A modified quantitative MRI analysis method to identify the direction of shift in lumbar disc hydration over the axial MRI in response to side-bending

Abstract

BACKGROUND: The direction of loading to disc tissue in response to postural changes has been investigated. However, the reported original method may not adequately represent the disc hydration profile.

OBJECTIVE: To investigate the reliability of a modified method to measure the direction of shift of disc hydration in response to side-bending.

METHODS: Each lumbar disc of 10 healthy subjects was scanned with T2-weighted magnetic resonance imaging in supine (neutral) position and left side-bending position in lying. A mid-point over the pixel intensity profile line was obtained. The differences in mid-point values between the neutral and side-bending positions were plotted in each direction on a X–Y coordinates plane and the angle (θ) of each plotted point from the X-axis was calculated.

RESULTS: The Intraclass Correlation Coefficients (ICC) of the mid-point values during left side-bending along the sagittal and coronal directions of the disc ranged between 0.72–0.98 and 0.80–0.96, respectively. It was estimated that calculating θ values twice yielded an ICC of 0.8 at each disc level.

CONCLUSIONS: The modified method was reliable in capturing gross disc hydration profiles. The direction of shift in disc hydration in response to side-bending can be identified using averaged values of θ that was calculated twice.

Keywords

Intervertebral disc magnetic resonance imaging reliability and validity

1. Introduction

The intervertebral disc can be a common source of low back pain [1]. The direction of loading to disc tissue in response to postural changes in the lumbar spine has been investigated in many studies [2–6] and it can be analyzed in vivo via disc hydration profile using Magnetic Resonance Imaging (MRI) [7–12]. Particularly, understanding the direction of loading on the axial plane would provide more information than that on the sagittal or coronal plane to consider the kinematical mechanism of discogenic pain.

Previously, Fazey and colleagues [7,12] reported the direction of shift in disc hydration in response to lumbar rotation by analyzing disc hydration over three 1-pixel wide lines on the axial MRI through the center of the disc over the mid-sagittal and coronal directions (i.e. original method). Lumbar rotation would not shift disc hydration considerably in the sagittal and coronal directions. Hence, this original method could still capture a good representation of the disc hydration profile and enable us to understand the direction of shift in disc hydration over the axial plane by combining shifts in disc hydration profiles over the sagittal and coronal directions. However, the original method would not be able to provide an adequate representation of the disc hydration profile in the sagittal and coronal directions when the disc hydration deviates away from the area covered by the three 1-pixel wide lines in the sagittal and coronal directions of the disc such as side-bending [8]. Thus, the application of the original method is limited.

Hence, it is necessary to modify the original method by obtaining additional data over the antero- and postero-lateral regions of the disc in order to understand the direction of shift in disc hydration over the axial plane in response to various postural changes in the lumbar spine. Such understanding requires many steps to analyze MRI data by the assessor. Therefore, reliability assessments are necessary. The purpose of this study was to investigate reliability and the number of times of analysis in identifying the direction of shift in disc hydration in response to side-bending using a modified method.

2. Methods

2.1. Participants

Eleven asymptomatic subjects without history of low back pain and contraindications for a MRI scan were recruited from a local university. Exclusion criteria included degenerated discs (Grades III to V [13]) found in generic T1- and T2-weighted sagittal and axial MRI. One subject was excluded because of moderate degeneration of the L4–L5 disc. Consequently, 10 subjects (five males), with a mean age of 25.0 years (range, 20–34 years), completed this study. This study was cleared by the institutional human medical ethics committee. Informed consent was completed prior to data collection.

