The Validity of Rasterstereography: A Systematic Review.

To investigate and monitor the progression of scoliosis and other spinal deformities in patients following idiopathic scoliosis (IS), non-invasive and radiation-free techniques are recommended because of the need for repeated radiographs. In a clinical setting, spine parameters can be quickly, cheaply and easily assessed using rasterstereography (RS). To assess the validity of the radiation-free technique RS based on surface topography compared with radiographs. MEDLINE, the Cochrane Library and EMBASE were systematically searched for studies which investigate the validity of rasterstereography compared with x-ray measurements. Studies published between January 1, 1990 and July 31, 2013 in English, German and French were included. Studies dealing with magnetic resonance imaging were excluded. Twelve studies with 570 patients were included; these articles were published between 1990 and 2013. The majority of studies investigated patients with IS, but other spinal pathologies included were thoracic hyperkyphosis and Scheuermann's disease. With regard to the quality assessment criteria for the included studies, three out of twelve studies were evaluated using a twelve point scale and two used a scale with eleven points. We conclude that RS facilitates clinical practice by analysing the spinal column. It is completely radiation-free and could help to monitor scoliosis progression.

Introduction

Various subjective and objective methods have been developed to analyze and quantify scoliosis and other spinal deformities.[1-5] To evaluate the degree of deformity in the diagnosis and treatment of scoliosis and other spinal deformities examinations such as roentgenograms or computed tomography are frequently used.[4,6,7] The use of two-dimensional anterior-posterior (a.p.) full-length spine radiologic investigation is accepted as the mainstay to attest the medical diagnosis of idiopathic scoliosis, despite providing data only in two dimensions.[8] The obvious disadvantage of such instrumental assessment method is the fear of increased carcinogenic risk and infertility from repeated exposure to ionizing radiation.[9-12]

The reason for the repeated radiographs is the requirement to measure the type, the flexibility and progression of the spinal curvatures in follow-up examinations in definite time intervals, which is obtained from two-dimensional a.p. spine radiographs.[8] In the study of Nash and colleagues teenage girls with idiopathic scoliosis received over a treatment period of three years 22 roentgenograms.[11] One alternative examination suitable for this purpose is the light-sectioning method rasterstereography (RS), which is a precise, radiation-free and inexpensive and that is in routine clinical use in many scoliosis centers throughout the world.[13] The method, which was developed by Drerup and Hierholzer in the 1980s, has been confirmed to be reliable both in pre- and postsurgical scoliosis patients for supplementing radiological and clinical examinations.[14-18] By detecting anatomical landmarks with characteristic shape parameters – the vertebra prominens and the two spina iliaca posterior superior – coordinate data of back surface points and the line of symmetry can be determined.[19] RS provides a reliable method for three-dimensional back shape analysis and reconstruction of spinal deformities.[20,21] Several studies have evaluated the validity of RS compared with X-ray.[22-25] Therefore, the aim of the present research was to evaluate the validity and accuracy of RS compared with X-ray in a systematic literature review.

Inclusion criteria and study identification

A systematic review of the literature was conducted in July 2013 using the PubMed data base of the National Library of Medicine, Embase and the Cochrane library for relevant trials indexed between January 1, 1990, and July 31, 2013. To be included in the literature review, articles had to meet the following criteria: i) original studies that investigated the validity of RS compared with X-ray measurement and ii) published between January 1, 1990, and July 31, 2013, in the English, French or German language. The following search items were used: rasterstereography, rasterstereographic AND X-ray, rasterstereography OR rasterstereographic. Studies were also excluded if they lacked standard X-ray measurement of the spine, e.g. MR tomography or a lack of evaluation of spine parameters or if they analysis of difference in leg length or pelvic obliquity.

Those papers satisfying these criteria were retrieved and included in the review.

