DEXA GE model (DPx-pro). BMD, Z-score, and T-score were calculated according to International Society for Clinical Densitometry (ISCD) guidelines (a minimum of two consecutive levels are used for lumbar T-score evaluation, after exclusion of vertebrae with focal structural abnormalities, degenerative changes, tubing artifact). The calculated values for the post-menopausal women (> 50 years) are compared with those of the healthy control group reference population values. T-scores are complicated statistical scores: T-score > 1.0 SD (normal), between − 1.0 and − 2.5 SD (osteopenia), ≤ − 2.5 SD (osteoporosis), and ≤ − 2.5 SD with 1 or more fragility fractures (severe osteoporosis).
Magnetic resonance imaging methods and image analysis
All patients underwent lumbar MRI by using 1.5 Tesla MR scanner (Philips Medical Systems, Achieva) and post-processing was done by using a dedicated Philips workstation (Extended workspace, version 2.6).
Patient’s preparation
The ferromagnetic materials were taken off. The procedure was explained to the patient, and then the patient was asked to lie supine and instructed not to move during the study.
MRI protocol and technique
Routine lumbar spine assessment from L1–L4 was performed usually by obtaining sagittal T1W spin-echo sequence (TR = 7, TE = 400–600, slice thickness = 4 mm, fov = 280 mm, matrix = 320 × 320), the most useful sequence for the assessment of bone marrow. Region of interest (ROI) was manually placed as a circle in the vertebral body excluding the cortical bone, subchondral abnormalities, focal lesions (e.g., hemangioma), and posterior venous plexus. Three ROIs were used for each vertebra, and each of them acquired on a different slice (with their mean used for analysis; that signal-to-noise ratio is not equivalent within vertebrae). Another ROI was placed in an artifact-free site outside the patient to measure the noise.
MRI image analysis
All MR images were reviewed by a consensus of two experienced radiologists blinded to the clinical information and DEXA scan results. Images were transferred to an offline workstation for post-processing. Signal-to-noise ratio (SNR) was obtained by dividing the intra-vertebral intensity by the standard deviation of the noise.
The diagnostic performance of SNRL1–L4 was estimated for each patient and used to obtain M-score (MRI-based score) for the diagnosis of osteoporosis. The SNRL1–L4 of the control group was also used in the equation by using their mean (SNRref) and standard deviation (SDref) [10].
The M-score was defined according to the formula as follows:
\( \mathrm{M}-\mathrm{score}=\frac{{\mathrm{SNR}}_{\mathrm{L}1-\mathrm{L}4\kern0.75em }-{\mathrm{SNR}}_{\mathrm{ref}}}{{\mathrm{SD}}_{\mathrm{ref}}} \)
Statistical analysis
Statistical analysis and tests were specified according to the variable type. A commercially available PC-based software package (SPSS) was used.
Results were compared with the final diagnosis and the predictive values of MRI sequences were calculated by obtaining positive predictive values (PPV), negative predictive values (NPV), sensitivity, specificity, and total accuracy in the detection of osteoporosis.
Receiver operating curve (ROC) analysis was used to define the cutoff values of M-score and SNRL1–L4 for the discrimination of osteoporotic from non-osteoporotic females.
Chi-squared test was used for analyzing differences between the results of MRI (calculated M-score) and other imaging modalities (DEXA scan). The sensitivity and specificity of MRI and other imaging modalities were calculated and compared.