2.2. Data acquisition method

MRI of each disc was obtained with a 0.2-T horizontally open unit (AIRIS mate, HITACHI Inc., Sapporo, Japan) in the evening (6 PM–9 PM) in all subjects. The subjects were placed in the supine position with knees extended beneath the flexible receiver surface coil (MR-JCL-72 separate type, HITACHI Inc., Sapporo, Japan). The pelvis was firmly stabilized to the measurement table with straps to prevent rotation of the pelvis. Axial T2-weighted images ( $FSE, FOV = 260$ , $TR / TE = 3120 / 120 ms$ , $Thickness = 8.0 mm$ , $Interval = 8.0 mm$ , $Scan time = 6 : 52$ ) of the central portion of the discs from L1–L2 to L5–S1 were obtained in the neutral position first. The subject was then positioned actively into maximum left side-bending. An assistant applied overpressure through the shoulders to ensure that end-range left side-bending position was maintained. Subjects’ knees were also manually held by another assistant during MRI scanning. The same assistants were involved during data collection in this study. The axial T2-weighted images were then taken in the end-range left side-bending position in lying.

2.3. Image analysis

The disc rotation of the MRI was corrected using the Image J1.6 software (National Institute of Mental Health, Bethesda, USA). Mid-lines of the disc were drawn from the left to right edges and the anterior to posterior edges, respectively (Fig. 1). Subsequently, deviations from the true horizontal and vertical lines were determined and the alignment of the disc was adjusted for the extent of rotation.

Fig. 1.

Diagram of measurement of disc rotation. $Rotation of the disc = (φ 1 + φ 2) / 2$ . (Colors are visible in the online version of the article; https://dx-doi-org.web.bisu.edu.cn/10.3233/BME-151273.)

Fig. 2.

Mid-point over a pixel intensity profile line. A – anterior; P – posterior; L – left; R – right; X – mid-point over a pixel intensity profile line. (Colors are visible in the online version of the article; https://dx-doi-org.web.bisu.edu.cn/10.3233/BME-151273.)

Two different methods of analysis (the original and modified methods) were utilized to identify a mid-point over an intensity profile line, where it divides the total pixel intensity into equal halves on either side (Fig. 2). A pixel was defined as the smallest unit of image of the MRI.

2.3.1. Original method [8,12]

A first line which is 1-pixel wide was drawn along the coronal direction through the center of the disc (Fig. 3). A pixel intensity profile line was then computed along this line. This line of data was normalized from 0% on the left edge to 100% on the right edge using Excel (Microsoft Corporation, Redmond, USA). Similarly, a second and a third 1-pixel wide line was drawn beneath and above the first line (Fig. 3) and the line of data was normalized, respectively. Subsequently, the three mid-points over the three lines were averaged and used for statistical analyses. This averaged mid-point value indicates the disc hydration along the coronal direction of the disc over the axial MRI. Similarly, the same procedure was repeated along the sagittal direction through the center of the disc and data was normalized from 0% on the posterior edge to 100% on the anterior edge using Excel (Fig. 3).

Fig. 3.

Diagram of data acquisition area in the original (left) and modified (right) methods of analysis. Enlarged each line of 1-pixel (original method) is presented below the two disc images. One box of square represents one pixel. (Colors are visible in the online version of the article; https://dx-doi-org.web.bisu.edu.cn/10.3233/BME-151273.)

2.3.2. Modified method

First, each disc image was cored using generic software (Paint, Microsoft Windows XP). A 55-pixel wide line was drawn along the coronal direction through the center of the disc using the Image J1.6 software (Fig. 3). The pixel intensity profile line was normalized from 0% on the left edge to 100% on the right edge using Excel and the mid-point was calculated. This mid-point, which indicated the gross disc hydration along the coronal direction of the disc over the axial MRI, was used for subsequent statistical analyses. Similarly, the same procedure was repeated along the sagittal direction through the center of the disc and data was normalized from 0% on the posterior edge to 100% on the anterior edge using Excel (Fig. 3).

2.4. Identifying the direction of shift in disc hydration over the axial MRI

The modified method was used to identify the direction of shift in disc hydration over the axial MRI in response to end-range left side-bending. The mid-point values over the pixel intensity profile line during side-bending were subtracted from the values in the neutral position over each of the sagittal and coronal directions of the disc. Each subtracted value was then plotted on a X–Y coordinates plane (sagittal direction; Y-axis, coronal direction; X-axis) and the angle (θ) formed by the plotted point with reference to the X-axis was calculated. Thus, for example, a θ value of $0 ° < θ < 90 °$ and $- 90 ° < θ < 0 °$ indicate a shift in disc hydration over the axial MRI towards the right-anterior direction and the right-posterior direction in response to left side-bending, respectively.