Data abstraction

An abstraction form was created. Data were extracted independently on the basis of their full text by one reviewer, and verified by a second. The reviewers were not blinded to the journal or the author’s name. The accuracy of the data abstraction was randomly confirmed in 10% of cases by the initial reviewer, as well as the second reviewer. The data abstraction form included: author’s name, year of publication, investigated parameters, study population, x-ray measurement, statistical methods, outcomes and the QUADAS tool.

Methodological quality assessment

The studies included in this review were independently appraised for quality by two authors using the 14-item QUADAS appraisal tool.[26,27] It was developed to assess the diagnostic accuracy of primary diagnostic studies used in systematic reviews.[26]

Using this tool, two reviewers (MM and SS) independently assessed the methodological quality of each included paper. Any disagreement in respect of study eligibility, data extraction or methodological quality assessment was settled through discussion between the reviewers. If no agreement was reached, a third reviewer (AS) acted as an adjudicator to determine the consensus.

Study identification

An appropriate search strategy was constructed to ensure that all relevant trials published during the study period were identified (Table 1). A total of 62 citations were identified through the literature search. All full-text articles were assessed for eligibility. Twelve studies were included in the review after this procedure (Figure 1).[22-25,28-34]

Table 1.

Medline, Cochrane library and EMBASE were checked in July 2013 for appropriate material following the in- and exclusion criteria of this systematic review. The same keywords were used in all three databases.

Keywords	Medline	Cochrane	EMBASE
Rasterstereography	40	1	50
Rasterstereography OR rasterstereographic	46	1	51
Rasterstereographic AND x-ray	13	0	2
Total	99	2	103

Limits: publication date (1990-2013); “humans”; language (English, French. German).

Figure 1.

Flow of the studies through the systematic review.

Investigated parameters

The vertebral rotation was measured radiographically in nine studies out of the twelve investigated and compared with the rasterstereographic surface rotation respectively.[14,15,22-25,28-40] The lateral deviation was the second most frequent parameter mentioned in three out of four studies by Hackenberg et al. with five mentions altogether.[22-24] Kyphotic angle and lumbar lordosis were investigated triply. Torso overhang, apex height, lumbar scoliosis, thoracic scoliosis and pelvic position were recorded once each. Liljenqvist et al. investigated the highest number of parameters in their study (Table 2).[25]

Table 2.

Parameters investigated in the primary studies.

	Vertebral rotation - surface rotation	Cobb angle	Torso overhang	Pelvic balance	Lateral deviation of spine	Kyphosis angle	Lumbar lordosis angle	Thoracic scoliosis	Apex height	Lumbar scoliosis	Total
Drerup et al.[34]	1	1			1				1		4
Liljenqvist et al.[25]	1	1	1	1		1	1				6
Hackenberg et al.[30]	1										1
Hackenberg et al.[22]	1				1						2
Hackenberg et al.[23]	1				1						2
Hackenberg et al.[24]	1				1						2
Schulte et al.[29]	1										1
Schulte et al.[28]	1	1			1						3
Weiss et al.[31]						1[a]					1
Crawford et al.40							1[b]				1
Frerich et al.[33]						1	1	1		1	4
Mangone et al.[32]	1[c]										1
Total	9	3	1	1	5	3	3	1	1	1

aAccording to Stagnara;[41]

bvia modified Cobb method;[39]

cvia Raimondi method.[38]

Study population

All studies investigated patients with spinal pathologies, e.g. idiopathic scoliosis (Table 3)[14,15,22-25,28-40]

Table 3.

