Whole spine localizers of magnetic resonance imaging detect unexpected vertebral fractures

Abstract

Background

Whole spine localizers (WS-loc) of magnetic resonance imaging (MRI) are performed for enumeration of the vertebrae but they can be also used for the evaluation of the spine.

Purpose

To assess the accuracy of fracture detection using WS-locs of MRI and compare the findings with standard high-resolution short tau inversion recovery (STIR) sequences, and to determine whether the review of WS-locs is useful and if additional information can be gained by assessing the thoracic spine section of the WS-locs.

Material and Methods

A total of 298 magnetic resonance (MR) examinations of the lumbar spine with WS-locs were evaluated. Two independent readers reviewed the images. In case of fracture detection, further characterization of the fracture was performed. To assess inter-reader agreement, unweighted Cohen’s kappa with 95% confidence intervals (CI) and Phi coefficients were calculated.

Results

The study sample included 187 female and 111 male patients (age range = 65–94 years; median age = 75.0 years). The WS-locs detected 42 fractures of the lumbar spine and 36 of the thoracic spine. Inter-reader agreement for fracture detection in the lumbar and thoracic spine was strong (K = 0.87, 95% CI = 0.78–0.95, Phi = 0.87, and K = 0.88, 95% CI = 0.79–0.96, Phi = 0.88, respectively).

Conclusion

WS-locs from MR examinations of the lumbar spine provide a good diagnostic tool for the detection and evaluation of unsuspected vertebral fractures. WS-locs show strong inter-reader agreement for fracture detection in the thoracic and lumbar spine.

Keywords

Spinal fracture magnetic resonance imaging MRI aged osteoporotic fractures/diagnosis

Introduction

Despite the rapid development of modern radiological techniques and evolving imaging possibilities, the diagnosis and therapy of vertebral fractures remain challenging (1). Up to 40% of symptomatic vertebral fractures are not diagnosed (2). Vertebral fractures are the most frequent of all osteoporotic fractures; their incidence in postmenopausal women is approximately 20%. Especially in elderly patients, the clinical manifestations of vertebral fractures may be obscured by additional co-morbidities. When postmenopausal women sustain one vertebral fracture, the risk for further vertebral fractures increases fivefold (3). However, it is unclear whether the risk for further fractures is also increased in patient groups other than postmenopausal women.

In patients aged > 70 years with low back pain (LBP), radiographs are initially performed, even if the radiographic detection of the vertebral fractures is rather poor (4). To exclude some rather rare reasons for LBP such as infection, tumor, or metastases, magnetic resonance imaging (MRI) can be additionally obtained (5). MRI may also help in fracture characterization (sub/acute versus old) and in the detection of unexpected fractures. Due to an increased risk of additional fractures (hip or other spine segments), associated medical and economic impact, and subsequently impaired function and disability, the relevance of unexpected vertebral fractures has been emphasized (5 –9).

Recent studies have assessed the reliability of localizers from computed tomography (CT) and MRI examinations for the recognition and evaluation of vertebral fractures (10 –13). MRI localizers of the lumbar spine (LS) were used to detect vertebral fractures and showed excellent inter- and intra-observer agreement (14). Due to the recent development of coils used in MRI, nowadays the whole spine localizers (WS-loc) are performed at the beginning of the MRI of the lumbar spine for enumeration of the vertebrae (15).

The aims of this study were as follows: to assess the accuracy of fracture detection of the lumbar spine using WS-locs of MRI and to compare the findings with standard high-resolution short tau inversion recovery (STIR) sequences; and to determine whether the review of WS-locs is useful for detection of additional fractures in thoracic and cervical spine

Material and Methods

Study design

This study was approved by the Ethical Review Committee (EKNZ Project-ID: 2016-02141).

A total of 298 MRI consecutive examinations of the lumbar spine with WS-locs performed between January and June 2017 were selected from the local Picture Archiving and Communication System (PACS) in the radiology digital archive. Patients aged >65 years who had an MRI of the lumbar spine with WS-locs were included in the study. The exclusion criteria were: (i) missing WS-locs; (ii) missing information on the referral reason; and (iii) metal hardware resulting in non-evaluable image quality. To investigate a representative patient population, the study was performed on all consecutive patients aged >65 years and not only >70 years as previously described (4).