2.5. Statistics

Intraclass Correlation Coefficient (ICC) was calculated for the assessment of inter- and intra-observer reliability to identify mid-point over the pixel intensity profile line during left side-bending over the sagittal and coronal directions of the disc using the modified method. One investigator (Investigator 1) analyzed the images twice on two separate occasions to calculate ${ICC}_{(1, 1)}$ . Two investigators (Investigators 1 and 2), who were blinded to each other’s assessment, analyzed the same images to calculate ${ICC}_{(3, 1)}$ .

Pearson’s r correlation coefficients for the mid-point values during left side-bending along the sagittal and coronal directions of the disc were calculated between the two methods. The mean mid-point values during left side-bending along the sagittal and coronal directions of the disc were compared between the two methods using paired t-tests to investigate the difference of these methods in capturing disc hydration profiles.

Intra-observer reliability of θ values was calculated in ${ICC}_{(1, 1)}$ from two analyses of the Investigator 1 at each disc level. Subsequently, the minimum number of times of analysis expected to yield an ICC of 0.8 were calculated using the Spearman–Brown formula.

All statistical analyses were performed using SPSS version 19.0 (IBM Corporation, New York, USA). Statistical significance was set at $P < 0.05$ . Normal distribution of data was confirmed in all variables by Shapiro–Wilk tests.

3. Results

Table 1 demonstrates inter- and intra-observer reliability of the modified method in mid-point values during left side-bending along the sagittal and coronal directions of the disc. The values of ICC ranged between 0.72–0.98 and 0.80–0.96, respectively.

Table 1
Intra-class correlation coefficients ( ${ICC}_{(1, 1)}$ and ${ICC}_{(3, 1)}$ ) of the values of mid-point over the pixel intensity profile line during end-range left side-bending along the sagittal and coronal directions over the axial MRI of the disc with the use of the modified method of analysis

Level	Sagittal		Coronal

	${ICC}_{(1, 1)}$	${ICC}_{(3, 1)}$	${ICC}_{(1, 1)}$	${ICC}_{(3, 1)}$
L1–L2	0.91	0.78	0.96	0.80
L2–L3	0.96	0.72	0.84	0.93
L3–L4	0.93	0.74	0.84	0.88
L4–L5	0.90	0.91	0.81	0.83
L5–S1	0.98	0.96	0.92	0.91

Table 2 shows correlations between the two methods in the mid-point value during left side-bending along the sagittal and coronal directions of the disc at each disc level. Pearson’s r ranged between 0.83–0.92 and 0.91–0.93, respectively, indicating adequate correlations in detecting the mid-point values during side-bending.

Table 2

Pearson’s r correlation coefficients of the mid-point values during end-range left side-bending along the sagittal and coronal directions over the axial MRI of the disc in between the original method of analysis and the modified method of analysis

Level	Sagittal		Coronal

	r	P-value	r	P-value
L1–L2	0.83	0.03	0.92	<0.01
L2–L3	0.87	0.01	0.92	<0.01
L3–L4	0.87	0.01	0.91	<0.01
L4–L5	0.91	<0.01	0.93	<0.01
L5–S1	0.92	<0.01	0.93	<0.01

Table 3 demonstrates the mean mid-point value during left side-bending along the sagittal and coronal directions of the disc at each disc level in each method. There was a significant difference (

P < 0.05

) in each level of the disc except L5–S1 along the sagittal direction of the disc and in L4–L5 along the coronal direction of the disc.

