Comparative effectiveness of computer optical topography and clinical photogrammetry in assessing spinal deformity in adolescents: results of a prospective study

Full Text

Introduction

An objective assessment of the severity of spinal deformity is a pressing issue in the current stage of development of pediatric traumatology and orthopedics [1]. This is due to the fact that the objectives of such an assessment are based on the basic principles of diagnostic measures in vertebrology, which can be outlined as follows:

1. The safety principle dictates the need to reduce radiation exposure during routine radiographic examinations [2].
2. The principle of high accuracy ensures timely detection of deviations in objective indicators testifying the nature of the dynamics of the pathological process [3].
3. The principle of reducing the qualification requirements for the operator of a diagnostic unit has a greater economic effect due to the subsequent reduction in costs for personnel training and the possibility of expanding the scope of research by increasing the throughput of the diagnostic unit [4].

These three pillars dictate the need to develop new diagnostic tools.

Modern tools for diagnosing spinal deformities in children [5] are represented by two main groups, each with its own shortcomings that prevent them from being fully integrated into the diagnostic and treatment process.

The first group is represented by radiological methods. This primarily includes spinal X-rays, the "gold standard" in diagnosing spinal deformities [5]. It was precisely this that gave rise to the classical nomenclature for classifying spinal deformities depending on the obtained angular values ​​ (according to Cobb and Chaklin) [6], which allow segregating specific clinical cases into groups depending on the severity of the pathological process. Multispiral X-ray computed tomography and EOS scanning (EOS Digital Radiographic System, EOS Imaging, France) are considered to be technologically close to routine radiography, which is essentially a low-dose X-ray imaging method.

The second group of diagnostic methods are optical methods based on the analysis of images of the dorsal surface of the human body with superimposed shadow "stripes" that are distorted depending on the characteristics of the deformation of this surface. The advanced diagnostic devices in this area are computer optical topography (hereinafter COT, ComOT, TODP, OOO "MESTOS", Novosibirsk) [7] and the DIERS system (DIERS International GmbH, Schlangenbad, Germany) [8; 9]. Both of these methods have demonstrated their accuracy in numerous comparative studies [3; 10], in which radiographic methods served as the standard, which fully confirms the compliance of both optical topographic and radiographic diagnostic methods with the principle of “accuracy” [11]. At the same time, the principle of “reducing qualification requirements” for operators of X-ray and optical topographic equipment cannot be fully applied, since each of them requires the participation of medical personnel with a high level of professional training [12]. The final safety principle is fully applicable to the description of optical topography, which, unlike radiography, is not associated with radiation exposure to the patient.

All these provisions, describing the current stage of development of methods for diagnosing spinal deformities, are in line with modern trends in the development of science and technology, and require the development of new – primarily digital and easily accessible – tools for diagnosing spinal deformities. Photogrammetry [13] – a method for spatial evaluation of a three-dimensional model of parts or the entire body, reconstructed from a series of photographs from different angles – could become such a promising method.

Previously, we [14] reported on the development of a new promising product – software for a personal telecommunications device (smartphone) ScolView™ for clinical photogrammetry of the torso, namely its dorsal surface. This digital product has the functionality to construct a three-dimensional image of the human body by registering fixed reference points on the surface of the three-dimensional model. The geometric characteristics of the spinal column are determined by assessing the spatial relationships of these points.

The aim of the study is to investigate the diagnostic representativeness of clinical photogrammetry ScolView™ in comparison with the results of computer optical topography.

Materials and Methods

This work was conducted as part of a monitoring study of a population of subjects – primary school students (grades 1–3, average age 8 ± 0.78 years). The total population of subjects consisted of 166 people, of which 86 were girls and 80 were boys. Prior to the study, informed consent for diagnostic intervention was obtained from all parents of the subjects. The study consisted of anonymized protocols obtained from the results of the methods described below.

All children underwent clinical photogrammetry using ScolView™. A three-dimensional image was constructed using 120 images taken from different angles. Based on the three-dimensional model, the following reference points were identified (Fig. 1): N₀ – midpoint of the posterior neck,

N₁ – left point of posterior neck,

N₂ – right point of posterior neck,

A₁ – left axillary point, A₂ – right axillary point, S₁ – point if the left shoulder joint, S₂ – point of the right shoulder joint, G₀ – apex of thoracic kyphosis, G₁ – left scapular angle, G₂ – right scapular angle, W₀ – apex of lumbar kyphosis, W₁ – left point of the waist, W₂ – right point of the waist, B₁ – left point of buttocks, B₂ – right point of buttocks, O – apex of gluteal fold.