Image analysis

Two independent blinded readers with experience in spine imaging (reader 1: radiologist (MK) with two years of experience; reader 2: orthopedic surgeon (JMB) with four years of experience) evaluated the WS-locs. Lumbar, thoracic, and cervical spine segments of the sagittal and coronal WS-loc were reviewed for fractures. In case of disagreement, a senior radiologist was consulted (FA).

After assessment of the WS-locs, comparison with the lumbar spine STIR sequences was performed. STIR sequences were used as the internal reference standard sequences for fracture detection in the lumbar spine. The fractures in the thoracic and cervical spine have been evaluated on the WS-locs; there was no additional sequence of the thoracic or cervical spine performed.

The reading sessions were done using GE Centricity RA 1000 PACS software, Version 4.0.11 (GE Healthcare, Barrington, IL, USA). Image quality was separately evaluated by both readers as 1 (poor), 2 (medium), or 3 (good).

Fracture characterization

When one or more fractures were detected, both readers performed further characterization of the most severe fracture, separately for each segment of the spine. The following criteria were analyzed: (i) height loss (>50% or <50%); and (ii) spinal canal stenosis.

MR examination

MR examinations were performed with 1.5-T and 3-T systems (Avanto Fit/Aera and Skyra, respectively; Siemens Medical Solutions, Erlangen, Germany) using integrated body coils with patients in the supine position. MR examinations of the lumbar spine that included WS-locs in the sagittal and coronal planes were used for fracture detection. Sagittal STIR sequences of the lumbar spine were used as the internal reference standard method to verify whether there was a fracture in the lumbar spine that correlated to the previously detected finding on the WS-locs. Detailed MR parameters of the WS-locs and STIR sequences are shown in Table 1.

Table 1.

MRI parameters.

	TR/TE (ms)	Section thickness (mm)	FOV (mm)	Matrix size	Acquisitions (n)
WS-localizers (3×)	7/2.44	5.5	400 × 400	179 × 256	2
STIR sagittal	3560/31	6	420 × 341	195 × 320	2

TR/TE, repetition time/echo time; FOV, field of view; STIR, short tau inversion recovery; 3×, three blocks.

Statistical methods

Nominal variables are presented as numbers with appropriate percentages, while continuous variables are presented as medians (Me) with interquartile ranges (IQR). To evaluate inter-reader agreement, unweighted Cohen’s kappa with 95% confidence intervals (CI) and Phi coefficients were calculated and interpreted as previously described (16 –18). STATISTICA 13.1 (Statsoft, Tulsa, OK, USA) was used for statistical analyses (19).

Results

In total, 298 patients (187 women, 111 men) were included in the study (age range = 65–94 years; median age = 75.0 years). The flow chart presenting patient selection is shown in Fig. 1.

The most common referral reason for performing MRI of the LS was lumbar pain (152/187 and 87/111 in female and male patients, respectively). The second most common referral reason for MRI of the lumbar spine was clinically suspected or known vertebral fracture (11/187 and 5/111 in female and male patients, respectively). Detailed patient characteristics of all patients are found in Table 2.

Table 2.

Detailed patient characteristics of all patients (n = 298).

Gender (female vs. male)	n (%)
Gender (female vs. male)	187 (62.75) vs. 111 (37.12)
Age (years)	Median (IQR)	Average (±)	Range
All patients	75.0 (71.0–81.0)	76.08 (±6.52)	65.00–94.00
Female	75.0 (71.0–80.0)	75.63 (±6.27)	65.00–94.00
Male	77.0 (70.0–83.0)	75.84 (±6.89)	66.00–90.00

Lumbar fractures

Reader 1 detected 42 fractures in the WS-locs; however, four of them turned out to be false positives in sagittal STIR sequences (sensitivity = 100%, specificity = 98%). Reader 2 detected 45 fractures of the LS in WS-locs, but two of them were false positives. Moreover, there were two false-negative findings in WS-locs identified by reader 2. Inter-reader agreement for the detection of fractures of the LS was strong (K = 0.87, 95% CI = 0.78–0.95, Phi = 0.87). Detailed fracture characteristics such as affected vertebral body, height loss, and spinal canal stenosis and inter-reader agreement are presented in Table 3. L1 was the most commonly affected vertebral body (25/42 of all LS fractures).

Table 3.