Study population, X-ray measurement and statistical methods

Study	Study population	X-ray measurement	Statistical methods 1
Drerup et al.[34]	113 patients with Cobb angles below 52 degrees	Anterior-posterior radiographs	RMS, standard deviation
Liljenqvist et al.[25]	95 patients with idiopathic scoliosis or scoliotic posture and 18 patients with thoracic hyperkyphosis and Scheuermann’s disease	Anterior-posterior radiography in standing posture, vertebral rotation according to Perdriolle 197937	RMS, mean
Hackenberg et al.[30]	31 patients with thoracic, thoracolumbar and lumbar idiopathic scoliosis	Anterior-posterior radiography in standing posture, sagital curvature according to Cobb 1948 andvertebral rotation according to Perdriolle 1979[37,39]	Wilcoxon sign rank test, mean, standard deviation, minimum, maximum
Hackenberg et al.[22]	25 patients with severe idiopathic scoliosis (Cobb angle of 57° on average)	Standard anterior-posterior radiographs were digitized according to Drerup[14,15]	RMS
Hackenberg et al.[23]	52 patients with thoracic, thoracolumbar und lumbar idiopathic scoliosis	Anterior-posterior radiography in standing posture, digitized according to the method of Drerup[14,15]	RMS
Hackenberg et al.[24]	25 patients with idiopathic scoliosis	Anterior-posterior radiography in relaxed standing posture, digitized according to the method of Drerup[14,15]	RMS, range
Schulte et al.[29]	43 patients with idiopathic scoliosis	Digital radiometric rotation according to Drerup anterior-posterior radiographs, relaxed standing posture[14,15]	Wilcoxon sign rank test, mean, median, minimum, maximum, standard deviation
Schulte et al.[28]	16 patients with idiopathic right convex scoliosis	Radiographs digitized in accordance with Drerup’s method[14,15]	RMS, maximum, correlation coefficient r2
Weiss et al.[31]	53 patients (26 with Scheuermann’s disease,3 with thoracolumbar kyphosis, 15 with idiopathic kyphosis.9 with kyphosis other origin)	Lateral x-ray	Mean, standard deviation, Pearson correlation, t-test.
Crawford et al.[40]	10 patients prepared for lumbar spine surgery	Relaxed clavicle position with hands placed over ipsilateral clavicles, standardized erect lateral x-ray positioning guide, lordodic angle via the modified Cobb-method using the superior endplates of L1 and S1 for reference	Nonparametric correlation coefficient Spearman’s rho, mean, standard deviation, coefficient of variations
Frerich et al.[33]	64 patients with adolescent idiopathic scoliosis (A1S). Cobb angle between 10-50 degrees	Standard anterior-posterior radiograph	Correlation coefficient, average difference, range of difference
Mangone et al.[32]	25 patients with diagnosis of A1S	Vertebral rotation (Raimondi method regolo)[33]	Kolmogorov-Smirrnov test, one way ANOVA intra-class-correlation coefficient (ICC), paired t-test, Spearman’s correlation coefficient by rank (rs)
Total	570 patients

SD, Standard deviation; RMS, root mean square.

Statistics

The root mean square (RMS) was used for data evaluation in six studies. Mean values were also found in five studies. Weiss 2008, Schulte 2008, Frerich 2012 and Magone 2013 et al. used correlation coefficients for statistical comparison of the parameters (Table 3).[28,31-33]

X-ray methods

Full-length radiographs of spines on two planes were made by Liljenqvist et al.[ and Hackenberg et al.[30] These two authors were the only ones who determined vertebral rotation using the Perdriolle method (Table 3).

Study outcomes

The studies produced different results. The highest RMS for vertebral rotation was measured at 7.9 degrees. Smaller values for this parameter were found by Hackenberg et al. and Drerup et al.[22-24,34] Crawford et al.[40] illustrated a nonsignificant association between the two methods using a t-test. Weiss et al.[31] produced a significant difference which will be discussed later (Table 4).[22-25,28-3340]

Table 4.

