- Email: [email protected]
Parametric characterization of spinal motions in osteoporotic vertebral fracture at level T12 with fluoroscopy
Available online at www.sciencedirect.com
Medical Engineering & Physics 31 (2009) 346–355
Parametric characterization of spinal motions in osteoporotic vertebral fracture at level T12 with fluoroscopy Shing Chun Benny Lam a , Robert Allen a,∗ , Gill Pearson b , Cyrus Cooper c,d b
a Institute of Sound and Vibration Research, University of Southampton, Southampton, UK Department of Medical Physics and Bioengineering, Southampton University Hospital NHS Trust, Southampton, UK c MRC Epidemiology Resource Centre, Southampton University Hospital NHS Trust, Southampton, UK d Botnar Research Institute, University of Oxford, Oxford, UK
Received 21 January 2008; received in revised form 24 May 2008; accepted 1 June 2008
Abstract Vertebral fractures due to osteoporosis are a common skeletal disorder affecting the mobility of the patients, although little is known about the relationship between spinal kinematics and osteoporotic fracture. The purpose of this study was to characterize the motions of the thoracolumbar spine affected by osteoporotic vertebral fracture at level T12 and compare the results with those of non-fracture osteoporosis subjects. We examined the continuous segmental kinematics of the vertebrae, and describe the segmental motion of the spine when a fracture at T12 is present. Fluoroscopy sequences of the thoracolumbar spines during sagittal and lateral flexion were collected from 16 subjects with osteoporosis of their spine (6 with vertebral fractures at T12, 10 without a fracture). Vertebrae T10–L2 in each frame of the sequences were landmarked. Kinematic parameters were calculated based on the landmarks and motion graphs were constructed. Compared to the control subjects who did not have a fracture, fracture subjects had a more asymmetric lateral range of motion (RoM) and required a longer time to complete certain phases of the motion cycle which are parameterized as lateral flexion ratio and percentage of motion cycle, respectively. Prolonged deflection was more frequently found from the fracture group. Characterizing the motions of the fractured vertebra together with its neighboring vertebrae with these kinematic parameters is useful in quantifying the dysfunction and may be a valuable aid to tracking progress of treatment. © 2008 IPEM. Published by Elsevier Ltd. All rights reserved. Keywords: Spine kinematics; Fluoroscopy; Osteoporosis; Vertebral fracture; Motion analysis
1. Introduction Osteoporosis is a common skeletal disorder characterized by compromised bone strength predisposing the sufferer to an increased risk of fracture [1]. Patients with osteoporotic spines can suffer from reduced mobility due to fractures. Reoccurrence rate leading to multiple fractures is high and the need of hospitalization is significant [2]. Furthermore, we have shown in [3] that osteoporotic fractures in elderly people were associated with their daily life functions. Previous studies also reported a significant correlation between quality of ∗ Corresponding author at: Signal Processing and Control Group, Institute of Sound and Vibration Research, University of Southampton, Southampton SO17 1BJ, UK. Tel.: +44 2380 593082. E-mail address: [email protected] (R. Allen).
life, number of vertebral fractures and lumbar lordosis angle, as well as spinal range of motion (RoM) [4–7]. However, none of these studies addressed the relationship between the motions of the spine and osteoporotic fracture. Stable spinal condition relies upon the idea that the anatomical and physiological conditions of every component of the functional spinal unit are normal and can coordinate effectively with each other [8]. Osteoporosis reduces the bone density, often leading to osteoporotic fracture which alters the normal shape of the vertebral body. Due to the structural changes of the vertebral body, the motions of the spinal column would be disturbed. Although the direct causeand-effect relationships between abnormal vertebral motions and spinal disorders have not been fully understood, previous studies had reported that spinal disorders such as low back pain, spondylolisthesis and disc degeneration could be
1350-4533/$ – see front matter © 2008 IPEM. Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.medengphy.2008.06.004
S.C.B. Lam et al. / Medical Engineering & Physics 31 (2009) 346–355
revealed in abnormal spinal motions, including increased mobility in one particular direction, abnormal spinal kinematics and altered instantaneous axes of rotation [8–11]. Fluoroscopy is an X-ray-based imaging modality employed in this study. As mentioned in previous studies, actual motions of each vertebra can be found from the fluoroscopy sequences [8,12–17]. They investigated the spine in different conditions by monitoring the changes in motions of the vertebrae during functional motions in different setups. Moreover, fluoroscopy allows a direct and objective assessment of the kinematics with a robust landmarking protocol [18]. More importantly, fluoroscopy has the advantage of involving relatively low radiation exposures. These features of fluoroscopy enable us to study the continuous motion of the vertebrae and no other devices can currently offer such an insight. The purpose of this study is to characterize the spine motions captured using fluoroscopy in vivo and verify the usefulness of the kinematic parameters used to characterize and differentiate subjects with an osteoporotic fracture at level T12 from controls. The technique would be used as a basis for assessing functional recovery with treatment and for assessing the patient at different times. The experimental setup and methods to extract and analyze the spinal motions are reported.
2. Method 2.1. Subjects The study was reviewed and approved by the National Health Services Local Research Ethics Committee and registered with the Southampton University Hospital Trust Research & Development Central Administrative Office. Advice from the Radiation Protection Adviser was sought. Two groups of female subjects with osteoporosis were recruited: a control group and a fracture group. These women were selected from subjects who had been routinely scanned for assessment of osteoporosis, and who had vertebral bone mineral density values of at least 2.5 standard deviations (S.D.) below the young adult mean (i.e. T-score at, or below, −2.5) using the Hologic Discovery bone densitometry (DXA) system (Hologic Inc., Waltham, MA, USA). Full informed consent was obtained from each of the subjects. Although some of the subjects had experienced back pain occasionally during their daily life, none of them reported any back pain or discomfort during the whole course of data acquisition. Their physical details are summarized in Table 1. The control group consisted of 10 female osteoporotic control subjects aged 66–81 years (mean 72.0). They were selected from the Southampton University Hospital Trust osteoporosis database. None of them had any evidence of vertebral fractures. The fracture group consisted of 6 women aged 66–78 years (mean 72.3), who had sustained a vertebral fracture due
347
to osteoporosis at level T12. Their fractures were clinically diagnosed and identified using Instant Vertebral Assessment (IVA) with the Hologic Discovery system mentioned above. The resulting morphological scans of the vertebrae were viewed and graded by clinicians with the semi-quantitative system of Genant et al. [19]. All fractures found were more than 3 months by the time of data collection and had been healed. All subjects had been undergoing routine medical management of osteoporosis as per local guidelines. There was no diagnosis for diseases (e.g. MRI for oedema) other than vertebral fracture. Potential subjects for both groups were excluded if they had other spinal disorders, impaired limb function, vertebroplasty, spinal implant or fusion. These criteria were confirmed with the chosen subjects via medical history and clinical examination. 2.2. Experimental setup and data collection Fluoroscopy sequences were collected with a Philips MultiDiagnost Eleva system (Philips Medical Systems; Best, The Netherlands) in its upright position. The diameter of flat panel detector was 38 cm. Pulse mode was used as it could reduce blurring and provide better quality when the object motion is large. Parameter settings (kVp and mA), In-Pulse control and, beam filtration were optimized using DoseWise (Philips Medical Systems, Best, The Netherlands) which is tailored to this fluoroscopy system for dose management and best image quality. Each subject was asked (i) to flex her trunk forward from the neutral position to her maximum range of flexion (sagittal flexion-outward phase) and then return back to the neutral position (sagittal flexion-inward phase); (ii) to flex maximally sideways to the left (left lateral flexion-outward phase) and then back to neutral (left lateral flexion-inward phase), and maximally to the right (right lateral flexion-outward phase) and then finally back to neutral position (right lateral flexioninward phase) as shown in Fig. 1(a) and (b), respectively. A tailor-made apparatus modified from a pump was designed to keep the shoulders aligned with the handle to minimize axial rotation and out-of-plane motion during sagittal flexion. When performing lateral flexion, the subject was asked to keep her head, shoulders and buttocks as close to the table as possible to prevent forward tilt and axial rotation. It started with the subject sliding her left hand along the lateral side of her left leg slowly. The subjects practiced several non-imaged motion cycles before fluoroscopy screening started to ensure the subject was comfortable and familiar with the motions and environment. The X-ray tube was centered on the T12 region for both control and fracture groups so that the region containing vertebrae of interest (i.e. from T10 to L2) was imaged. During sagittal flexion, sequences were taken as a lateral projection in which the X-rays entered the subject’s body laterally from the left and exited from the right side of the body. For lateral flexion, images were taken as an anteropos-
348
S.C.B. Lam et al. / Medical Engineering & Physics 31 (2009) 346–355
Table 1 Summary of subject characteristics and radiation dosages
Control group Fracture group
Number of subjects
Age (years)
Height (m)
Weight (kg)
BMI (kg/m2 )
Total effective dose (mSv)
Total exposure time (s)
10 6
72.0 (5.2) 72.3 (5.9)
1.63 (0.09) 1.60 (0.06)
70.8 (11.6) 72.2 (12.5)
26.5 (3.9) 28.2 (5.5)
0.38 (0.34) 0.25 (0.13)
91.2 (23.4) 99.5 (30.3)
Note. Values are presented in mean (S.D.). BMI stands for body mass index.