Detailed fracture characteristics for lumbar spine (LS) such as affected vertebral body, height loss, and spinal canal stenosis with inter-reader agreement.

Reader 1, LS	N/N total	Inter-reader agreement
Fractures (n)		K = 0.87, 95% CI = 0.78–0.95, Phi = 0.87
1	25/42
2	8/42
3	7/42
5	2/42
Affected vertebral body		K = 0.96, 95% CI = 0.88–1, Phi = 1.89
1	18/42
2	8/42
3	4/42
4	7/42
5	4/42
50%		K = 0.68, 95% CI = 0.42–0.94, Phi = 0.72
1	33/42	K = 0.68, 95% CI = 0.42–0.94, Phi = 0.72
2	9/42
Spinal canal stenosis		K = 0.91, 95% CI = 0.73–1, Phi = 0.91
0	36/42	K = 0.91, 95% CI = 0.73–1, Phi = 0.91
1	6/42

Thoracic fractures

Reader 1 detected 36 fractures in the sagittal WS-locs; 33 of these were also visible in the coronal WS-locs. Reader 2 identified 33 fractures of the TS in sagittal WS-locs, with 32 of them evident in the coronal WS-locs. Inter-reader agreement for fracture detection was strong (0.88, 95% CI = 0.79–0.96, Phi = 0.88). Detailed fracture characteristics such as affected vertebral body, height loss, and spinal canal stenosis with inter-reader agreement are presented in Table 4. Th12 was the most commonly affected vertebral body (15/36 of all identified TS fractures).

Table 4.

Detailed fracture characteristics for thoracic spine (TS) such as affected vertebral body, height loss, and spinal canal stenosis with inter-reader agreement.

Reader 1, TS	Total (n/n)	Inter-reader agreement
Fractures (n)		K = 0.88, 95% CI = 0.79–0.96, Phi = 0.88
1	26/36
2	6/36
3	3/36
4	1/36
Affected vertebral body		K = 0.81, 95% CI = 0.63–0.98, Phi = 2.2
3	2/36
5	1/36
7	1/36
8	2/36
9	5/36
10	2/36
11	8/36
12	15/36
50%		K = 0.80 (95% CI = 0.58–1, Phi = 0.81
1	24/36
2	12/36
Spinal canal stenosis		K = 0.66 (95% CI = 0.30–1, Phi = 0.67
0	29/36
1	7/36

In 17 of patients who were diagnosed with a fracture in the LS, an additional fracture in the TS was identified. Moreover, in 22 out of 216 patients who did not have a fracture of the LS, at least one fracture of the TS was found. Examples of the MRI WS-loc with fractures of the LS and TS are presented in Figs. 2 –4.

Cervical fractures

There were no fractures or any pathologic signal abnormalities identified in the cervical spine.

Image quality

Image quality was good in 69.0%, medium in 25.3%, and poor in 5.7% of all cases.

Additional information

In all analyzed cases (n = 281), all findings of the LS and TS were mentioned in the final radiology report and there was no need to call the referring physician. A total of 245 patients underwent the MR examination via 1.5-T and 36 patients via 3-T scanners.

Discussion

This study provides a comprehensive evaluation using the WS-locs of MRI. The results suggest that WS-locs of MRI can be used for fracture detection and these outcomes are consistent with previous studies that focused on only one section of the spine, rather than the whole spine (10, 12, 14). Our study shows that WS-locs of MRI provide a reliable tool for the detection of vertebral fractures in the thoracic and lumbar spines. Despite a complete depiction of the cervical spine offered by the WS-locs, there were no fractures or pathologic findings observed in this region.

WS-locs of MRI depict the entire spine in both sagittal and coronal planes. Once a fracture is detected in any of the planes, it is possible to scroll for a correlative finding in both planes. As this study demonstrates, the agreement between sagittal and coronal WS-locs for fracture detection in the thoracic spine is very high (K= 0.95, 95% CI = 0.9–1). The possibility of scrolling between images, both in the sagittal and coronal WS-locs allows for even higher confidence during the evaluation of vertebral fractures.

WS-locs of MRI are obtained as three different blocks of the cervical, thoracic, and lumbar spine and then automatically reconstructed into the WS-locs. They have been described as a reliable and time-efficient tool for enumeration of the vertebrae; their implication in spine MR protocols has been primarily suggested for this reason (15).