Comparison between x-ray measurement and rasterstereography

Study	Outcomes
Drerup et al.[34]	RMS deviation of the spinal midline (frontal) 3.6 mm (4.6 mm), RMS deviation of rotation 2,7 degrees (3.1°), SD of lateral curve amplitude 4.0 mm, SD of rotation amplitude 2.5, difference in apex height –4.6 mm, SD of Cobb angle 7.9°
Liljenqvist et al.[25]	Cobb angle: double-major scoliosis upper curvature 7.1° RMS, lower curvature 8.2°, thoracic scoliosis 7.8° RMS. thoracic and lumbar scoliosis 6.9° RMS, scoliotic malposition 7.4° RMS. Thoracic kyphosis with Scheuermann’s disease 5.6° RMS. Lumbar lordosis with Scheuermann’s disease 14.0° RMS. Torso overhang mean statistical spread 1.07 cm. Pelvic balance mean statistical spread 0.65 cm. Rotation of apical vertebra on average 7.9° RMS (depends on type of scoliosis)
Hackenberg et al.[30]	Radiography apical vertebral rotation 29.2° preoperatively, 16.7° postoperatively, rasterstereographic max. surface rotation 16.5° preoperatively, 10.8° postoperatively, RMS preoperatively 14.8C
Hackenberg et al.[22]	Average RMS difference of lateral deviation 5.8 mm, for vertebral and surface rotation curves 4.4°
Hackenberg et al.[23]	Lateral deviation RMS difference 6.4 mm preoperatively, 3.4 mm postoperatively. Vertebral rotation (VR) without correction of trunk torsion RMS difference 5.3° preoperatively, 4.1° postoperatively, with correction of trunk torsion the RMS difference 4.5° preoperatively, 3.2° postoperatively
Hackenberg et al.[24]	Lateral deviation RMS difference 5.8 mm preoperatively, 4.5 mm postoperatively. Vertebral rotation without correction of torsion RMS difference of 5.5° preoperatively, 5.2° without correction of torsion postoperatively, vertebral rotation with correction of torsion RMS 4.8° preoperatively, 4.3°with correction of torsion RMS postoperatively.
Schulte et al.[29]	Group 1: Radiography maximum thoracic vertebral rotation 12.7° preoperatively, 6.8° follow-up. Secondary lumbar curve rotation 12.7° preoperatively, 10.9° at follow-up. Rasterstereography thoracic vertebral derotation 19.5° preoperatively, 10.4° follow-up. Secondary lumbar curve 9.6° preoperatively, 4.9° follow-up.Group 2: Radiography maximum lumbar vertebral rotation 21.9° preoperatively, 9.0° follow-up. Secondary thoracic curve vertebral rotation 5.0° preoperatively, 6.5° follow-up. Rasterstereography lumbar vertebral rotation 15.5° preoperatively, 6.0° postoperatively, secondary thoracic curve rotation 8.5° preoperatively, 11.3° postoperatively.
Schulte et al.[28]	RMS differences 3.2 mm for vertebral lateral deviation and 2.5 degrees for vertebral rotation. Correlation coefficient Cobb angle radiography and lateral deviation RMS 0.7 rasterstereography and rotation RMS 0.5 rasterstereography.
Weiss et al.[31]	Average kyphosis angle radiography 49°, SD 17°, rasterstereography 63°, SD 13°, Pearson correlation of 0.78, significant difference of 14 degrees in the t test (t 9.6 P<0.001).
Crawford et al.[40]	Lumbar lordosis radiography at baseline mean 56.4°, SD 10.7° preoperatively, mean 53.5°, SD 12.1° six week follow-up. Lumbar lordosis rasterstereography baseline mean preoperatively 46.2°, SD 11.5, mean postoperatively 6-week follow up 43.1°, SD 7.6. Lumbar lordosis rasterstereography and radiography correlation preoperatively, P=0.28 (not significant), postoperatively, P=0.26 (not significant).
Frerich et al.[33]	Lumbar curve average difference 9.40 degrees, range of difference 0-22, r=0.758 Thoracic curve average difference 7.00 degrees, range of difference 0-19, r=0.872, Thoracic kyphosis average difference 10.6 degrees, range of difference 1-24, r=0.799, Lumbar lordosis average difference 8 degrees, range of difference 0-21, r=0.813
Mangone et al.[32]	Ventral rotation (VR) by x-ray mean 9.93, SD 5.38, mean by RS 4.99, SD 3.50, t-test significant (t=10.48, P<0.0001. Lumbar VR by x-ray mean ± SD 9.18, SD 3.33, by RS mean 5.52, SD 3.01. Thoracic VR by x-ray mean 10.18, SD 5.90, by RS mean 4.82, SD 3.65. r=0.52 entire spine, r=0.47, Cobb angle <30° (subgroup) r=0.42 Cobb angle >30° (subgroup) r=0.30 lumbar ventral rotation (subgroup) r=0.47 thoracic ventral rotation (subgroup)