As a result of the identification of paired reference points, the spatial characteristics of the segments formed by them in relation to the horizon were estimated: ε/ε' – angle of the buttock line, ζ/ζ' – waist line angle, ι/ι' – angle of the position of the line of the shoulder blades, ν/ν' – angle of the position of armpit line, ξ/ξ' – shoulder line angle, τ/τ' – angle of the neck line. The designation without the symbol «'» is the position of the segment in the frontal plane (skew); the designation with the symbol «'» is the position of the segment in the horizontal plane (rotation).

It is important to note that, in addition to the indicated points, ScolView™ identifies the spinal line – the N₀O curve – the corresponding line connecting the projections of the spinous processes of the vertebrae, which was stratified into 17 equal segments (hereinafter referred to as the V-segment), which were designated as V₀–V₁, V₂–V₃, V₃–V₄, V₄–V₅, V₅–V₆, V₆–V₇, V₇–V₈, V₈–V₉, V₉–V₁₀, V₁₀–V₁₁, V₁₁–V₁₂, V₁₂–V₁₃, V₁₃–V₁₄, V₁₄–V₁₅, V₁₅–V₁₆, V₁₆–V₁₇, V₁₇–V₁₈, where the location of point V₀ coincides with the position of N₀, and O – V₁₈, respectively. To assess the rotation of individual vertebrae in the horizontal plane, perpendiculars PnVn were plotted from each point V_n (where 0 ≤ n ≤ 18, n is the ordinal number of the V-segment in the rostral-caudal direction, P_n is the intersection point of the plotted perpendicular with N₀O) to the segment N₀O (V₀V₁₈). It should be noted that in this case the length of the perpendiculars laid off from V₀ and V₁₈ is zero, and V₀ and V₁₈ coincide with the location of N₀ and O, respectively. Thus, the rotation angles of each segment were calculated: in the thoracic (θn) and lumbar (ηn) arcs of the N₀O curve. It should be noted that the nomenclature correlation of the arcs to the thoracic and lumbar spine was conditional, since the segregation of the rotation angles of the V-segments was built on the principle of their belonging to the “upper” (thoracic) and “lower” (lumbar) part of the N₀O curve, divided by the point of its intersection with the N₀O segment.

 

Fig. 1. Samples of three-dimensional images of the dorsal surface of the human body generated by ScolView™, where the reference points are marked in red; a, c – right/left semi-profile view (oblique projection), respectively; b – frontal view (direct projection from behind). From here on: the color gradient (from red to blue) highlights the “height” of the points on the surface of the object (the anteroposterior size of the three-dimensional model – from larger to smaller, respectively)

 

Fig. 2. Samples of 3D images of the human dorsal body surface generated by ScolView™, where the N_O curve is indicated in green; a, b – view in the frontal and sagittal plane (view from the back and side, respectively), the symbols indicate the angles of deformation of the spinal line (hereinafter in the text)

 

An additional pool of ScolView™ output data indicating spinal deformity in the sagittal and frontal planes is the N₀O curve deformation angle values ​​in the frontal and sagittal planes (posterior and lateral views, respectively). As an example, these angles are shown in Fig. 2.

The angles identified during the analysis of the spatial distribution of the N₀O curve indicate the topographic features of the spinal line. A detailed interpretation of the designations of these angles is presented in Table 1.

The total volume of ScolView™ output data describing the spatial relationship of reference points is represented by 123 indicators, which will be named if there are reliable correlations with individual results obtained using the COT.

Computerized optical topography (COT) was used as a reference tool. This study examined baseline COT parameters.

The nonparametric Spearman's rank order (SR) test was chosen as the statistical evaluation method (Statistica 12.0) for correlations. A value of p ≤ 0.05 was considered significant. As a justification for the choice of the static evaluation method, it should be noted that the typical criteria for assessing the analytical reliability of the method turned out to be inapplicable for comparison, since the technologies of the methods used differ radically, being united solely by the subject of study – the relief of the dorsal surface of the human torso.

