Scoliosis is characterized by three-dimensional deformities of the spine, showing a lateral curvature of the spine of at least 10° and axial rotation around the vertical body axis. This inside deformation has an impact on the outside torso contour. For the diagnosis of scoliosis, in addition to clinical examinations an X-ray is required at the rst consultation with an orthopaedic specialist. In case of fast progression of scoliosis specially during growing stage, regular clinical monitoring is required, including follow-up X-rays. As a consequence of frequent Xrays, patients are often exposed to signi cant ionizing radiation and have a higher risk of radiation-related health problems. To reduce the number of X-rays a 3D body scanner is proposed as an ionizing radiation-free method, supplementary to clinical examinations, for assessing scoliosis and its progression. A mathematical method is proposed in combination with a self developed model of rib cage and vertebral column to analyze the spinal curvature from the image obtained from 3D body scanner. Here we present the idea, challenges and possible solutions at the example scan from a healthy person. In future, the proposed methods will be transferred and applied to patients with di erent kinds of scoliosis, as an alternative for X-rays.
Scoliosis is a medical condition characterized by three-dimensional (3D) spinal
deformities, showing at least 10° of lateral curvature in the coronal plane and
a vertebral rotation [
Adolescent scoliosis, either idiopathic or neurogenic, is one of the most
common pediatric diseases, which has a higher risk of progression and aggravation
of the scoliotic deformities. The severity of deformity is usually assessed by
radiography examinations, and quanti ed by the Cobb angle [
For the same purpose, a 3D body scanner has been developed with a camera
system [
For this purpose, a basic skeletal structure of the essential components of the
ribcage and vertebral column have been modeled with computer aided design
(CAD) software [
Even though the method presented here based on the scan image of a healthy person and the bony model is presented for a normal con guration of human spine, it is intended in the future to generate patient speci c spinal con gurations by means of simulations. Generally, an X-ray image is essential, as a standard of treatment for the rst appointment with an orthopaedic specialist, in order to con rm the suspected diagnosis after clinical examinations and to verify the severity of scoliosis to start appropriate treatment. In perspective of future clinical applications, the rst time taken X-ray will be used in order to get the initial con guration of the vertebral column for the simulation of the bony model. On the same day, a body scan image will be captured. Then the scan image will be oriented according to the quantities introduced here to obtain the best match with the bony model and put them together in the same reference frame. Later in the follow-up examination, in addition to clinical examinations, a scan image will be captured and natural growth will be noted in case of young patients. In order to account the patient's body change due to natural growth, initially obtained bony model will be scaled along the principle axes and compared with the scan image. Thus the model will replace the X-ray image in the follow-up examination. If there is a mismatch between the model and the follow-up scan image, the model will be simulated to obtain a good match with the scan image in di erent planes. The new approximated vertebral body positions from the bony model will then give the spinal curve normally obtained from X-rays. Thus one has the possibility to reduce number of follow-up X-rays. 2 2.1
Here we analyzed a torso image of a 25 years old healthy male, obtained using
the 3D body scanner. At the beginning, parts not belonging to the torso contour,
i.e. the portion of head, arms, were removed from the image. Only the torso part
of the image was extracted for analysis in terms of 3D Cartesian coordinates
(see Fig. 1(a)). The geometry of a bony model of rib cage and vertebral column
of an average human, based on real geometry and anatomical parameters
[
The dimensions of the bony model for a male and the extracted region of the
scanned image were then normalized along the principle axis in between 0 and 1
(see Fig. 1). 2D cross sections transverse to the vertical body axis were extracted
from the model and from the scan image. Their contours or the outlines of this
2D transverse cross sections were presented in terms of Cartesian coordinates
(see Fig. 2). The positions of the spinous processes (marked as `SP' in Fig. 2
(upper panel)), were located from the scanned image, by nding the dip in the
back part of the 2D cross sections transverse to the vertical body axis. A self
developed analysis tool extracted the 3D coordinates of the spinous processes
positions following the automated algorithm described in [
Sh1 Sc1 Pl1
Sh2
T1
T2
T3
T4
T5
T6
T7
TT89 Sc2
T10
T11
T12
L1
L2
L3
L4
L5 Pl2
(c)
Fig. 1. (a) 3D image of a healthy male obtained using the 3D body scanner. (b) The
bony model of ribcage and vertebral column of an adult male, based on anatomical
geometry and dimension parameters after [
In addition to spinous processes positions left and right extreme points on the shoulder level (Sh1, Sh2), on the scapulas level (Sc1, Sc2) and on the pelvic level (Pl1, Pl2) were manually extracted from both the model and the scanned image. For the model, T1, T8 and L5 vertebral levels were assumed to be at shoulder, scapulas and pelvic levels respectively. These points are also marked in
T2 0.7 LRASM = 0.5875
ASR = 0.022755
The principle axes of the Cartesian coordinate system (X,Y,Z in Fig. 1 (a) and 1(b)) of the bony model are parallel to the normals of the main anatomical planes of the human body. That is, any anatomical position of the bony model can be described with respect to these three planes: { the transverse plane parallel to the ground separates the body into top and bottom halves { the coronal plane perpendicular to the ground separates the front (anterior) from the back (posterior) { the sagittal plane is perpendicular to the transverse and coronal and separates the left from the right.