In this study, parallel imaging was used to acquire the WS-locs in the sagittal and coronal planes, and the acquisition time was 86 s (approximately 28 s for each part of the spine). Because this method is performed using reduced volume of k-space data, the acceleration of acquisition time by factors of 1.5–3 is provided (20,21). The increased acquisition speed is the main advantage of this technique; however, its application is limited by artefacts and reduced image quality, such as signal-to-noise ratio (20, 21).

Our study shows that parallel imaging can be applied for acquisition of the WS-locs and it provides reliable fracture detection in the thoracic and lumbar spine. However, this approach may not be applicable in patients with diffuse osseous metastases with hyperintense signal in the whole spine, thus impeding detection of acute fractures.

The study of Xu et al. found that vertebral body height loss is an important indicator for the presence of vertebral fractures (22). For this reason, we sought to characterize the height loss of detected fractures in the WS-locs and applied such criteria as height loss of < 50% or > 50%. The inter-reader agreement was moderate (K = 0.68, 95% CI = 0.42–0.94) and strong (K = 0.80, 95% CI = 0.58–1) for characterization of the height loss in the LS and TS, respectively. However, malignant metastatic osseous disease may lead to kyphotic deformity and spinal malalignment, so in this case the loss of vertebral height cannot be used as a fracture indicator in the WS-locs. It must be also stated that the differentiation between benign and malignant vertebral fracture is not possible using the WS-locs.

To investigate a representative patient population, the study was performed on all consecutive patients aged > 65 years, without any additional information on patient medical history and co-morbidities (such as osteoporosis, metabolic bone disease, or medication). Reviewing the WS-locs, we found 42 and 36 fractures of the LS and TS, respectively. Sagittal STIR sequences of the lumbar spine were used as a reference standard method to verify whether the fracture of the LS visible in the WS-locs was truly positive. Only four of these turned out to be false positives in sagittal STIR sequences, resulting in 100% sensitivity and 98% specificity of the WS-locs. Moreover, 36 fractures of the thoracic spine were detected in the sagittal WS-locs, 33 of which were also visible in the coronal WS-loc. They were correctly described in the final report; however, they may have been considered unexpected findings, as the primary imaging focus was the lumbar spine.

As the study shows, the WS-locs may reveal additional fractures of the thoracic spine; these findings may require further imaging and focused clinical examination. The study of Kim et al. found that if there is a fracture involving > 2 vertebral bodies, MRI is required to make the diagnosis, because the possibility of confirming the hot uptake lesion as a new fracture by bone scanning is low (23). In this situation, the WS-locs can help in exact planning of the MRI exam with focus on the affected areas, without the need for the whole spine imaging.

The coronal WS-locs were used only for evaluation of the spine hence incidental findings such as renal and liver cysts were reported separately but not included in the study.

Our findings suggest that time efficient WS-locs are a helpful tool for the detection of unexpected vertebral fractures, including in the spine segments other than those initially suspected.

Our study has several limitations. First, the included patient group was relatively heterogeneous, with patients aged 65–94 years. Decreased bone mineralization and loss of muscle mass and function in elderly patients, known as sarcopenia, are potential risk factors for vertebral fractures, especially in women (24,25). However, we were not able to validate this hypothesis as there was no control group of younger patients. Second, all MR examinations were performed without specific information on bone diseases, for example, osteoporosis. Finally, we used sagittal STIR sequences of the lumbar spine to verify whether the detected fractures of LS were true positives. There was no additional sequence of the thoracic spine performed to verify whether the suspected findings were truly positive. The MR examination was performed primarily due to lumbar pain and the WS-locs were a usual part of the MR study. Detected fractures of the thoracic spine were reported separately and were subsequently followed by the referring clinicians. The addition of verifying sequences of the thoracic and lumbar spine in future studies with WS-locs of MRI may be valuable and could potentially provide supplementary results.

In conclusion, time-efficient WS-locs from MR examinations of the lumbar spine provide information for the evaluation of vertebral fractures and can be used not only for enumeration of the vertebrae, but also for the detection of unsuspected vertebral fractures. These fractures may need further imaging and should be reviewed separately. The WS-locs of MRI show a strong inter-reader agreement for fracture detection of the thoracic and lumbar spine and can be helpful for both radiologists and orthopedic surgeons.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

References

Mitchell

Jewell

Javaid

et al.