Results and Discussion

The results obtained in the study should be considered sequentially, in relation to the ScolView™ data, which were divided into four main groups: data on the coordinates of reference points, and data on the frontal, sagittal, and horizontal planes, respectively.

When assessing the correlations between the coordinates of individual reference points and the segments formed by paired reference points, a number of logical, reliable criterion values ​​were found.

 

Table 1. Angles describing the topographic features of the N_OO curve, identified on a 3D model of the human body using ScolView™

No.	Reference designator	Interpretation
1	γ1	The angle of lateral deviation of the spinal column in the thoracic spine in the frontal plane
2	γ2	The angle of lateral deviation of the spinal column in the lumbar spine in the frontal plane
3	χ	The angle of transition of the lateral deviation of the spinal column in the thoracic and lumbar regions
4	FB	Frontal balance is the ratio of the lateral deviation of the spinal column in the thoracic region to that in the lumbar region
5	μ	Thoracic kyphosis angle
6	λ	Lumbar lordosis angle
7	ψ	Thoracic kyphosis and lumbar lordosis transition angle
8	SB	Sagittal balance is the ratio of the thoracic kyphosis angle to the lumbar lordosis angle

Note: angles χ, ψ formed by rays of adjacent angles in the corresponding plane.

 

Thus, it turned out that the value of the “depth” of the apex of the lumbar lordosis significantly negatively correlates with the indicator of the vertical orientation of the body in the frontal plane (p = – 0.667). This is confirmed by the fact that the less pronounced the lumbar lordosis, the more vertical the subjects' body axis. Similarly, the "height" value of this same point significantly negatively correlated with the magnitude of kyphosis (p = – 0.72) and the displacement of the kyphosis/lordosis transition boundary relative to 45° (p = – 0.717). This is justified by the fact that a more rostral position of the lumbar lordosis ensures a lesser expression of the thoracic kyphosis and, in principle, shifts the topography of the transition of the thoracic kyphosis to the lumbar lordosis.

The value of the “depth” of the segment formed by the points of both armpits had a reliably positive correlation with the indicators of rib asymmetry (p = 0.748), which, in turn, ensured a reliable correlation with the data on the severity of the rotation of the pelvic girdle (p = 0.714). This phenomenon is explained by the fact that the greater value of the “depth” of the A₁A₂ segment is directly related to the asymmetry of the shoulder girdle, which could be compensated by counter-rotation of the pelvic girdle of the limbs.

The expected finding of the conducted analysis was a reliable correlation coefficient (p = 0.75) between the “height” of the segment formed by the points of the dorsal surface of the neck in its “narrowest” part and the value of the vertical orientation of the body in the sagittal plane.

At the same time, the “heights” of the segments formed by paired points of the armpits and the angles of the scapulae significantly negatively correlated with the value of the deviation of the line of the spinous processes from the midline both to the right and to the left (p = – 0.8; – 0.82 and p = – 0.694; – 0.698). This is likely due to the fact that the increase in the deviation of the spinous process line is largely due to rotation, rather than the distortion of individual points on the surface of the 3D body model. This thesis is fully supported by the significant positive correlation between the "height" of the paired segment formed by the scapular angles and the indicators of scapular angle line distortion relative to the horizontal (p = 0.683) and scapular angle rotation (p = 0.7).

In addition, the “height” of the position of the segment formed by the points of the dorsal surface of the body at the level of its “narrowest” part – between the right and left waist points, reliably positively (p = 0.667) correlated with the value of the pelvic girdle line tilt relative to the horizontal, which is quite expected due to the significant anatomical proximity of these objects. Moreover, the “length” of these same paired points had a reliable positive correlation (p = 0.767) with the value of the rotation of the shoulder girdle relative to the pelvis, while the discovery of a reliable negative correlation (p = – 0.667) between the value of the “length” of the segment formed by the “buttock points” was unexpected and at first glance had no logical explanation.

When assessing the correlations between the ScolView™ indicators testifying spinal deformation in the frontal plane (skewness of segments formed by paired points), reliable values of the Spearman criterion were also found.