terior (AP) projection in which the X-ray beam traversed the body from anterior to posterior. The average total screening time for all subjects was 92.4 s (S.D. 22.2). With fluorograb, the averaged total effective dose was 0.3 mSv (S.D. 0.2) which is less than two lateral radiographs of the lumbar spine. Table 1 also lists the screening time and effective dose for both groups. The dose varies with the subject’s body size and bone and tissue densities. The image capture frame rate was 5 frames/s. Acquired data was stored as sequences of images by the built-in function of the system onto the hard drive which was then converted and stored in portable network graphics (PNG) format by Philips’ ViewForum software. 2.3. Fluoroscopy sequence landmarking All images were transferred to a PC for landmarking and analysis. Fig. 2(a) shows a frame of lateral projection from a fracture subject. We landmarked the four corners of the vertebral bodies from T10 to L2 for both projections. Fig. 2(b) shows the landmarks on a frame from a lateral flexion sequence taken by AP projection. A graphical user interface (GUI) was developed in Matlab (The MathWorks, Inc., Nat-
ick, MA, USA) for the landmarking procedure which can be downloaded from the author’s website [20]. The images of each motion were displayed as a sequence of frames through which the observer could go back and forth with the slide bar of the GUI. The image was shown in its actual pixel size in the GUI. 2.4. Kinematic analysis Based on the landmarked locations of the vertebrae in each frame, the rotation angle of each vertebra between two consecutive frames was calculated by finding the angle between the midplane lines [18] of that vertebra in these two frames. The angles were represented relative to their initial positions (i.e. the first frame). The resulting rotation angles were then smoothed using a moving average filter of window size 3 to reduce the random error and a median filter of order 3 to minimize the effects the outliers during the process of landmarking. The filter orders were selected by trial and error as the best compromise between filtering random errors while preserving kinematic features. All these functions were implemented in the GUI. The maximum angle of rotation of each flexion was defined as the RoM. Lateral flexion ratio
Fig. 1. Illustration of data acquisition and motions. The table of the DVF system was set upright. (a) Sagittal flexion. A device modified from a bicycle pump to provide resistance to the subject to maintain a stable speed when carrying out the motion. The subject held the handle of the device with her arm straight. Image sequences were taken as lateral projection. (b) Lateral flexion. The subject stood against the table with her head, shoulder and buttock lying closely to the table to prevent tilt forward and axial rotation. Lateral flexion started by sliding her left hand along the lateral side of her left leg slowly. Images were taken as anteroposterior (AP) projection.
S.C.B. Lam et al. / Medical Engineering & Physics 31 (2009) 346–355
349
Fig. 2. Example frames from fluoroscopy sequences taken from a fracture subject with T12 severe wedge fracture. (a) Lateral projection during sagittal flexion and (b) AP projection during lateral flexion with landmarks placed.
(LFR) was calculated using Eq. (1) to determine the symmetry of RoM towards the left and the right: RoMleft /RoMright if RoMleft < RoMright LFR = , (1) RoMright /RoMleft if RoMleft > RoMright where RoMleft and RoMright are the RoM of left and right lateral flexion, respectively. The closer the LFR to unity, the more symmetric the lateral flexion is. In order to examine the time taken to complete each motion phase across subjects, the whole motion cycle was normalized to unity since each subject carried out her motion at her own speed in which they took different time to complete their motions. Angles of rotation were plotted against the normalized motion cycle to obtain the motion curve. The percentages of a motion cycle at which the curve reached the peaks and zero crossing (τ A , τ B , τ C and τ D in Fig. 3) were determined. These points divided a motion cycle into the phases as defined in Section 2.2 and the proportion of time in a cycle taken to complete these phases could then be calculated for comparison. The 2-S.D. limit from the control mean was adopted as the normal range for each parameter. For the fracture subjects, the values of each parameter were compared with the normal range. A reproducibility test was performed using a set of 320 images (10 consecutive frames from each subject for each flexion) lankmarked three times on separate occasions by the same investigator. The overall standard deviations of the angle of rotation of all levels among the three landmarkings were 0.36◦ in sagittal flexion and 0.61◦ in lateral flexion.
3. Results The rotation angles of all five vertebrae were plotted against the normalized motion cycle as motion curves to visualize and depict the kinematics. Figs. 3a and 4a show motion curves during sagittal flexion averaged over the con-
trol group and fracture group, respectively. Among the 10 control subjects, the mean percentages of cycle attained sagittal flexion RoM (τ A in Fig. 3a) were about 55% for the five vertebrae. The mean percentages of τ A for the six fracture subjects ranged from 63% to 66% for the five vertebrae. Significant differences between two groups were found for all vertebrae (all P < 0.05). The mean RoM values during sagittal flexion (RoMs ) of the control group decreased from vertebra T10–L2 which were ranging from 44.2◦ (S.D. 8.3◦ ) to 35.3◦ (S.D. 9.8◦ ). Regarding the fracture group, RoMs ranged from 41.8◦ (S.D. 11.7◦ ) to 37.8◦ (S.D. 10.4◦ ). However, there was no statistical difference between the two groups (all P > 0.2). Each fracture subject has at least one vertebra of which the motion fell outside the normal range for the control subjects. Grouping in terms of vertebra level, out of the six fracture subjects, the number of fractured subjects whose motion fell outside the normal range of τ A in each level ranges from 1 to 3. Regarding the RoMs , there was only one subject whose motion fell outside the normal ranges for T10–L1 and none for L2. Figs. 3b and 4b show the angle of rotation motion curves of lateral flexion for the control and fracture groups, respectively. As indicated in Fig. 3b, the percentages of cycle attained RoMleft (τ B ), neutral position (τ C ) and RoMright (τ D ) were about 29%, 52%, and 78%, respectively, for the control group. The percentages for the fracture group were about 37%, between 57% and 60%, and about 85% for τ B , τ C and τ D , respectively. Significant differences were found between the two groups at τ B , τ C and τ D for all vertebrae (P < 0.05). The mean RoMleft values of the control group during lateral flexion ranged from 19.6◦ (S.D. 6.5◦ ) to 9.8◦ (S.D. 4.3◦ ), and the mean RoMright ranged from 21.7◦ (S.D. 5.9◦ ) to 10.4◦ (S.D. 4.9◦ ). In the fracture group, the values ranged from 25.0◦ (S.D. 5.8◦ ) to 16.9◦ (S.D. 5.8◦ ) for RoMleft , and from 20.6◦ (S.D. 7.9◦ ) to 16.7◦ (S.D. 5.8◦ ) for RoMright . All RoM values of both groups decreased along the spine from T10 to L2. Except the RoMleft of L2, there was no statistical difference between the two groups in terms of RoM (RoMleft
350
S.C.B. Lam et al. / Medical Engineering & Physics 31 (2009) 346–355
Fig. 3. Angle of rotation motion curves of vertebra T10–L2 averaged over the control group with phases defined: (a) sagittal flexion and (b) lateral flexion. τ A , τ B , τ C and τ D are the percentage of motion cycle taken to attain points A, B, C and D, respectively, as indicated in the curves.
Fig. 4. Angle of rotation motion curves of vertebra T10–L2 averaged over the fracture group with phases defined: (a) sagittal flexion and (b) lateral flexion.
S.C.B. Lam et al. / Medical Engineering & Physics 31 (2009) 346–355
of L2: P = 0.04; all other RoMs: P > 0.2). The mean LFR of each vertebra varied from 0.88 to 0.92 for the control group and from 0.72 to 0.76 for the fracture group. All vertebra levels showed statistical significant differences between the two groups (all P < 0.05). This suggests that fracture group had more asymmetric lateral flexions than the control. All fracture subjects had at least three vertebrae with abnormal motion. Grouping in terms of vertebra level, at least four out of the six fracture subjects whose motions fell outside the normal ranges of τ B , τ C and τ D for all vertebra levels. There were one and three fracture subjects whose RoMleft of L1 and L2 fell outside the normal ranges, respectively. Concerning the RoMright , one fracture subject fell outside normal ranges in each of the T10 and L1. There were at least four fracture subjects whose LFR fell outside the normal ranges for the five vertebrae. All kinematic results were summarized in Tables 2 and 3. The averaged angular velocities were 3.9◦ (S.D. 0.9◦ ) s−1 and 4.3◦ (S.D. 1.2◦ ) s−1 in the control group and fracture group, respectively, during sagittal flexion, and 4.6◦ (S.D. 1.6◦ ) s−1 and 4.0◦ (S.D. 1.6◦ ) s−1 in the control group and fracture group, respectively, during lateral flexion. They also showed no statistical difference (all P > 0.2). Looking at the motion curves of individual subjects, prolonged deflections in rotation were found in several vertebrae. They failed to rotate along with other vertebrae while the subject was flexing her back. Furthermore, they increased their speed to make themselves aligned with the other vertebrae when the subject almost reached at the RoM or returned to the neutral position. As indicated by the arrows in Fig. 5a, vertebra T10, T11, T12 and L2 exhibited prolonged deflections in the inward phase while returning to the neutral position during sagittal flexion. From lateral flexion motion curves (Fig. 5b), L1 and L2 were found to have prolonged deflections in the left-outward phase. Similarly, prolonged deflections were found from the T10 and L1 in the left-inward phase and L2 in the right-outward phase. Table 4 shows the number of prolonged deflections of each vertebra found from each phase. The number that found from the fracture group was more than the control group in most phases and vertebra levels.