For the scan image the situation is more di cult. Since the camera of the scanner system needs to be adjusted according to the individual height and standing position of each patient, the viewing axis and thus the camera angle relative to the patient is di erent for all scans at all times. The scan image thus always has an oblique coordinate system, with di erent obliquity at each and every time. Therefore it is important to have the exact angular information about the obtained 3D torso image initially for future comparison.
The 3D coordinates of the marker positions were used by the analysis program to identify and check the orientation of the scanned torso in relation to the laboratory coordinate system and to compute the parameters described in the next section. 2.2
The analysis parameters introduced here were computed independently for the scan image and for the bony model. These parameters were designed to identify and quantify the characteristics of scoliosis from the scan image with the help of bony model. They were intentionally de ned in relation to the three main body planes and thus independent of the coordinate system. Further, thereby the impact of growth and posture is reduced, which is important for future applications to adolescents. The parameters introduced here are not very typical in medical literature, but the way the tool is designed has the possibility to evaluate the standard parameters used in medical literature e.g. thoracic kyphosis, lumbar lordosis.
{ Extraction of speci c lines from the scan image and from the bony
model:
(a) Vertebrae line or principal axis: The vertebrae line or principal axis was
de ned as the straight line between the positions of the spinous processes
at the levels T1 and L5 for the scanned image and accordingly between
the corresponding vertebral bodies at the bony model.
(b) Vertical and horizontal line: A vertical line parallel to z axis and a
horizontal line parallel to x axis were de ned at positions T1 and L5. In
principle, at any marker position these lines can be created.
(c) Shoulder, Scapulas and pelvic line: Straight lines were created at shoulder
(Sh1{Sh2) , scapulas (Sc1 {Sc2) and pelvic levels (Pl1{Pl2), connecting
left and right points.
{ De nition of the parameters:
(a) Torso inclination angle: Angle between the vertebra line and the
vertical axis containing T1 was de ned as torso inclination angle. This is a
parameter in sagittal plane. In principle, it would be possible to divide
the sagittal plane by thoracic and lumbar part.
(b) Pelvic obliquity: Angle between the pelvic line (Pl1{Pl2) and the
horizontal line at L5 was de ned as pelvic obliquity. This is a parameter in
the frontal plane.
(c) Shoulder obliquity: Angle between the shoulder line (Sh1{Sh2) and the
horizontal line at T1 was de ned as shoulder obliquity. This one is also
a parameter in the coronal plane.
(d) Pelvic Rotation: Angle between the pelvic line (Pl1{Pl2) and scapulas
line (Sc1 {Sc2) was de ned as pelvic rotation. This is a parameter in
transverse plane.
(e) Left-right area asymmetry: The regions spanned by the contour of the
transverse cross sections were divided by left and right area following
the similar approach published [
L R
L + R
;
where L and R are the areas on the left and right sides to the spinous
processes positions respectively of the transverse cross sections.