The reliable correlations found in relation to the deviation indicators of the spinous process line – both to the right and to the left were the most representative in this aspect of the ScolView™ assessment. Thus, the values of the angles of skew of the line connecting the armpits and the angles of the shoulder blades were significantly positively correlated with the deviation of the line of the spinous processes to the left (p = 0.809; 0.82), and the segments connecting the points of the shoulder joints and points N₁ and N₂, respectively, were correlated with the deviation of the line of the spinous processes to the right (p = 0.749; 0.822). Such connections can be interpreted by the occurrence of asymmetry of the limb girdles during the development of three-plane deformation of the spine, which is fully confirmed by the presence of a reliable negative correlation (p = – 0.683) between the pelvic girdle rotation indicator and the value of the distortion of the segment connecting the points of the shoulder joints.

The most informative from the point of view of the subject of the study were the correlations of the values indicating the deformation of the spinal line – the N₀O curve in the frontal and sagittal plane.

It turned out that the value of the angle of lateral deviation of the spinal column in the thoracic region in the frontal plane was significantly positively correlated (p = 0.7) with the value of the costal and muscular asymmetry index, which is logically explained by the common origin of these values. At the same time, the value of the angle of lateral deviation of the spinal column in the lumbar region in the frontal plane had a reliable negative correlation (p = – 0.805) with the magnitude of the tilt of the line of the angles of the shoulder blades relative to the horizontal plane. This is due to the counter torsion of the torso, which logically explains the significant positive value of the Spearman criterion (p = 0.731) found between the magnitude of the tilt of the scapula angles and the value of the frontal balance. This same value-the frontal balance coefficient-had a significant positive correlation (p = 0.757) with the vertical orientation of the torso (in the frontal plane) and a negative correlation (p = – 0.74) with the magnitude of the rotation of the scapula angles. All this confirms the importance of the coefficients of the ratios of the arc sizes as an integrative indicator of spinal deformity.

When assessing the parameters of the spine line in the sagittal plane, reliable negative correlations were found (p = – 0.767; – 0.683) between the magnitude of thoracic kyphosis and the vertical orientation of the torso - both in the frontal and sagittal planes, which is explained by the expected more significant verticalization of the torso with less pronounced thoracic kyphosis. A rather unusual finding was a significant negative correlation (p = – 0.75) between the angle of transition from thoracic kyphosis to lumbar lordosis and the pelvic girdle rotation index, which may partly be due to the worsening of pelvic girdle rotation with increasing scoliosis severity. The most obvious finding, which does not require logical confirmation, was a significant positive correlation (p = 0.75) between the values of lumbar lordosis both when using ScolView™ and when assessing the results of the COT. Another indicator that confirmed its integral function was the value of sagittal balance, which had reliable negative correlations with the values of the vertical orientation of the trunk in the frontal plane (p = – 0.833), the value of thoracic kyphosis (p = – 0.783) and the displacement of the “kyphosis/lordosis” transition boundary relative to 45° (p = – 0.783).

When assessing the correlations of ScolView™ output data indicating a violation of the geometry of a three-dimensional object in the horizontal plane, it is advisable to consistently describe the relationships between individual features that are not anatomically clearly associated with the spinal column and indicators directly related to the structure of the spinal column.

Thus, the first group of such indicators includes the value of the rotation of the buttock line, which revealed a reliable positive correlation (p = 0.694) with the magnitude of the deviation of the line of the spinous processes from the midline to the right. At the same time, the rotation angle of the scapular line was significantly positively correlated (p = 0.883) with the values of the pelvic girdle line tilt relative to the horizontal. In turn, the rotation angle of the neck line had a reliable positive correlation with the indicators of the vertical orientation of the torso in the frontal plane (p = 0.867) and the value of the tilt of the line of the angles of the scapula relative to the horizontal (p = 0.75), and a reliable negative correlation with the magnitude of the rotation of the angles of the scapula (p = – 0.667). Thus, these correlation interactions describe the natural phenomena of the combined violation of the harmonious geometry of the relief of the surface of the back, which at the same time are a manifestation of balanced biomechanical phenomena that support the vertical orientation of the body.