4. Discussion Throughout the years, the feasibility of using fluoroscopy to study the kinematics of the spine in vivo has been verified and discussed by several groups (e.g. [13–17,21,22]). They studied asymptomatic subjects and subjects with different spinal disorders and used different experimental settings. The present study characterized the spinal motions from the fluoroscopy sequence with kinematic parameters and verified the usefulness of these parameters in differentiating the subjects with osteoporotic fractures at T12 from those without a fracture. The results provide an insight into the motions of the fractured spine and indicate the usefulness of quan-
351
titative characterization of the spine motions for functional assessment during diagnosis and rehabilitation. Sagittal and lateral flexions were investigated from the subjects in this study. These motions were initialized and carried out actively by the subject herself in a standing position with an assistive device for sagittal bending and clear procedural guidelines due to the unstable spinal conditions of the fractured subjects. Because of the different conditions in each individual subject, they were only able to carry out the motions voluntarily at their own rates with assistance of the pump which provided them sense of security and prevented them from falling. As we wished to study participants carrying out bending tasks at their own rates, it was necessary to normalize the kinematic parameters with respect to a unity motion cycle in order to make standardized comparisons. This approach is justifiable, so long as there was not a systematic difference in the angular velocities of the vertebrae between the fracture and non-fracture groups. Although we confirmed this observation, we clearly lacked statistical power for biologically relevant differences between the two groups (statistical power: ∼0.45 in both flexions). It is possible that Type II error could have influenced our findings. Nevertheless, the present data acquisition protocol was simple and could be transferable to study other kinds of spinal disorders with minimal modification. Pain may affect the bending motions but in the present study none of the subjects reported pain during data acquisition. It is not possible to eliminate the errors generated by manual landmarking. However, in order to minimize the effect, we post-processed the angle of rotation values with a moving average filter and a median filter. This preserved the original shape of the curve in the flexion graphs with the abrupt changes due to instability and reduced the effects of random noise and the outliers due to landmarking. Our repeatability test showed that the overall standard deviations of the angle of rotation of all levels were 0.36◦ and 0.61◦ in sagittal and lateral flexion, respectively. The errors attributed to the landmarking protocol used in this study were considered as acceptable, particularly given the point raised by Muggleton and Allen [23] that the angle value is subjected to a quantization error of ±0.25◦ . The values of the kinematic parameters of each fractured subject were compared with their 2-S.D. normal ranges which cover the 95% confidence intervals of the control means. The majority of the fractured subjects demonstrated abnormalities in LFR, percentage of cycle and prolonged deflections in most phases as they fell out of the defined normal range from the control mean for these kinematic parameters. Furthermore, statistically significant differences were obtained from the two groups for LFR and percentage of cycle. Prolonged deflections have also been found in subjects with degenerative spondylolisthesis which suggests that it might be related to the slipped segment or weaker muscles in the subjects [8]. In this study, it may be associated with altered vertebra shape due to fracture which disturbs the coordination of surrounding muscles.
352
S.C.B. Lam et al. / Medical Engineering & Physics 31 (2009) 346–355
Table 2 Mean values and standard deviations of range of motion (RoM) and lateral flexion ratio (LFR) of thoracolumbar vertebrae (T10–L2) Range of motion (RoM) T10
T11
T12
L1
L2
Sagittal flexion (RoMs ) Control group Fracture group Out of normal range#
44.2 (8.3) 41.8 (11.7) 1
42.5 (8.5) 41.3 (9.8) 1
41.1 (8.7) 41.1 (10.6) 1
39.2 (7.9) 36.5 (10.9) 1
35.3 (9.8) 37.8 (10.4) 0
Left lateral flexion (RoMleft ) Control group Fracture group Out of normal range#
19.6 (6.5) 25.0 (5.8) 0
17.8 (6.9) 23.5 (5.9) 0
15.9 (5.6) 21.2 (5.4) 0
12.9 (5.3) 19.3 (5.6) 1
9.8 (4.3) 16.9 (5.8)* 3
Right lateral flexion (RoMright ) Control group Fracture group Out of normal range#
21.7 (5.9) 20.6 (7.9) 1
18.7 (5.9) 19.7 (6.7) 0
16.3 (5.3) 18.9 (6.4) 0
13.7 (4.7) 17.1 (6.1) 1
10.4 (4.9) 16.7 (6.2) 0
Lateral flexion ratio (LFR) Control group Fracture group Out of normal range#
0.89 (0.06) 0.73 (0.23)* 4
0.90 (0.07) 0.76 (0.18)* 4
0.92 (0.07) 0.76 (0.17)* 5
0.88 (0.10) 0.72 (0.18)* 4
0.88 (0.09) 0.74 (0.15)* 4
Note. Values are presented in mean (S.D.). * P < 0.05 between control and fracture group. # Number of fracture subject fell out of the 2-S.D. normal range from the control mean.
Based on the results, LFR, percentage of cycle in each phase and detection of prolonged deflections would be useful to characterize spinal disorder as they could differentiate the fracture from the control. Furthermore, analysis of lateral flexion could provide important insights which most previous studies on spinal disorders with fluoroscopy have ignored. In addition to the fracture level (T12), abnormalities were found from neighboring vertebrae for all these parameters. This implies that the conditions of the neighboring vertebrae were disturbed by the fracture and suggests
the need to analyze several vertebrae as a system. Okawa et al. [8] also suggested that the motion of the whole lumbar spine might be affected by a single vertebral segment having spondylolisthesis. RoM of both sagittal and lateral flexions could not differentiate fractured from the control in this study. No statistical significance was found in previous studies either [8,22]. This result supports previous studies and suggests that RoM may be not a suitable parameter to characterize the motions. Instantaneous center of
Table 3 Mean values and standard deviations of percentage of motion cycle which thoracolumbar vertebrae (T10–L2) attained RoMs and neutral positions Percentage of cycle T11
T12
L1
L2
Attained sagittal flexion RoM—point A (τ A ) Control group 55.8 (6.5) Fracture group 63.4 (6.3)* Out of normal range# 1
T10
55.8 (6.5) 63.7 (5.8)* 2
55.8 (6.0) 64.2 (6.3)* 2
55.6 (6.8) 64.8 (7.5)* 2
56.7 (6.3) 66.2 (5.4)* 3
Attained left lateral flexion RoM—point B (τ B ) Control group 28.9 (2.7) Fracture group 37.2 (5.0)* Out of normal range# 5
29.3 (3.3) 36.8 (5.9)* 4
29.3 (3.3) 37.0 (7.7)* 5
29.6 (3.1) 36.7 (6.1)* 4
29.3 (3.3) 37.0 (5.1)* 5
Attained lateral flexion neutral position—point C (τ C ) Control group 51.9 (2.8) Fracture group 57.7 (4.7)* Out of normal range# 4
51.5 (2.1) 58.3 (2.9)* 4
51.9 (2.3) 59.8 (5.0)* 4
52.2 (2.9) 58.2 (5.8)* 4
51.8 (2.6) 57.2 (4.9)* 4
Attained right lateral flexion RoM—point D (τ D ) Control group 78.3 (2.0) Fracture group 85.6 (4.4)* # Out of normal range 5
78.1 (2.3) 85.7 (4.1)* 5
78.1 (2.3) 85.5 (4.5)* 6
78.3 (2.6) 84.8 (4.4)* 4
78.1 (1.4) 84.3 (4.5)* 6
Note. Values are presented in percent and mean (S.D.). * P < 0.05 between control and fracture group. # Number of fracture subject fell outside the 2-S.D. normal range from the control mean.
S.C.B. Lam et al. / Medical Engineering & Physics 31 (2009) 346–355
353
Fig. 5. Motion curves from a fracture subject. Arrows indicate prolonged deflections. (a) Sagittal flexion, prolonged deflection was found from T10, T11, T12 and L2 in the inward phase; (b) lateral flexion, prolonged deflection was found from L1 and L2 in the left-outward phase, T10 and L1 in the left-inward phase and L2 in the right-outward phase.
rotation (ICR) was not calculated here because of the controversy of its usefulness in spinal kinematic analysis especially data extracted from fluoroscopy sequences [24].
It would be desirable to determine when prolonged deflections appear. To do this, averaging the repeated motions by the same subject is needed to minimize the effect of variability. This will require further reduction of radiation exposure.
Table 4 Occurrence of prolonged deflections Number of prolonged deflections T10
T11
T12
L1
L2
Sagittal flexion-outward phase Control group Fracture group
0 0
0 0
0 1 (16.7%)
0 0
0 1 (16.7%)
0 2
Sagittal flexion-inward phase Control group Fracture group
0 2 (33.3%)
0 3 (50%)
1 (10%) 5 (83.3%)
0 2 (33.3%)
0 2 (33.3%)
1 14
Left lateral flexion-outward phase Control group 1 (10%) Fracture group 1 (16.7%)
1 (10%) 0
1 (10%) 2 (33.3%)
1 (10%) 3 (50%)
1 (10%) 2 (33.3%)
5 8
Left lateral flexion-inward phase Control group 1 (10%) Fracture group 1 (16.7%)
1 (10%) 2 (33.3%)
1 (10%) 1 (16.7%)
1 (10%) 3 (50%)
1 (10%) 2 (33.3%)
5 9
Right lateral flexion-outward phase Control group 0 Fracture group 3 (50%)
0 2 (33.3%)
0 3 (50%)
0 0
1 (10%) 3 (50%)
1 11
Right lateral flexion-inward phase Control group 1 (10%) Fracture group 2 (33.3%)
1 (10%) 3 (50%)
1 (10%) 2 (33.3%)
1 (10%) 2 (33.3%)
1 (10%) 3 (50%)
5 12
Note. Percentage occurrence over the group is in bracket.