SD, Standard deviation; RMS, root mean square.

QUADAS

Three studies scoring 12 points and two studies scoring 11 points on the QUADAS scale were evaluated; one study received four out of 14 possible points (Table 5).[22-25,28-33,40]

Discussion

Our systematic review of twelve studies evaluating the validity of RS shows that the accuracy of this method varies. Liljenqvist reported an unacceptably high root-mean-square difference for a vertebral rotation of 7.9 degrees.[25] The RMS difference of vertebral rotation comparing digitized x-rays and rasterstereography was considerably lower in other studies.[23,24,30,34] Postoperatively, Hackenberg et al. state problems when recording radiographic vertebral rotation according to the Perdriolle method.[30] A source of inaccuracy of vertebral rotation could occur due to metallic implants hiding the contours of the vertebral bodies and pedicles.[29]

In some cases the patients had an idiopathic scoliosis located in the thoracic, thoracolumbar and lumbar parts of the spinal column. The highest Cobb angle of the scoliosis amounted to 92 degrees in Hackenberg.[30] Other spinal pathologies were hyperkyphoses and Morbus Scheuermann. Healthy volunteers were not examined.

Some authors investigated more than the parameters in Table 2. But they didn’t match them with the radiographic or rasterstereographic correlate because some parameters are not directly comparable.

The statistical methods diverged. The RMS was frequently used as a statistical method. By calculating the RMS the square results were extracted; these have a smoothing effect on the absolute values.

Another problem when comparing absolute values became apparent in the study by Weiss.[31] When the rasterstereographic kyphotic angle was compared to the radiometric kyphotic angle, a correlation of 0.78 according to Pearson was found. A t-test revealed a significant difference of 14 degrees (P<0.001) between the two methods, attributed to the fact that the radiometric kyphotic angle was measured from Th4 to Th12, but the rasterstereographic kyphotic angle was measured from Thl to Th12.

The most studies scored seven or more of possible 14 points on the QUADAS scale, which means that these studies minimum have sufficient methodological quality.

However, quantitative evaluation of the validity of RS is difficult because RS is based on external measurements of the outer contour. RS will reference outer contours formed by spinous processes. Furthermore, it is influenced by overlying subcutaneous tissue that may affect the evaluation of parameters related to the internal morphology e.g. the surrounding soft tissue rotates less than the spine itself. In contrast, radiographic measures are derived directly from the internal morphology and consider variability of the spine.[32,40]

Mohokum et al. however were able to show that a higher body mass index has no influence on the reliability of the method when used on healthy test volunteers.[20]

To be able to further improve research, it is important to standardise all applied statistical methods. Useful and suitable methods, in addition to mean values, standard deviation, scattering etc., would be a continuous evaluation of the correlation coefficient (e.g. Pearson) and the RMS. Ideally, in diagnostic studies involving rasterstereography, i.e. during an evaluation if the method appropriately detects spinal disorders according to current standards, a statement about the sensitivity and specificity should be made. Questions about influential factors such as the thickness of skin folds, body weight, body height, scars, etc. should be systematically included in further study designs.

Conclusions

Rasterstereography facilitates clinical practice by examining the spinal column. Further, it is completely radiation free and could help to monitor scoliosis progression. It can be used for screening examinations as well as for follow-ups and a diagnostic method for spinal scoliosis.