Table 3

Means and standard deviations of the mid-point values^* during end-range left side-bending along the sagittal and coronal directions over the axial MRI of the disc in the original method of analysis and the modified methods of analysis

Level	Sagittal			Coronal

	Original $M \pm SD$	Modified $M \pm SD$	P-value^†	Original $M \pm SD$	Modified $M \pm SD$	P-value^†
L1–L2	46.4 ± 3.1	43.3 ± 2.1	<0.01	54.9 ± 3.5	54.6 ± 2.3	0.56
L2–L3	46.7 ± 3.0	43.9 ± 2.5	<0.01	56.4 ± 2.4	56.1 ± 2.2	0.43
L3–L4	48.2 ± 2.8	46.3 ± 2.2	<0.01	56.7 ± 2.7	56.2 ± 1.8	0.32
L4–L5	49.1 ± 3.6	48.1 ± 2.9	0.04	56.2 ± 1.7	55.3 ± 2.2	<0.01
L5–S1	48.5 ± 5.0	48.7 ± 4.0	0.69	53.4 ± 2.1	53.2 ± 1.8	0.56

100 indicates anterior or right edge of the disc, 50 indicates the center of the disc and 0 indicates posterior or left edge of the disc.

^†

Paired t-tests.

Table 4

Intra-class correlation coefficients ( ${ICC}_{(1, 1)}$ ) of θ values in the modified method of analysis and the minimum number of times needed to calculate θ in order to yield an ICC of 0.8 based on the Spearman–Brown formula

Level	${ICC}_{(1, 1)}$	Number of analysis
L1–L2	0.71	2
L2–L3	0.71	2
L3–L4	0.75	2
L4–L5	0.69	2
L5–S1	0.80	1

Table 4 shows intra-observer reliability of θ values and the minimum number of times of analysis expected to yield an ICC of 0.8. The ICC of the θ values ranged between 0.69–0.80 and it was estimated that calculating θ values twice yielded an ICC of 0.8 over every disc levels. The mean (standard deviation) of the averaged θ values analyzed twice were $19.0 ° \pm 34.2 °$ at the L1–L2 level, $6.5 ° \pm 15.0 °$ at the L2–L3 level, $0.1 ° \pm 12.3 °$ at the L3–L4 level, $3.1 ° \pm 24.5 °$ at the L4–L5 level and $- 20.9 ° \pm 42.7 °$ at the L5–S1 level.

4. Discussions

This study proposed a modified method of analysis to quantify the shift in disc hydration over the sagittal and coronal planes. It is noteworthy that the modified method captures the shift in disc hydration more accurately than the original method when the disc hydration deviates away from the mid-sagittal and coronal lines of the disc. Thus, the modified method enables the investigation of the direction of hydration shift over the axial plane in response to various postural changes in the lumbar spine. This study has demonstrated the reliability of identifying the direction of disc hydration shift when the θ value was calculated twice followed by averaging it.

The disc hydration profiles identified using the modified method correlated with those of the original method along the sagittal and coronal directions. This indicates a similar trend between the modified and original method. However, there were significant differences in the mean mid-point values between the modified method and the original method along the sagittal direction of the disc at most of the disc levels and along the coronal direction of the disc at L4–L5. These concurs with our hypothesis that the modified method is able to provide a better representation of the disc hydration profile as opposed to the original method when the disc hydration displaces from the center of the disc.

The modified method may enable us to understand disc kinematics in response to side-gliding of the lumbar spine, which is a common therapeutic exercise for unilateral back and leg symptoms [14]. It is interesting to investigate if there is any kinematical differences between side-bending and side-gliding.

There are three potential limitations in this study. First, MRI scanning was conducted only once in each lumbar posture at night in this study. We did not consider stability of the mid-point values and the θ values between days and within a day. Thus, the reliability of the analysis is likely to be consistent in studies with a cross-sectional design for now. Second, only healthy discs were scanned in this study. There may be a reduction in analysis reproducibility for degenerated discs due to the reduction of signals on T2-weighted images. Further research is required to extend the application of this modified method beyond the limited condition.

5. Conclusion

The modified method was reliable in capturing gross disc hydration profiles. The direction of shift in disc hydration over the axial MRI of the lumbar disc in response to side-bending can be identified using averaged values of θ which was calculated twice.

Footnotes

Acknowledgements

The authors wish to acknowledge Yoshikazu Ikemoto, MD, PhD for the inspection of the MRI data and Kozue Miura, JT; Taichi Gotou, JT; Yuta Nakai, JT; Shouta Kaneko, OT; Takeshi Iizawa, PT for their assistance.