Total
354
S.C.B. Lam et al. / Medical Engineering & Physics 31 (2009) 346–355
The present study did not aim to establish a method for clinical diagnosis of osteoporotic fracture since it was a pilot study and its clinical relevance still has not been fully understood. Instead, the present findings serve as a stepping-stone for future studies on parameterization of spine motions. A metric system combining all named parameters above should be developed to objectively and comprehensively quantify the spine motions. According to the semi-quantitative vertebral fracture assessment system of Genant et al. [19], types and severity of fracture are also the factors of osteoporotic fracture. In addition to change in vertebral body shape, other factors such as intravertebral body mobility and vacuum phenomenon [25,26] and discal or apophyseal joints alterations [27,28] would also affect spinal mobility. Having these studied with radiograph and MRI, together with information about vertebral fracture, more thorough understand regarding these multi-parametric causes to abnormal spinal motion can be obtained. Due to limited number of subjects in this study, categorization according to these variables was not feasible. The next step could be a multi-dimensional, parametric analysis of these parameters which requires a large population size to obtain results with statistical significance. Furthermore, clinical relevance can also be evaluated. A complementary development to make the present study clinically feasible would be having a large-scale validation of the landmarking technique for fractured vertebrae. Ultimately, it would be desirable to have an automated tracking technique such as [29] to replace the current laborintensive manual landmarking procedure. This will bring incentive for the community to employ the current spine motion analysis technique to a greater extent. Development of stand-up and open MRI technology would provide dynamics details of the discs and other tissues around the vertebrae during motions. Having this with fluoroscopy, researchers could study more in-depth how spinal motions are affected by changes in the components of the functional spinal unit [30,31]. On the other hand, this will start a new branch of research area involving dynamic image registration for images acquired from different imaging modalities.
5. Conclusion This study has indicated the usefulness of comparing the left and right RoM of lateral flexion (LFR), percentages of motion cycle and number of prolonged deflections in characterizing the spinal motions from, and differentiating between, subjects with and without osteoporotic fracture. The approach may provide a useful method of quantifying the degree of dysfunction in osteoporotic vertebral fracture and provide a basis for following the progress of treatment and rehabilitation.
Acknowledgements The authors would like to thank Southampton Osteoporosis Research Fund for financial support. We would also like to acknowledge the grateful help of the following personnel from Southampton General Hospital: Ms Jackie Stapleton and colleagues from X-ray Department for fluoroscopy sequences collection, Ms Sanchia Triggs and nurses from WTCRF for subject recruitment and their help with the study, Mr Philips Clewer for radiation protection advice and Ms Angela Jackson for assistance in the preparation of the ethics application. The inputs from reviewers are appreciated. Conflict of interest statement There is no conflict of interest. References [1] NIH. Osteoporosis prevention, diagnosis, and therapy. NIH Consens Statement 2000;17:1–45. [2] Dennison E, Cooper C. Epidemiology of osteoporotic fractures. Horm Res 2000;54:58–63. [3] Cooper C. The crippling consequences of fractures and their impact on quality of life. Am J Med 1997;103:12S–7S. [4] Miyakoshi N, Itoi E, Kobayashi M, Kodama H. Impact of postural deformities and spinal mobility on quality of life in postmenopausal osteoporosis. Osteoporos Int 2003;14:1007–12. [5] Liu-Ambrose T, Eng JJ, Khan KM, Mallinson A, Carter ND, McKay HA. The influence of back pain on balance and functional mobility in 65- to 75-year-old women with osteoporosis. Osteoporos Int 2002;13:868–73. [6] Purser JL, Pieper CF, Branch LG, Shipp KM, Gold DT, Lyles KW. Spinal deformity and mobility self-confidence among women with osteoporosis and vertebral fractures. Aging (Milano) 1999;11:235–45. [7] White A, Panjabi M. Clinical Biomechanics of the Spine. 2nd ed. Philadelphia: Williams and Wilkins/Lippincott; 1990. [8] Okawa A, Shinomiya K, Komori H, Muneta T, Arai Y, Nakai O. Dynamic motion study of the whole lumbar spine by videofluoroscopy. Spine 1998;23:1743–9. [9] Gertzbein SD, Seligman J, Holtby R, Chan KH, Kapasouri A, Tile M, et al. Centrode patterns and segmental instability in degenerative disc disease. Spine 1985;10:257–61. [10] Stokes IA, Wilder DG, Frymoyer JW, Pope MH. Assessment of patients with low-back pain by biplanar radiographic measurement of intervertebral motion. Spine 1981;6:233–40. [11] Schneider G, Pearcy MJ, Bogduk N. Abnormal motion in spondylolytic spondylolisthesis. Spine 2005;30:1159–64. [12] Breen AC, Allen R, Morris A. A digital videofluoroscopic technique for spine kinematics. J Med Eng Technol 1989;13:109–13. [13] Takayanagi K, Takahashi K, Yamagata M, Moriya H, Kitahara H, Tamaki T. Using cineradiography for continuous dynamic-motion analysis of the lumbar spine. Spine 2001;26:1858–65. [14] Vander Kooi D, Abad G, Basford JR, Maus TP, Yaszemski MJ, Kaufman KR. Lumbar spine stabilization with a thoracolumbosacral orthosis: evaluation with video fluoroscopy. Spine 2004;29:100–4. [15] Lee SW, Wong KW, Chan MK, Yeung HM, Chiu JL, Leong JC. Development and validation of a new technique for assessing lumbar spine motion. Spine 2002;27:E215–20. [16] Teyhen DS, Flynn TW, Bovik AC, Abraham LD. A new technique for digital fluoroscopic video assessment of sagittal plane lumbar spine motion. Spine 2005;30:E406–13.
S.C.B. Lam et al. / Medical Engineering & Physics 31 (2009) 346–355 [17] Breen AC, Muggleton JM, Mellor FE. An objective spinal motion imaging assessment (OSMIA): reliability, accuracy and exposure data. BMC Musculoskelet Disord 2006;7:pp. 1. [18] Frobin W, Brinckmann P, Leivseth G, Biggemann M, Reikeras O. Precision measurement of segmental motion from flexion-extension radiographs of the lumbar spine. Clin Biomech (Bristol, Avon) 1996;11:457–65. [19] Genant HK, Wu CY, van Kuijk C, Nevitt MC. Vertebral fracture assessment using a semiquantitative technique. J Bone Miner Res 1993;8:1137–48. [20] S.B. Lam, http://shingchunlam.googlepages.com. [21] Kanayama M, Abumi K, Kaneda K, Tadano S, Ukai T. Phase lag of the intersegmental motion in flexion–extension of the lumbar and lumbosacral spine. An in vivo study. Spine 1996;21:1416–22. [22] Otani K, Okawa A, Shinomiya K, Nakai O. Spondylolisthesis with postural slip reduction shows different motion patterns with videofluoroscopic analysis. J Orthop Sci 2005;10:152–9. [23] Muggleton JM, Allen R. Automatic location of vertebrae in digitized videofluoroscopic images of the lumbar spine. Med Eng Phys 1997;19:77–89. [24] Harvey SB, Hukins DW. Measurement of lumbar spinal flexion–extension kinematics from lateral radiographs: simulation of the effects of out-of-plane movement and errors in reference point placement. Med Eng Phys 1998;20:403–9.
355
[25] Kim DY, Lee SH, Jang JS, Chung SK, Lee HY. Intravertebral vacuum phenomenon in osteoporotic compression fracture: report of 67 cases with quantitative evaluation of intravertebral instability. J Neurosurg Spine 2004;100(1 Suppl.):24–31. [26] Pappou IP, Papadopoulos EC, Swanson AN, Cammisa FP, Girardi FP. Osteoporotic vertebral fractures and collapse with intravertebral vacuum sign (Kümmel’s disease). Orthopedics 2008;31: 61–6. [27] Adams MA, Dolan P. Spine biomechanics. J Biomech 2005;38: 1972–83. [28] Adams MA, Pollintine P, Tobias JH, Wakley GK, Dolan P. Intervertebral disc degeneration can predispose to anterior vertebral fractures in the thoracolumbar spine. J Bone Miner Res 2006;21:1409– 16. [29] Zheng Y, Nixon MS, Allen R. Automated segmentation of lumbar vertebrae in digital videofluoroscopic images. IEEE Trans Med Imaging 2004;23:45–52. [30] Gilbert JW, Wheeler GR, Lingreen RA, Johnson RR. Open stand-up MRI: a new instrument for positional neuroimaging. J Spinal Disord Tech 2006;19:151–4. [31] Goto T, Hamada K, Ito T, Nagao H, Takahashi T, Hayashida Y, et al. Interactive scan control for kinematic study in open MRI. Magn Reson Med Sci 2008;6:241–8.
