Secondary Logo

Journal Logo


Morphometric Analysis of Anatomic Scoliotic Specimens

Parent, Stefan MD*†‡; Labelle, Hubert MD*†; Skalli, Wafa PhD; Latimer, Bruce PhD§; de Guise, Jacques PhD*†∥

Author Information
  • Free


Idiopathic scoliosis is a three-dimensional deformity affecting the orientation and position of the spinal elements in space. The regional and global changes are characterized by a deviation in the frontal plane, a modification of the sagittal profile, and modifications in the shape of the rib cage. Locally, the vertebral shape is also affected by complex tissue alterations, including vertebral wedging, neural arch deformation, transverse process orientation, and vertebral torsion. 12 The study of these local descriptors is hindered by the imaging techniques used in clinical practice. Plain radiographs provide good clinical information about the regional deformity but are poor indicators of the local vertebral changes. Although vertebral body and disc wedging can be relatively well assessed using plain radiographs, 14,17 the posterior element definition remains suboptimal. Computed tomography scan imaging provides higher-quality images and a better definition of the vertebral modifications, but the high levels of radiation to which patients are exposed preclude its routine clinical use to monitor the progression of scoliotic curves.

A review of the literature provides limited information about the local changes occurring in idiopathic scoliosis. Most reports are made on a few isolated anatomic specimens with major curvatures, and most of the knowledge we currently have regarding the local changes is based on these observational studies. Several representations of scoliotic vertebrae exist in the literature. However, often these are artistic representations of isolated cases and may therefore not represent accurately the local changes occurring at the vertebral level. 3–5,13 There is also a lot of controversy in the literature regarding isolated observations made by different authors or, in some cases, the same author. 2 In an attempt to quantify changes occurring at the level of the pedicles, Liljenqvist et al reported that pedicle size is significantly smaller in thoracic pedicles located on the concavity of thoracic scoliotic curves, 7 but the study was limited only to this aspect of scoliotic vertebrae.

The objective of this study was to characterize the local vertebral changes seen in idiopathic scoliosis. We postulate that there is a typical deformation pattern that describes the local changes at the vertebral level based on vertebral wedging, pedicle orientation and dimension, orientation of the posterior elements, localization of the curve (i.e., thoracic or lumbar), and the localization of the vertebral body (apical or transitional vertebra).

Materials and Methods

Thirty scoliotic specimens were identified from two major osteologic collections. The first 15 specimens are part of the Hamann-Todd Osteological Collection found at the Cleveland Museum of Natural History. This collection contains >3000 complete skeletal specimens including postcranial material. The other 15 specimens are part of the Robert J. Terry Collection found at the National Museum of Natural History in Washington, DC. This collection contains >1700 specimens. Scoliotic specimens were selected by assembling the thoracic and lumbar segments and observing their natural configuration and retaining those that presented typical characteristics of thoracic idiopathic scoliosis, namely, vertebral body wedging at the midthoracic apex and a rotational deformity of the curve. Vertebral wedging produced these configuration changes. Examination of the specimens did not include intervertebral disc changes because these had been resected at the time of dissection. Cobb angle and the exact degree of deformity were therefore not available. All of the 4700+ specimens were screened for scoliosis, and any specimen with evidence of a congenital deformity was excluded. All 30 scoliotic specimens were then matched for age, sex, race, height, and weight with a normal specimen from the corresponding collection. Both collections keep an extensive record on each specimen, including postmortem photographs, pathology, and dissection reports as well as anthropometric measurements taken at the time of dissection. All these parameters were reviewed and incorporated in our analysis. Radiographs are also available for selected specimens showing severe spinal deformities.

The following step was to obtain an accurate three-dimensional model of each thoracic and lumbar vertebra of the 30 scoliotic specimens and of the 30 age- and sex-matched normal specimens. This three-dimensional model was obtained by manual digitization. The instrument used to digitize the anatomic specimens was the Fastrack (Polhemus, VT). This device works by generating near, low frequency, magnetic field vectors from a single assembly of three co-located antennas that are fixed in space working as the transmitter. These magnetic field vectors are then detected by a set of co-located antennas remotely placed (stylus), acting as the receiver. This receiver is linked to a computer that transforms the received signal using a mathematical algorithm creating a set of three coordinates indicating the exact position of the receiver in relation to the transmitter referential.

The digitizing protocol was developed by our research group 6,16 and consisted of measuring specific anatomic landmarks on each specimen, thus creating a set of approximately 200 points depending on the vertebral level measured with regard to its particular geometry (190 points for lumbar vertebrae and 216 points for thoracic vertebrae). Each point was always located at the same landmark and points were taken in a specific order to diminish measurement errors. A PVC frame, to which the magnetic source was fixed, prevented each vertebra from moving during the measuring procedure by maintaining them rigidly (Figure 1). The source was connected to the Fastrack module, which calculated the exact three-dimensional position of the stylus (receiver) and relayed this information to the computer. The stylus’ position was calculated by pressing on the mouse button, thus reducing micromovement errors of the receiver as opposed to using the stylus’ button. The different points were acquired in a specific order and recorded in this sequence in a computer file.

Figure 1
Figure 1:
PVC frame with vertebra held in place. The source is rigidly fixed to the PVC frame reducing measurement errors.