(f) Aspect ratio di erence: This parameter computes the di erence between
the height and width ratios at two sides of the transverse cross sections
following the same approach described in [
ASR = abs
Right side width
Left side width : The method was developed by using Matlab2019a (The MathWorks, Inc., Natick, MA, USA) and the meshlab software. 3
Figure 1(a) shows the scan image from a healthy person and Figure 1(b) shows the bony model for an average male. In Figure 1(b) the normals of the main anatomical planes of the human body are parallel to the principle axes of the lab coordinate system. In case of the scanned image (see Fig. 1(a)) the normals of the main anatomical planes however are not parallel to the principal axes of the lab coordinate system.
Table 1 lists the corresponding angular parameters, computed for the case of the bony model and the scanned image. For the bony model, the value of the torso inclination angle re ects it's natural S-type shape of the course of the vertebral column in the sagittal plane. On the other hand, the S type nature of the course of the vertebral column for the case of scan image can be detected by the torso inclination angle ( 9°). Quite a good matching of this value indicates this parameter might be a good choice to capture the change in scoliosis in the sagittal plane. However, the larger values of pelvic obliquity and shoulder obliquity of the scan image imply that the image's anatomical plane is obliquely placed with respect to the lab coordinate system. This is the main di culty of the presented method to separate the e ects of body change due to the change in scoliosis from the e ects due to the scanning procedure (di erent position of the patient, camera etc.). Therefore, for future comparison, either the follow-up image should have the same orientation with the initial one or our analysis system needs to set a xed reference frame. For the rst option, more automation of the present program, specially automatic detection of shoulder points, pelvic points and scapulas points, will help, so that the obtained image can be simultaneously checked and patient position can be adjusted to reproduce the orientation of previously obtained image. For the other option, the analysis tool can transform the basis of the coordinate system into the bony model's basis. Either one of these is necessary to capture the change of torso shape due to scoliosis, which is our end aim.
Figure 2 illustrates transverse body contours, extracted at T2, T8 and T10 levels of the vertebral column. The upper panel shows the 2D transverse cuts from the scan image and the lower panel shows the same for the model. The corresponding left-right asymmetry (LRASM) and aspect ratio di erence (ASR) parameters of the cross sections extracted from the scan image are given in the annotations. The positions of spinous processes and approximated vertebral body positions are depicted with the black open circles. Qualitatively there is a change in the shape of the contours, from the scanned image from the left to the right panel. These changes in shape are re ected in a change in the values of the parameters: the LRASM parameters decrease continuously from left to right. In contrast the aspect ratio di erence asymmetry increases from the left to the center panel and decreases to the right panel. The equivalent values of the symmetry parameters can not be calculated for the transverse cross sections of the model (lower panel), because the scattered points are not continuous and not equally distributed at all levels. Proper tting function for shape matching, at this level, will help to calculate the above mentioned symmetry parameters from the bony model.
Our previous study supports the concept of these two symmetry parameters
by nding quite a good correlation with the apex position of scoliosis, when
comparing with computer tomography data set [
The current, as well as other non-invasive methods, however, cannot completely replace the conventional and established methods (e.g. X-ray, MRI) in the medical assessment of scoliosis, specially when surgical decisions are involved. Further X-rays might be necessary when signi cant changes in scoliosis are recognized clinically and also in the method presented here. In principle, the idea introduced here can be implemented to get a large number of quantities or indices along the vertebral column. Therefore there might be a good chance to nd a well correlated index with the standard method like Cobb angle. An implementation and development of such methods requires, however, analysis of a larger set of data. The idea here described constitutes nevertheless a preliminary step
Future work in this direction. In general, the present scanner system might have the potential to reduce the number of X-rays in follow-up examinations, which is needed for adolescents with scoliosis.
{ To reduce the e ect of camera position and di erent standing positions of the patients, more automation of the presented method will be considered and several experimental measurements with healthy participants without scoliosis will be done to calibrate the scanning procedure against external perturbations, like the e ect due to the position of the camera, position and posture of the patient etc. These steps will be considered in our future work. { Finding an algorithm for shape matching for the 2D transverse cross sections from the bony model and from the scan image will also be considered in future work. { Simulation of the bony model following nite element method simulation will be done in future to get the distorted con guration of vertebral column.