Medical Engineering & Physics 31 (2009) 346–355
Parametric characterization of spinal motions in osteoporotic vertebral fracture at level T12 with fluoroscopy Shing Chun Benny Lam a , Robert Allen a,∗ , Gill Pearson b , Cyrus Cooper c,d b
a Institute of Sound and Vibration Research, University of Southampton, Southampton, UK Department of Medical Physics and Bioengineering, Southampton University Hospital NHS Trust, Southampton, UK c MRC Epidemiology Resource Centre, Southampton University Hospital NHS Trust, Southampton, UK d Botnar Research Institute, University of Oxford, Oxford, UK
Received 21 January 2008; received in revised form 24 May 2008; accepted 1 June 2008
Abstract Vertebral fractures due to osteoporosis are a common skeletal disorder affecting the mobility of the patients, although little is known about the relationship between spinal kinematics and osteoporotic fracture. The purpose of this study was to characterize the motions of the thoracolumbar spine affected by osteoporotic vertebral fracture at level T12 and compare the results with those of non-fracture osteoporosis subjects. We examined the continuous segmental kinematics of the vertebrae, and describe the segmental motion of the spine when a fracture at T12 is present. Fluoroscopy sequences of the thoracolumbar spines during sagittal and lateral flexion were collected from 16 subjects with osteoporosis of their spine (6 with vertebral fractures at T12, 10 without a fracture). Vertebrae T10–L2 in each frame of the sequences were landmarked. Kinematic parameters were calculated based on the landmarks and motion graphs were constructed. Compared to the control subjects who did not have a fracture, fracture subjects had a more asymmetric lateral range of motion (RoM) and required a longer time to complete certain phases of the motion cycle which are parameterized as lateral flexion ratio and percentage of motion cycle, respectively. Prolonged deflection was more frequently found from the fracture group. Characterizing the motions of the fractured vertebra together with its neighboring vertebrae with these kinematic parameters is useful in quantifying the dysfunction and may be a valuable aid to tracking progress of treatment. © 2008 IPEM. Published by Elsevier Ltd. All rights reserved. Keywords: Spine kinematics; Fluoroscopy; Osteoporosis; Vertebral fracture; Motion analysis
1. Introduction Osteoporosis is a common skeletal disorder characterized by compromised bone strength predisposing the sufferer to an increased risk of fracture [1]. Patients with osteoporotic spines can suffer from reduced mobility due to fractures. Reoccurrence rate leading to multiple fractures is high and the need of hospitalization is significant [2]. Furthermore, we have shown in [3] that osteoporotic fractures in elderly people were associated with their daily life functions. Previous studies also reported a significant correlation between quality of ∗ Corresponding author at: Signal Processing and Control Group, Institute of Sound and Vibration Research, University of Southampton, Southampton SO17 1BJ, UK. Tel.: +44 2380 593082. E-mail address: [email protected] (R. Allen).
life, number of vertebral fractures and lumbar lordosis angle, as well as spinal range of motion (RoM) [4–7]. However, none of these studies addressed the relationship between the motions of the spine and osteoporotic fracture. Stable spinal condition relies upon the idea that the anatomical and physiological conditions of every component of the functional spinal unit are normal and can coordinate effectively with each other [8]. Osteoporosis reduces the bone density, often leading to osteoporotic fracture which alters the normal shape of the vertebral body. Due to the structural changes of the vertebral body, the motions of the spinal column would be disturbed. Although the direct causeand-effect relationships between abnormal vertebral motions and spinal disorders have not been fully understood, previous studies had reported that spinal disorders such as low back pain, spondylolisthesis and disc degeneration could be
1350-4533/$ – see front matter © 2008 IPEM. Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.medengphy.2008.06.004
S.C.B. Lam et al. / Medical Engineering & Physics 31 (2009) 346–355
revealed in abnormal spinal motions, including increased mobility in one particular direction, abnormal spinal kinematics and altered instantaneous axes of rotation [8–11]. Fluoroscopy is an X-ray-based imaging modality employed in this study. As mentioned in previous studies, actual motions of each vertebra can be found from the fluoroscopy sequences [8,12–17]. They investigated the spine in different conditions by monitoring the changes in motions of the vertebrae during functional motions in different setups. Moreover, fluoroscopy allows a direct and objective assessment of the kinematics with a robust landmarking protocol [18]. More importantly, fluoroscopy has the advantage of involving relatively low radiation exposures. These features of fluoroscopy enable us to study the continuous motion of the vertebrae and no other devices can currently offer such an insight. The purpose of this study is to characterize the spine motions captured using fluoroscopy in vivo and verify the usefulness of the kinematic parameters used to characterize and differentiate subjects with an osteoporotic fracture at level T12 from controls. The technique would be used as a basis for assessing functional recovery with treatment and for assessing the patient at different times. The experimental setup and methods to extract and analyze the spinal motions are reported.
2. Method 2.1. Subjects The study was reviewed and approved by the National Health Services Local Research Ethics Committee and registered with the Southampton University Hospital Trust Research & Development Central Administrative Office. Advice from the Radiation Protection Adviser was sought. Two groups of female subjects with osteoporosis were recruited: a control group and a fracture group. These women were selected from subjects who had been routinely scanned for assessment of osteoporosis, and who had vertebral bone mineral density values of at least 2.5 standard deviations (S.D.) below the young adult mean (i.e. T-score at, or below, −2.5) using the Hologic Discovery bone densitometry (DXA) system (Hologic Inc., Waltham, MA, USA). Full informed consent was obtained from each of the subjects. Although some of the subjects had experienced back pain occasionally during their daily life, none of them reported any back pain or discomfort during the whole course of data acquisition. Their physical details are summarized in Table 1. The control group consisted of 10 female osteoporotic control subjects aged 66–81 years (mean 72.0). They were selected from the Southampton University Hospital Trust osteoporosis database. None of them had any evidence of vertebral fractures. The fracture group consisted of 6 women aged 66–78 years (mean 72.3), who had sustained a vertebral fracture due
347
to osteoporosis at level T12. Their fractures were clinically diagnosed and identified using Instant Vertebral Assessment (IVA) with the Hologic Discovery system mentioned above. The resulting morphological scans of the vertebrae were viewed and graded by clinicians with the semi-quantitative system of Genant et al. [19]. All fractures found were more than 3 months by the time of data collection and had been healed. All subjects had been undergoing routine medical management of osteoporosis as per local guidelines. There was no diagnosis for diseases (e.g. MRI for oedema) other than vertebral fracture. Potential subjects for both groups were excluded if they had other spinal disorders, impaired limb function, vertebroplasty, spinal implant or fusion. These criteria were confirmed with the chosen subjects via medical history and clinical examination. 2.2. Experimental setup and data collection Fluoroscopy sequences were collected with a Philips MultiDiagnost Eleva system (Philips Medical Systems; Best, The Netherlands) in its upright position. The diameter of flat panel detector was 38 cm. Pulse mode was used as it could reduce blurring and provide better quality when the object motion is large. Parameter settings (kVp and mA), In-Pulse control and, beam filtration were optimized using DoseWise (Philips Medical Systems, Best, The Netherlands) which is tailored to this fluoroscopy system for dose management and best image quality. Each subject was asked (i) to flex her trunk forward from the neutral position to her maximum range of flexion (sagittal flexion-outward phase) and then return back to the neutral position (sagittal flexion-inward phase); (ii) to flex maximally sideways to the left (left lateral flexion-outward phase) and then back to neutral (left lateral flexion-inward phase), and maximally to the right (right lateral flexion-outward phase) and then finally back to neutral position (right lateral flexioninward phase) as shown in Fig. 1(a) and (b), respectively. A tailor-made apparatus modified from a pump was designed to keep the shoulders aligned with the handle to minimize axial rotation and out-of-plane motion during sagittal flexion. When performing lateral flexion, the subject was asked to keep her head, shoulders and buttocks as close to the table as possible to prevent forward tilt and axial rotation. It started with the subject sliding her left hand along the lateral side of her left leg slowly. The subjects practiced several non-imaged motion cycles before fluoroscopy screening started to ensure the subject was comfortable and familiar with the motions and environment. The X-ray tube was centered on the T12 region for both control and fracture groups so that the region containing vertebrae of interest (i.e. from T10 to L2) was imaged. During sagittal flexion, sequences were taken as a lateral projection in which the X-rays entered the subject’s body laterally from the left and exited from the right side of the body. For lateral flexion, images were taken as an anteropos-
348
S.C.B. Lam et al. / Medical Engineering & Physics 31 (2009) 346–355
Table 1 Summary of subject characteristics and radiation dosages
Control group Fracture group
Number of subjects
Age (years)
Height (m)
Weight (kg)
BMI (kg/m2 )
Total effective dose (mSv)
Total exposure time (s)
10 6
72.0 (5.2) 72.3 (5.9)
1.63 (0.09) 1.60 (0.06)
70.8 (11.6) 72.2 (12.5)
26.5 (3.9) 28.2 (5.5)
0.38 (0.34) 0.25 (0.13)
91.2 (23.4) 99.5 (30.3)
Note. Values are presented in mean (S.D.). BMI stands for body mass index.