Conflict of interest

There was no conflict of interest for any of the authors.

References

[1] Kramer

, Intervertebral Disk Diseases: Causes, Diagnosis, Treatment and Prophylaxis, 2nd edn, Georg Thieme Verlag, New York, 2008.

[2] Krag

M.H.

Seroussi

R.E.

Wilder

D.G.

and Pope

M.H.

, Internal displacement distribution from in vitro loading of human thoracic and lumbar spinal motion segments: Experimental results and theoretical predictions, Spine (Phila Pa 1976) 12(10) (1987), 1001–1007.

[3] Tsantrizos

Ito

Aebi

and Steffen

, Internal strains in healthy and degenerated lumbar intervertebral discs, Spine (Phila Pa 1976) 30(19) (2005), 2129–2137.

[4] Beattie

P.F.

Brooks

W.M.

Rothstein

J.M.

Sibbitt

W.L.

Jr. Robergs

R.A.

MacLean

et al., Effect of lordosis on the position of the nucleus pulposus in supine subjects. A study using magnetic resonance imaging, Spine (Phila Pa 1976) 19(18) (1994), 2096–2102.

[5] Fennell

A.J.

Jones

A.P.

and Hukins

D.W.

, Migration of the nucleus pulposus within the intervertebral disc during flexion and extension of the spine, Spine (Phila Pa 1976) 21(23) (1996), 2753–2757.

[6] Schnebel

B.E.

Simmons

J.W.

Chowning

and Davidson

, A digitizing technique for the study of movement of intradiscal dye in response to flexion and extension of the lumbar spine, Spine (Phila Pa 1976) 13(3) (1988), 309–312.

[7] Fazey

P.J.

Song

Monsas

Johansson

Haukalid

Price

R.I.

et al., An MRI investigation of intervertebral disc deformation in response to torsion, Clin. Biomech. (Bristol, Avon) 21(5) (2006), 538–542. doi:10.1016/j.clinbiomech.2005.12.008.

[8] Fazey

P.J.

Takasaki

and Singer

K.P.

, Nucleus pulposus deformation in response to lumbar spine lateral flexion: An in vivo MRI investigation, Eur. Spine J. 19(7) (2010), 1115–1120. doi:10.1007/s00586-010-1339-4.

[9] Brault

J.S.

Driscoll

D.M.

Laakso

L.L.

Kappler

R.E.

Allin

E.F.

and Glonek

, Quantification of lumbar intradiscal deformation during flexion and extension, by mathematical analysis of magnetic resonance imaging pixel intensity profiles, Spine (Phila Pa 1976) 22(18) (1997), 2066–2072.

10.

[10] Edmondston

S.J.

Song

Bricknell

R.V.

Davies

P.A.

Fersum

Humphries

et al., MRI evaluation of lumbar spine flexion and extension in asymptomatic individuals, Man. Ther. 5(3) (2000), 158–164. doi:10.1054/math.2000.0356.

11.

[11] Alexander

L.A.

Hancock

Agouris

Smith

F.W.

and MacSween

, The response of the nucleus pulposus of the lumbar intervertebral discs to functionally loaded positions, Spine (Phila Pa 1976) 32(14) (2007), 1508–1512. doi:10.1097/BRS.0b013e318067dccb.

12.

[12] Fazey

P.J.

Song

Price

R.I.

and Singer

K.P.

, Nucleus pulposus deformation in response to rotation at L1–2 and L4–5, Clin. Biomech. (Bristol, Avon) 28(5) (2013), 586–589. doi:10.1016/j.clinbiomech.2013.03.009.

13.

[13] Pfirrmann

C.W.

Metzdorf

Zanetti

Hodler

and Boos

, Magnetic resonance classification of lumbar intervertebral disc degeneration, Spine (Phila Pa 1976) 26(17) (2001), 1873–1878.

14.

[14] McKenzie

and May

, The Lumbar Spine: Mechanical Diagnosis and Therapy, 2nd edn, Spinal Publications New Zealand Ltd, Waikanae, 2003.