The second group of indicators are the rotation angles of the V-segments and the absolute values of the length of the perpendiculars PnVn, which directly indicate rotational changes in the spinal column and require closer attention to their own reliable correlations with the output data of the COT. Thus, the value of the rotation angle η_10–12 had a reliable positive correlation with the magnitude of the deviation of the line of the spinous processes to the left (p = 0.822; 0.73; 0.73), while the η6 indicator was reliably positively correlated with the value of the vertical orientation of the body (p = 0.667). In turn, the angle θ₉ demonstrated a reliable positive correlation (p = 0.782; 0.782) with the values of the shoulder girdle tilt of the extremities, including the line of the angles of the shoulder blades relative to the horizontal. At the same time, the angle θ₁₀ had a reliable positive correlation (p = 0.678) with the indicator of the vertical orientation of the body in the frontal plane, and the values of the angles θ_13–14 – like a number of angles η earlier – with the indicator of the deviation of the line of the spinous processes to the left (p = 0.833; 0.8), and the angle θ₁₃, in addition, with a similar indicator “to the right” (p = 0.767). All these correlations indicate a clear connection between the logically close indicators of the COT and ScolView™, pointing at the deformation of the projection of the spinal column on a three-dimensional image of the object of study.

Like the values of the V-segment rotation angles, the values of the P_nV_n perpendiculars had a sufficient volume of correlations with the COT indicators, which indicate the representativeness of ScolView™ in identifying spinal deformities. The values of the length of the perpendiculars P₁V₁, P₄V₄–P₁₀V₁₀ were the most significant from this position, which had a reliable correlation with the magnitude of the skew of the line of the angles of the shoulder blades relative to the horizontal plane (p = 0.75; 0.75; 0.833; 0.717; 0.817; 0.82; 0.8; 0.733). It is logically explained by the worsening of the asymmetry of the upper limb girdle with an increase in the magnitude of the scoliotic arc. The detected reliable correlation between the length of P₉V₉ and the index of vertical orientation of the body in the frontal plane (p = 0.667) is also explained by this. From the standpoint of the mathematical origin of the absolute value of the perpendiculars P_nV_n, reliable correlations between the values of thoracic kyphosis (p = 0.71; 0.753; 0.793; 0.833) and the displacement of the kyphosis/lordosis transition boundary relative to 45° (p = 0.7; 0.733; 0.744; 0.835)with the dimensions of the perpendiculars P₄V₄–P₅V₅, P₉V₉–P₁₀V₁₀ turned out to be logical, since an increase in the length of these dimensions is logically associated with an increase in the severity of thoracic kyphosis. Therefore, the value of the lumbar lordosis had a reliable positive correlation (p = 0.767) with the length of the perpendicular P₁₄V₁₄, and a negative one (p = – 0.7) with the same value of P₆V₆. The final fact in the description of the correlation links between the indicators of spinal column deformation in the horizontal plane was the discovery of another negative correlation link (p = – 0.167) between anatomically close indicators – the rotation of the angles of the scapula and the length of the perpendicular P₆V₆. These data indicate a less pronounced shoulder girdle rotation in subjects with flattened thoracic kyphosis.

Accumulated global experience with photogrammetry in terms of its practical use in assessing spinal deformities does not fully allow the results obtained in this study to be extrapolated to the work of other research groups. This is due to the fact that their findings are relatively fragmentary and describe isolated milestones in the violation of the geometry of the spinal column from the perspective of the modern theory of the etiopathogenesis [15] of scoliosis as a multiplanar spinal deformity. All modern results in the development of the problem of photogrammetric assessment of spinal deformity can be summarized by the outstanding works of Brazilian researchers. Thus, according to the results of a meta-analysis carried out by employees of the University of Rio Grande do Sul (UFRGS, Rio Grande, Brazil), it was shown that the photogrammetry method demonstrates the most valid values of the Cobb angle and spinal rotation indicators in relation to the radiographic method of research. The only drawback of this fact was that this assessment was carried out exclusively when studying the parameters of the cervical spine [16]. The same group of scientists assessed the sensitivity and specificity of photogrammetry in the analysis of torsion of the torso, and the sensitivity and specificity of the method were confirmed using a scoliometer. Despite the relatively high assessment results (the sensitivity and specificity of the photogrammetric method for assessing torsion were higher than 83 and 78 %, respectively) [17], a drawback of the study was its relative inaccuracy. This is due to the fact that the subject of the study was costal gibbus, which is largely formed due to deformation of the costal part of the rib cage [18].