terior (AP) projection in which the X-ray beam traversed the body from anterior to posterior. The average total screening time for all subjects was 92.4 s (S.D. 22.2). With fluorograb, the averaged total effective dose was 0.3 mSv (S.D. 0.2) which is less than two lateral radiographs of the lumbar spine. Table 1 also lists the screening time and effective dose for both groups. The dose varies with the subject’s body size and bone and tissue densities. The image capture frame rate was 5 frames/s. Acquired data was stored as sequences of images by the built-in function of the system onto the hard drive which was then converted and stored in portable network graphics (PNG) format by Philips’ ViewForum software. 2.3. Fluoroscopy sequence landmarking All images were transferred to a PC for landmarking and analysis. Fig. 2(a) shows a frame of lateral projection from a fracture subject. We landmarked the four corners of the vertebral bodies from T10 to L2 for both projections. Fig. 2(b) shows the landmarks on a frame from a lateral flexion sequence taken by AP projection. A graphical user interface (GUI) was developed in Matlab (The MathWorks, Inc., Nat-
ick, MA, USA) for the landmarking procedure which can be downloaded from the author’s website [20]. The images of each motion were displayed as a sequence of frames through which the observer could go back and forth with the slide bar of the GUI. The image was shown in its actual pixel size in the GUI. 2.4. Kinematic analysis Based on the landmarked locations of the vertebrae in each frame, the rotation angle of each vertebra between two consecutive frames was calculated by finding the angle between the midplane lines [18] of that vertebra in these two frames. The angles were represented relative to their initial positions (i.e. the first frame). The resulting rotation angles were then smoothed using a moving average filter of window size 3 to reduce the random error and a median filter of order 3 to minimize the effects the outliers during the process of landmarking. The filter orders were selected by trial and error as the best compromise between filtering random errors while preserving kinematic features. All these functions were implemented in the GUI. The maximum angle of rotation of each flexion was defined as the RoM. Lateral flexion ratio
Fig. 1. Illustration of data acquisition and motions. The table of the DVF system was set upright. (a) Sagittal flexion. A device modified from a bicycle pump to provide resistance to the subject to maintain a stable speed when carrying out the motion. The subject held the handle of the device with her arm straight. Image sequences were taken as lateral projection. (b) Lateral flexion. The subject stood against the table with her head, shoulder and buttock lying closely to the table to prevent tilt forward and axial rotation. Lateral flexion started by sliding her left hand along the lateral side of her left leg slowly. Images were taken as anteroposterior (AP) projection.
S.C.B. Lam et al. / Medical Engineering & Physics 31 (2009) 346–355
349
Fig. 2. Example frames from fluoroscopy sequences taken from a fracture subject with T12 severe wedge fracture. (a) Lateral projection during sagittal flexion and (b) AP projection during lateral flexion with landmarks placed.
(LFR) was calculated using Eq. (1) to determine the symmetry of RoM towards the left and the right: RoMleft /RoMright if RoMleft < RoMright LFR = , (1) RoMright /RoMleft if RoMleft > RoMright where RoMleft and RoMright are the RoM of left and right lateral flexion, respectively. The closer the LFR to unity, the more symmetric the lateral flexion is. In order to examine the time taken to complete each motion phase across subjects, the whole motion cycle was normalized to unity since each subject carried out her motion at her own speed in which they took different time to complete their motions. Angles of rotation were plotted against the normalized motion cycle to obtain the motion curve. The percentages of a motion cycle at which the curve reached the peaks and zero crossing (τ A , τ B , τ C and τ D in Fig. 3) were determined. These points divided a motion cycle into the phases as defined in Section 2.2 and the proportion of time in a cycle taken to complete these phases could then be calculated for comparison. The 2-S.D. limit from the control mean was adopted as the normal range for each parameter. For the fracture subjects, the values of each parameter were compared with the normal range. A reproducibility test was performed using a set of 320 images (10 consecutive frames from each subject for each flexion) lankmarked three times on separate occasions by the same investigator. The overall standard deviations of the angle of rotation of all levels among the three landmarkings were 0.36◦ in sagittal flexion and 0.61◦ in lateral flexion.
3. Results The rotation angles of all five vertebrae were plotted against the normalized motion cycle as motion curves to visualize and depict the kinematics. Figs. 3a and 4a show motion curves during sagittal flexion averaged over the con-
trol group and fracture group, respectively. Among the 10 control subjects, the mean percentages of cycle attained sagittal flexion RoM (τ A in Fig. 3a) were about 55% for the five vertebrae. The mean percentages of τ A for the six fracture subjects ranged from 63% to 66% for the five vertebrae. Significant differences between two groups were found for all vertebrae (all P < 0.05). The mean RoM values during sagittal flexion (RoMs ) of the control group decreased from vertebra T10–L2 which were ranging from 44.2◦ (S.D. 8.3◦ ) to 35.3◦ (S.D. 9.8◦ ). Regarding the fracture group, RoMs ranged from 41.8◦ (S.D. 11.7◦ ) to 37.8◦ (S.D. 10.4◦ ). However, there was no statistical difference between the two groups (all P > 0.2). Each fracture subject has at least one vertebra of which the motion fell outside the normal range for the control subjects. Grouping in terms of vertebra level, out of the six fracture subjects, the number of fractured subjects whose motion fell outside the normal range of τ A in each level ranges from 1 to 3. Regarding the RoMs , there was only one subject whose motion fell outside the normal ranges for T10–L1 and none for L2. Figs. 3b and 4b show the angle of rotation motion curves of lateral flexion for the control and fracture groups, respectively. As indicated in Fig. 3b, the percentages of cycle attained RoMleft (τ B ), neutral position (τ C ) and RoMright (τ D ) were about 29%, 52%, and 78%, respectively, for the control group. The percentages for the fracture group were about 37%, between 57% and 60%, and about 85% for τ B , τ C and τ D , respectively. Significant differences were found between the two groups at τ B , τ C and τ D for all vertebrae (P < 0.05). The mean RoMleft values of the control group during lateral flexion ranged from 19.6◦ (S.D. 6.5◦ ) to 9.8◦ (S.D. 4.3◦ ), and the mean RoMright ranged from 21.7◦ (S.D. 5.9◦ ) to 10.4◦ (S.D. 4.9◦ ). In the fracture group, the values ranged from 25.0◦ (S.D. 5.8◦ ) to 16.9◦ (S.D. 5.8◦ ) for RoMleft , and from 20.6◦ (S.D. 7.9◦ ) to 16.7◦ (S.D. 5.8◦ ) for RoMright . All RoM values of both groups decreased along the spine from T10 to L2. Except the RoMleft of L2, there was no statistical difference between the two groups in terms of RoM (RoMleft
350
S.C.B. Lam et al. / Medical Engineering & Physics 31 (2009) 346–355
Fig. 3. Angle of rotation motion curves of vertebra T10–L2 averaged over the control group with phases defined: (a) sagittal flexion and (b) lateral flexion. τ A , τ B , τ C and τ D are the percentage of motion cycle taken to attain points A, B, C and D, respectively, as indicated in the curves.
Fig. 4. Angle of rotation motion curves of vertebra T10–L2 averaged over the fracture group with phases defined: (a) sagittal flexion and (b) lateral flexion.
S.C.B. Lam et al. / Medical Engineering & Physics 31 (2009) 346–355
of L2: P = 0.04; all other RoMs: P > 0.2). The mean LFR of each vertebra varied from 0.88 to 0.92 for the control group and from 0.72 to 0.76 for the fracture group. All vertebra levels showed statistical significant differences between the two groups (all P < 0.05). This suggests that fracture group had more asymmetric lateral flexions than the control. All fracture subjects had at least three vertebrae with abnormal motion. Grouping in terms of vertebra level, at least four out of the six fracture subjects whose motions fell outside the normal ranges of τ B , τ C and τ D for all vertebra levels. There were one and three fracture subjects whose RoMleft of L1 and L2 fell outside the normal ranges, respectively. Concerning the RoMright , one fracture subject fell outside normal ranges in each of the T10 and L1. There were at least four fracture subjects whose LFR fell outside the normal ranges for the five vertebrae. All kinematic results were summarized in Tables 2 and 3. The averaged angular velocities were 3.9◦ (S.D. 0.9◦ ) s−1 and 4.3◦ (S.D. 1.2◦ ) s−1 in the control group and fracture group, respectively, during sagittal flexion, and 4.6◦ (S.D. 1.6◦ ) s−1 and 4.0◦ (S.D. 1.6◦ ) s−1 in the control group and fracture group, respectively, during lateral flexion. They also showed no statistical difference (all P > 0.2). Looking at the motion curves of individual subjects, prolonged deflections in rotation were found in several vertebrae. They failed to rotate along with other vertebrae while the subject was flexing her back. Furthermore, they increased their speed to make themselves aligned with the other vertebrae when the subject almost reached at the RoM or returned to the neutral position. As indicated by the arrows in Fig. 5a, vertebra T10, T11, T12 and L2 exhibited prolonged deflections in the inward phase while returning to the neutral position during sagittal flexion. From lateral flexion motion curves (Fig. 5b), L1 and L2 were found to have prolonged deflections in the left-outward phase. Similarly, prolonged deflections were found from the T10 and L1 in the left-inward phase and L2 in the right-outward phase. Table 4 shows the number of prolonged deflections of each vertebra found from each phase. The number that found from the fracture group was more than the control group in most phases and vertebra levels.