Reporting of vertebral fragility fractures: can radiologists help reduce the number of hip fractures?

Arch Osteoporos 2017; 12:1–6.

Ross

Clinical consequences of vertebral fractures. Am J Med 1997; 103:30–43.

Delmas

van de Langerijt

Watts

et al . Underdiagnosis of vertebral fractures is a worldwide problem: the IMPACT study. J Bone Miner Res 2005; 20:557–563.

Terakado

Orita

Inage

et al . A clinical prospective observational cohort study on the prevalence and primary diagnostic accuracy of occult vertebral fractures in aged women with acute lower back pain using magnetic resonance imaging. Pain Res Manag 2017;2017:1–5.

Roudsari

Jarvik

JG.

Lumbar spine MRI for low back pain: indications and yield.

Am J Roentgenol 2010; 195:550–559.

Arima

Abe

Nishimura

et al . Association of vertebral compression fractures with physical performance measures among community-dwelling Japanese women aged 40 years and older. BMC Musculoskelet Disord 2017; 18:1–8.

Gehlbach

Bigelow

Heimisdottir

et al . Recognition of vertebral fracture in a clinical setting. Osteoporos Int 2000; 11:577–582.

Klotzbuecher

Ross

Landsman

et al . Patients with prior fractures have an increased risk of future fractures: a summary of the literature and statistical synthesis. J Bone Miner Res 2000; 15:721–739.

Lindsay

Silverman

Cooper

et al . Risk of new vertebral fracture in the year following a fracture. JAMA 2001; 285:320–323.

10.

Samelson

Christiansen

Demissie

et al . Reliability of vertebral fracture assessment using multidetector CT lateral scout views: the Framingham Osteoporosis Study. Osteoporos Int 2011; 22:1123–1131.

11.

Takada

Lang

et al . Vertebral fracture assessment using the lateral scoutview of computed tomography in comparison with radiographs. Osteoporos Int 1998; 8:197–203.

12.

Bazzocchi

Spinnato

Albisinni

et al . A careful evaluation of scout CT lateral radiograph may prevent unreported vertebral fractures. Eur J Radiol 2012; 81:2353–2357.

13.

Bazzocchi

Spinnato

Garzillo

et al . Detection of incidental vertebral fractures in breast imaging: the potential role of MR localisers. Eur Radiol 2012; 22:2617–2623.

14.

Bazzocchi

Garzillo

Fuzzi

et al . Localizer sequences of magnetic resonance imaging accurately identify osteoporotic vertebral fractures. Bone 2014; 61:158–163.

15.

Weiss

Storrs

Banto

Automated Spine Survey Iterative Scan Technique.

Radiology 2006; 239:255–262.

16.

McHugh

Interrater reliability: the kappa statistic.

Biochem Med (Zagreb) 2012; 22:276–282.

17.

Sinacore

Connell

Olthoff

et al . A method for measuring interrater agreement on checklists. Eval Health Prof 1999; 22:221–234.

18.

Zegers

Coefficients for interrater agreement.

Appl Psychol Meas 1991; 15:321–333.

19.

Statsoft. STATISTICA. 13.1 ed. Palo Alto, CA: TIBCO Statistica, 2016. Available at: https://www.tibco.com/products/tibco-statistica.

20.

Deshmane

Gulani

Griswold

et al . Parallel MR imaging. J Magn Reson Imaging 2012; 36:55–72.

21.

Glockner

Stanley

et al . Parallel MR imaging: a user’s guide. Radiographics 2005; 25:1279–1297.

22.

Perera

Medich

et al . Height loss, vertebral fractures, and the misclassification of osteoporosis. Bone 2011; 48:307–311.

23.

Kim

Jang

et al . The comparison of bone scan and MRI in osteoporotic compression fractures. Asian Spine J 2010; 4:89–95.

24.

Cruz-Jentoft

Baeyens

Bauer

et al . Sarcopenia: European consensus on definition and diagnosis: Report of the European Working Group on Sarcopenia in Older People. Age Ageing 2010; 39:412–423.

25.

Hida

Shimokata

Sakai

et al . Sarcopenia and sarcopenic leg as potential risk factors for acute osteoporotic vertebral fracture among older women. Eur Spine J 2016; 25:3424–3431.