All this allows us to interpret the obtained results of ScolView™ as pioneering. However, this complicates the extrapolation of the obtained data to the sphere of practical medicine in the absence of a reliable tool for assessing validity. At the same time, based on the large-scale correlation analysis conducted, it can be confirmed that the output data of ScolView™ are truly representative. This is confirmed by multiple correlations with individual COT results. Therefore, to systematize the obtained data, it seems appropriate to evaluate the obtained correlations in terms of their segregation relative to the three planes of spinal deformity.

Thus, all the obtained reliable values ​​of the Spearman criterion were divided into groups (Table 2), and in each of the groups, its average value and the number of reliable values ​​of this criterion were estimated based on the belonging of the indicators to one or another plane. It should be noted that such a maneuver was carried out solely to ensure clarity (Fig. 3); the information from Table 2 does not include data on specific indicators-of the spatial location of reference points and segments formed by paired reference points, since these are native, and their coordinates provide the emergence of other integrative indicators.

 

Table 2. General information on the Spearman S_R, criteria obtained from the correlation analysis of the output data of the COT and ScolView™

SV/COT plane index	Plane
	frontal			sagittal			horizontal
	M	m	n	M	m	n	M	m	n
Frontal	0.770	0.043	9	-	-	0	0.712	0.040	2
Sagittal	0.800	0.047	2	0.750	0.047	4	0.750	-	1
Horizontal	0.773	0.059	22	0.749	0.055	11	0.756	0.102	3

Note: M is the mean value; m is the root-mean-square deviation; n is the number of reliable values of the Spearman criterion in the group.

 

Fig. 3. Mean values and number of reliable values of the Spearman S_R  criterion in relation to the output data of the COT and ScolView™

 

Summarizing the work carried out, it should be noted that the average value of the Spearman criterion in assessing correlation relationships was 0.758 ± 0.025, which is very encouraging in the prospect of widespread implementation of the photogrammetry method in routine monitoring of the dynamics of spinal deformation in population groups.

Conclusions

Photogrammetry is a relatively new method in clinical medicine, which is only just beginning to be integrated into clinical practice. The most promising area of ​​clinical medicine for the use of photogrammetry is traumatology and orthopedics, specifically pediatric vertebrology. Specialists in this field should be interested in the development of new and valid tools for diagnosing spinal deformities. The developed ScolView™ software can become such a tool, which, however, requires a large-scale prospective study with segregation of subjects into comparison groups and identification of specific clinical signs, followed by the preparation of practical recommendations for the use of photogrammetry in clinical practice.

About the authors

Ivan D. Shitoev

Perm National Research Polytechnic University; Yord Tech

Email: ShitoevID@yord.tech
ORCID iD: 0000-0002-6391-9271

Assistant of the Department of Computational Mathematics, Mechanics, Biomechanics, CEO

Russian Federation, Perm; Perm

Sergey V. Muravyev

Ye.A. Vagner Perm State Medical University

Email: sergey89.m@mail.ru
ORCID iD: 0000-0002-3342-4710

PhD (Medicine), Associate Professor of the Department of Physical and Rehabilitation Medicine with a Course in Medical and Social Expertise

Russian Federation, Perm

Yulia V. Karakulova

Ye.A. Vagner Perm State Medical University

Email: pgmu-kafnerv@mail.ru
ORCID iD: 0000-0002-7536-2060

DSc (Medicine), Professor, Head of the Department of Neurology and Medical Genetics

Russian Federation, Perm

Gayane Z. Kloyan

Yord Tech; Ye.A. Vagner Perm State Medical University

Author for correspondence.
Email: kloyang@mail.ru
ORCID iD: 0000-0001-6615-8159

Biomechanist, Lecturer of the Department of Medical Informatics and Management in Medical Systems

Russian Federation, Perm; Perm

Pavel N. Chaynikov

Ye.A. Vagner Perm State Medical University

Email: chainikov.p.n@gmail.com
ORCID iD: 0000-0002-3158-2969

PhD (Medicine), Associate Professor, Head of the Department of Medical Rehabilitation, Sports Medicine, Physical Culture and Health

Russian Federation, Perm

Maxim A. Kovalev

Ye.A. Vagner Perm State Medical University

Email: maksim-k99@list.ru
ORCID iD: 0000-0003-2873-1553

Resident of the Department of Neurology and Medical Genetics

Russian Federation, Perm

References

Supplementary files