4. Discussion Throughout the years, the feasibility of using fluoroscopy to study the kinematics of the spine in vivo has been verified and discussed by several groups (e.g. [13–17,21,22]). They studied asymptomatic subjects and subjects with different spinal disorders and used different experimental settings. The present study characterized the spinal motions from the fluoroscopy sequence with kinematic parameters and verified the usefulness of these parameters in differentiating the subjects with osteoporotic fractures at T12 from those without a fracture. The results provide an insight into the motions of the fractured spine and indicate the usefulness of quan-
351
titative characterization of the spine motions for functional assessment during diagnosis and rehabilitation. Sagittal and lateral flexions were investigated from the subjects in this study. These motions were initialized and carried out actively by the subject herself in a standing position with an assistive device for sagittal bending and clear procedural guidelines due to the unstable spinal conditions of the fractured subjects. Because of the different conditions in each individual subject, they were only able to carry out the motions voluntarily at their own rates with assistance of the pump which provided them sense of security and prevented them from falling. As we wished to study participants carrying out bending tasks at their own rates, it was necessary to normalize the kinematic parameters with respect to a unity motion cycle in order to make standardized comparisons. This approach is justifiable, so long as there was not a systematic difference in the angular velocities of the vertebrae between the fracture and non-fracture groups. Although we confirmed this observation, we clearly lacked statistical power for biologically relevant differences between the two groups (statistical power: ∼0.45 in both flexions). It is possible that Type II error could have influenced our findings. Nevertheless, the present data acquisition protocol was simple and could be transferable to study other kinds of spinal disorders with minimal modification. Pain may affect the bending motions but in the present study none of the subjects reported pain during data acquisition. It is not possible to eliminate the errors generated by manual landmarking. However, in order to minimize the effect, we post-processed the angle of rotation values with a moving average filter and a median filter. This preserved the original shape of the curve in the flexion graphs with the abrupt changes due to instability and reduced the effects of random noise and the outliers due to landmarking. Our repeatability test showed that the overall standard deviations of the angle of rotation of all levels were 0.36◦ and 0.61◦ in sagittal and lateral flexion, respectively. The errors attributed to the landmarking protocol used in this study were considered as acceptable, particularly given the point raised by Muggleton and Allen [23] that the angle value is subjected to a quantization error of ±0.25◦ . The values of the kinematic parameters of each fractured subject were compared with their 2-S.D. normal ranges which cover the 95% confidence intervals of the control means. The majority of the fractured subjects demonstrated abnormalities in LFR, percentage of cycle and prolonged deflections in most phases as they fell out of the defined normal range from the control mean for these kinematic parameters. Furthermore, statistically significant differences were obtained from the two groups for LFR and percentage of cycle. Prolonged deflections have also been found in subjects with degenerative spondylolisthesis which suggests that it might be related to the slipped segment or weaker muscles in the subjects [8]. In this study, it may be associated with altered vertebra shape due to fracture which disturbs the coordination of surrounding muscles.
352
S.C.B. Lam et al. / Medical Engineering & Physics 31 (2009) 346–355
Table 2 Mean values and standard deviations of range of motion (RoM) and lateral flexion ratio (LFR) of thoracolumbar vertebrae (T10–L2) Range of motion (RoM) T10
T11
T12
L1
L2
Sagittal flexion (RoMs ) Control group Fracture group Out of normal range#
44.2 (8.3) 41.8 (11.7) 1
42.5 (8.5) 41.3 (9.8) 1
41.1 (8.7) 41.1 (10.6) 1
39.2 (7.9) 36.5 (10.9) 1
35.3 (9.8) 37.8 (10.4) 0
Left lateral flexion (RoMleft ) Control group Fracture group Out of normal range#
19.6 (6.5) 25.0 (5.8) 0
17.8 (6.9) 23.5 (5.9) 0
15.9 (5.6) 21.2 (5.4) 0
12.9 (5.3) 19.3 (5.6) 1
9.8 (4.3) 16.9 (5.8)* 3
Right lateral flexion (RoMright ) Control group Fracture group Out of normal range#
21.7 (5.9) 20.6 (7.9) 1
18.7 (5.9) 19.7 (6.7) 0
16.3 (5.3) 18.9 (6.4) 0
13.7 (4.7) 17.1 (6.1) 1
10.4 (4.9) 16.7 (6.2) 0
Lateral flexion ratio (LFR) Control group Fracture group Out of normal range#
0.89 (0.06) 0.73 (0.23)* 4
0.90 (0.07) 0.76 (0.18)* 4
0.92 (0.07) 0.76 (0.17)* 5
0.88 (0.10) 0.72 (0.18)* 4
0.88 (0.09) 0.74 (0.15)* 4
Note. Values are presented in mean (S.D.). * P < 0.05 between control and fracture group. # Number of fracture subject fell out of the 2-S.D. normal range from the control mean.
Based on the results, LFR, percentage of cycle in each phase and detection of prolonged deflections would be useful to characterize spinal disorder as they could differentiate the fracture from the control. Furthermore, analysis of lateral flexion could provide important insights which most previous studies on spinal disorders with fluoroscopy have ignored. In addition to the fracture level (T12), abnormalities were found from neighboring vertebrae for all these parameters. This implies that the conditions of the neighboring vertebrae were disturbed by the fracture and suggests
the need to analyze several vertebrae as a system. Okawa et al. [8] also suggested that the motion of the whole lumbar spine might be affected by a single vertebral segment having spondylolisthesis. RoM of both sagittal and lateral flexions could not differentiate fractured from the control in this study. No statistical significance was found in previous studies either [8,22]. This result supports previous studies and suggests that RoM may be not a suitable parameter to characterize the motions. Instantaneous center of
Table 3 Mean values and standard deviations of percentage of motion cycle which thoracolumbar vertebrae (T10–L2) attained RoMs and neutral positions Percentage of cycle T11
T12
L1
L2
Attained sagittal flexion RoM—point A (τ A ) Control group 55.8 (6.5) Fracture group 63.4 (6.3)* Out of normal range# 1
T10
55.8 (6.5) 63.7 (5.8)* 2
55.8 (6.0) 64.2 (6.3)* 2
55.6 (6.8) 64.8 (7.5)* 2
56.7 (6.3) 66.2 (5.4)* 3
Attained left lateral flexion RoM—point B (τ B ) Control group 28.9 (2.7) Fracture group 37.2 (5.0)* Out of normal range# 5
29.3 (3.3) 36.8 (5.9)* 4
29.3 (3.3) 37.0 (7.7)* 5
29.6 (3.1) 36.7 (6.1)* 4
29.3 (3.3) 37.0 (5.1)* 5
Attained lateral flexion neutral position—point C (τ C ) Control group 51.9 (2.8) Fracture group 57.7 (4.7)* Out of normal range# 4
51.5 (2.1) 58.3 (2.9)* 4
51.9 (2.3) 59.8 (5.0)* 4
52.2 (2.9) 58.2 (5.8)* 4
51.8 (2.6) 57.2 (4.9)* 4
Attained right lateral flexion RoM—point D (τ D ) Control group 78.3 (2.0) Fracture group 85.6 (4.4)* # Out of normal range 5
78.1 (2.3) 85.7 (4.1)* 5
78.1 (2.3) 85.5 (4.5)* 6
78.3 (2.6) 84.8 (4.4)* 4
78.1 (1.4) 84.3 (4.5)* 6
Note. Values are presented in percent and mean (S.D.). * P < 0.05 between control and fracture group. # Number of fracture subject fell outside the 2-S.D. normal range from the control mean.
S.C.B. Lam et al. / Medical Engineering & Physics 31 (2009) 346–355
353
Fig. 5. Motion curves from a fracture subject. Arrows indicate prolonged deflections. (a) Sagittal flexion, prolonged deflection was found from T10, T11, T12 and L2 in the inward phase; (b) lateral flexion, prolonged deflection was found from L1 and L2 in the left-outward phase, T10 and L1 in the left-inward phase and L2 in the right-outward phase.
rotation (ICR) was not calculated here because of the controversy of its usefulness in spinal kinematic analysis especially data extracted from fluoroscopy sequences [24].
It would be desirable to determine when prolonged deflections appear. To do this, averaging the repeated motions by the same subject is needed to minimize the effect of variability. This will require further reduction of radiation exposure.
Table 4 Occurrence of prolonged deflections Number of prolonged deflections T10
T11
T12
L1
L2
Sagittal flexion-outward phase Control group Fracture group
0 0
0 0
0 1 (16.7%)
0 0
0 1 (16.7%)
0 2
Sagittal flexion-inward phase Control group Fracture group
0 2 (33.3%)
0 3 (50%)
1 (10%) 5 (83.3%)
0 2 (33.3%)
0 2 (33.3%)
1 14
Left lateral flexion-outward phase Control group 1 (10%) Fracture group 1 (16.7%)
1 (10%) 0
1 (10%) 2 (33.3%)
1 (10%) 3 (50%)
1 (10%) 2 (33.3%)
5 8
Left lateral flexion-inward phase Control group 1 (10%) Fracture group 1 (16.7%)
1 (10%) 2 (33.3%)
1 (10%) 1 (16.7%)
1 (10%) 3 (50%)
1 (10%) 2 (33.3%)
5 9
Right lateral flexion-outward phase Control group 0 Fracture group 3 (50%)
0 2 (33.3%)
0 3 (50%)
0 0
1 (10%) 3 (50%)
1 11
Right lateral flexion-inward phase Control group 1 (10%) Fracture group 2 (33.3%)
1 (10%) 3 (50%)
1 (10%) 2 (33.3%)
1 (10%) 2 (33.3%)
1 (10%) 3 (50%)
5 12
Note. Percentage occurrence over the group is in bracket.
Total
354
S.C.B. Lam et al. / Medical Engineering & Physics 31 (2009) 346–355
The present study did not aim to establish a method for clinical diagnosis of osteoporotic fracture since it was a pilot study and its clinical relevance still has not been fully understood. Instead, the present findings serve as a stepping-stone for future studies on parameterization of spine motions. A metric system combining all named parameters above should be developed to objectively and comprehensively quantify the spine motions. According to the semi-quantitative vertebral fracture assessment system of Genant et al. [19], types and severity of fracture are also the factors of osteoporotic fracture. In addition to change in vertebral body shape, other factors such as intravertebral body mobility and vacuum phenomenon [25,26] and discal or apophyseal joints alterations [27,28] would also affect spinal mobility. Having these studied with radiograph and MRI, together with information about vertebral fracture, more thorough understand regarding these multi-parametric causes to abnormal spinal motion can be obtained. Due to limited number of subjects in this study, categorization according to these variables was not feasible. The next step could be a multi-dimensional, parametric analysis of these parameters which requires a large population size to obtain results with statistical significance. Furthermore, clinical relevance can also be evaluated. A complementary development to make the present study clinically feasible would be having a large-scale validation of the landmarking technique for fractured vertebrae. Ultimately, it would be desirable to have an automated tracking technique such as [29] to replace the current laborintensive manual landmarking procedure. This will bring incentive for the community to employ the current spine motion analysis technique to a greater extent. Development of stand-up and open MRI technology would provide dynamics details of the discs and other tissues around the vertebrae during motions. Having this with fluoroscopy, researchers could study more in-depth how spinal motions are affected by changes in the components of the functional spinal unit [30,31]. On the other hand, this will start a new branch of research area involving dynamic image registration for images acquired from different imaging modalities.
5. Conclusion This study has indicated the usefulness of comparing the left and right RoM of lateral flexion (LFR), percentages of motion cycle and number of prolonged deflections in characterizing the spinal motions from, and differentiating between, subjects with and without osteoporotic fracture. The approach may provide a useful method of quantifying the degree of dysfunction in osteoporotic vertebral fracture and provide a basis for following the progress of treatment and rehabilitation.
Acknowledgements The authors would like to thank Southampton Osteoporosis Research Fund for financial support. We would also like to acknowledge the grateful help of the following personnel from Southampton General Hospital: Ms Jackie Stapleton and colleagues from X-ray Department for fluoroscopy sequences collection, Ms Sanchia Triggs and nurses from WTCRF for subject recruitment and their help with the study, Mr Philips Clewer for radiation protection advice and Ms Angela Jackson for assistance in the preparation of the ethics application. The inputs from reviewers are appreciated. Conflict of interest statement There is no conflict of interest. References [1] NIH. Osteoporosis prevention, diagnosis, and therapy. NIH Consens Statement 2000;17:1–45. [2] Dennison E, Cooper C. Epidemiology of osteoporotic fractures. Horm Res 2000;54:58–63. [3] Cooper C. The crippling consequences of fractures and their impact on quality of life. Am J Med 1997;103:12S–7S. [4] Miyakoshi N, Itoi E, Kobayashi M, Kodama H. Impact of postural deformities and spinal mobility on quality of life in postmenopausal osteoporosis. Osteoporos Int 2003;14:1007–12. [5] Liu-Ambrose T, Eng JJ, Khan KM, Mallinson A, Carter ND, McKay HA. The influence of back pain on balance and functional mobility in 65- to 75-year-old women with osteoporosis. Osteoporos Int 2002;13:868–73. [6] Purser JL, Pieper CF, Branch LG, Shipp KM, Gold DT, Lyles KW. Spinal deformity and mobility self-confidence among women with osteoporosis and vertebral fractures. Aging (Milano) 1999;11:235–45. [7] White A, Panjabi M. Clinical Biomechanics of the Spine. 2nd ed. Philadelphia: Williams and Wilkins/Lippincott; 1990. [8] Okawa A, Shinomiya K, Komori H, Muneta T, Arai Y, Nakai O. Dynamic motion study of the whole lumbar spine by videofluoroscopy. Spine 1998;23:1743–9. [9] Gertzbein SD, Seligman J, Holtby R, Chan KH, Kapasouri A, Tile M, et al. Centrode patterns and segmental instability in degenerative disc disease. Spine 1985;10:257–61. [10] Stokes IA, Wilder DG, Frymoyer JW, Pope MH. Assessment of patients with low-back pain by biplanar radiographic measurement of intervertebral motion. Spine 1981;6:233–40. [11] Schneider G, Pearcy MJ, Bogduk N. Abnormal motion in spondylolytic spondylolisthesis. Spine 2005;30:1159–64. [12] Breen AC, Allen R, Morris A. A digital videofluoroscopic technique for spine kinematics. J Med Eng Technol 1989;13:109–13. [13] Takayanagi K, Takahashi K, Yamagata M, Moriya H, Kitahara H, Tamaki T. Using cineradiography for continuous dynamic-motion analysis of the lumbar spine. Spine 2001;26:1858–65. [14] Vander Kooi D, Abad G, Basford JR, Maus TP, Yaszemski MJ, Kaufman KR. Lumbar spine stabilization with a thoracolumbosacral orthosis: evaluation with video fluoroscopy. Spine 2004;29:100–4. [15] Lee SW, Wong KW, Chan MK, Yeung HM, Chiu JL, Leong JC. Development and validation of a new technique for assessing lumbar spine motion. Spine 2002;27:E215–20. [16] Teyhen DS, Flynn TW, Bovik AC, Abraham LD. A new technique for digital fluoroscopic video assessment of sagittal plane lumbar spine motion. Spine 2005;30:E406–13.
S.C.B. Lam et al. / Medical Engineering & Physics 31 (2009) 346–355 [17] Breen AC, Muggleton JM, Mellor FE. An objective spinal motion imaging assessment (OSMIA): reliability, accuracy and exposure data. BMC Musculoskelet Disord 2006;7:pp. 1. [18] Frobin W, Brinckmann P, Leivseth G, Biggemann M, Reikeras O. Precision measurement of segmental motion from flexion-extension radiographs of the lumbar spine. Clin Biomech (Bristol, Avon) 1996;11:457–65. [19] Genant HK, Wu CY, van Kuijk C, Nevitt MC. Vertebral fracture assessment using a semiquantitative technique. J Bone Miner Res 1993;8:1137–48. [20] S.B. Lam, http://shingchunlam.googlepages.com. [21] Kanayama M, Abumi K, Kaneda K, Tadano S, Ukai T. Phase lag of the intersegmental motion in flexion–extension of the lumbar and lumbosacral spine. An in vivo study. Spine 1996;21:1416–22. [22] Otani K, Okawa A, Shinomiya K, Nakai O. Spondylolisthesis with postural slip reduction shows different motion patterns with videofluoroscopic analysis. J Orthop Sci 2005;10:152–9. [23] Muggleton JM, Allen R. Automatic location of vertebrae in digitized videofluoroscopic images of the lumbar spine. Med Eng Phys 1997;19:77–89. [24] Harvey SB, Hukins DW. Measurement of lumbar spinal flexion–extension kinematics from lateral radiographs: simulation of the effects of out-of-plane movement and errors in reference point placement. Med Eng Phys 1998;20:403–9.
355
[25] Kim DY, Lee SH, Jang JS, Chung SK, Lee HY. Intravertebral vacuum phenomenon in osteoporotic compression fracture: report of 67 cases with quantitative evaluation of intravertebral instability. J Neurosurg Spine 2004;100(1 Suppl.):24–31. [26] Pappou IP, Papadopoulos EC, Swanson AN, Cammisa FP, Girardi FP. Osteoporotic vertebral fractures and collapse with intravertebral vacuum sign (Kümmel’s disease). Orthopedics 2008;31: 61–6. [27] Adams MA, Dolan P. Spine biomechanics. J Biomech 2005;38: 1972–83. [28] Adams MA, Pollintine P, Tobias JH, Wakley GK, Dolan P. Intervertebral disc degeneration can predispose to anterior vertebral fractures in the thoracolumbar spine. J Bone Miner Res 2006;21:1409– 16. [29] Zheng Y, Nixon MS, Allen R. Automated segmentation of lumbar vertebrae in digital videofluoroscopic images. IEEE Trans Med Imaging 2004;23:45–52. [30] Gilbert JW, Wheeler GR, Lingreen RA, Johnson RR. Open stand-up MRI: a new instrument for positional neuroimaging. J Spinal Disord Tech 2006;19:151–4. [31] Goto T, Hamada K, Ito T, Nagao H, Takahashi T, Hayashida Y, et al. Interactive scan control for kinematic study in open MRI. Magn Reson Med Sci 2008;6:241–8.