After the measurements were done, each vertebra was then reconstructed using computer graphics software while being still held in place in the PVC frame (Figure 2). This provides an excellent quality control, virtually eliminating any gross measurement error. Modifications to the dataset can be made immediately after visual inspection of the three-dimensional reconstruction while the vertebra is still held in the PVC frame, ensuring a high degree of reliability. 16

Figure 2
Figure 2:
Normal lumbar vertebrae (top) with reconstruction (bottom left) and both images superposed (bottom right).

Each set of points was then relocalized in a local coordinate system. This is a crucial step to allow for comparison between different specimens, especially when comparing different angles calculated in the local coordinate system. Every dataset underwent a mathematical modification based on a specific computer algorithm. The center of the vertebral body was defined as described hereafter. The origin is located at mid-distance between the two planes formed by the two vertebral endplates. These planes were fitted to a three-dimensional plane using the sum of least squares method from the set of points representing each endplate. The origin is located at two-thirds of the vertebral depth from the anterior in the sagittal plane. A computer program defined an algorithm to transform these points to their own anatomic local coordinate system with the center of the vertebral body being relocated at the intersection of three planes. The local coordinate system used was the one adopted by the Scoliosis Research Society, defining the frontal plane as Y0Z, the sagittal plane as X0Z, and the horizontal plane as X0Y. This transformation was necessary to allow comparison of different parameters for different vertebrae.

Computer software was then developed to calculate specific parameters from these sets of points. Each parameter represents a measure of length, width, height, surface area, or a specific angle. A total of 127 parameters were developed for lumbar vertebrae and 145 parameters for thoracic vertebrae.

A repeatability study done by two separate observers on 28 vertebrae using the digitizing protocol showed a mean measurement error of 0.5 mm, and reproducibility of the end results (parameters) as measured by the intraclass correlation coefficient was found to be >0.95. 11


Thirty scoliotic specimens and 30 normal specimens were measured from both osteologic sources. The mean age and sex distribution as well as the total number of vertebrae can be found in Table 1. The relatively high male-to-female ratio results from the higher proportion of males in each collection. This can be explained by the fact that the two collections were assembled with unclaimed bodies between 1920 and 1940, which were usually male bodies. Table 2 presents the type of curves according to King’s classification. As seen in Table 2, the scoliotic specimens were mostly right thoracic curves. The discrepancy between the number of measured vertebrae from scoliotic and normal spines is the result of the fact that some vertebrae were fused in the scoliotic specimens.

Table 1
Table 1:
Sex Distribution and Mean Age of Scoliotic and Normal Specimens
Table 2
Table 2:
Distribution of Specimens According to King’s Classification

Vertebral wedging measured in the coronal plane was significantly greater for the scoliotic specimens in the midthoracic area on the concave side of a typical right thoracic curve. Vertebral wedging in the coronal plane was the projection of the angle formed by the two planes representing the endplates as measured in the coronal plane (90° projection). Vertebral wedging was also present in significant amounts on the compensatory curve located in the high thoracic area on the convex side of a typical right thoracic curve (Figure 3).

Figure 3
Figure 3:
A: Wedging between the superior and inferior endplates seen in the frontal plane for normal and scoliotic thoracic vertebrae. B: Image of a typical scoliotic thoracic vertebra (T5) demonstrating vertebral wedging at the apex of the curve.

Minimal pedicle width represented the distance between two points taken at the equatorial line of a midsection of a pedicle. Left pedicle width was found to be significantly diminished on the concavity of a typical right thoracic curve. Apical vertebrae were most significantly altered with mean differences ranging from 1.03 to 1.37 mm between the concave side of scoliotic vertebrae when compared with the corresponding side of normal thoracic vertebrae from T6 and T12 with a maximal mean difference of 1.37 mm at T8 (95% confidence interval 0.80–1.95) (Figure 4). Right pedicle width was also significantly diminished on the concavity of the high thoracic compensatory curve with mean differences ranging from 0.68 to 1.68 mm and a maximal difference of 1.68 mm at T4 (95% confidence interval 1.08–2.28) (Figure 5). Changes in the thoracolumbar or lumbar compensatory curve were more subtle but were significant for T10, T12, and L2 with mean differences ranging from 0.9 to 1.18 mm.

Figure 4
Figure 4:
A: Mean left pedicular width distribution and mean age of scoliotic and normal specimens. B: Typical thoracic vertebra presenting a characteristic deformation of the pedicle located on the concavity of a scoliotic curve.
Figure 5
Figure 5:
A: Mean right pedicular width. B: Right pedicle width shown to be markedly decreased on concavity of a superiorly adjacent thoracic curve.

Orientation of the maximal deformation was found by calculating the intersection of the two vertebral endplate planes. The orientation was then evaluated in the transverse plane as seen from above by using a 360° scale, starting at the center portion of the vertebral posterior wall and increasing clockwise all the way around the vertebral endplate (Figure 6A). Figure 6B shows that the orientation of maximal deformation was mostly found in the frontal plane.

Figure 6
Figure 6:
A: Orientation of maximal deformity. B: Orientation of maximal deformity in the transverse plane for normal and scoliotic vertebrae.

Vertebral heights taken at different points around the vertebrae were also found to be statistically different when compared with normal vertebrae, especially in areas where vertebral wedging was predominant. Facet surface was measured by adding the surface of eight triangles forming the three-dimensional representation of the facet (Figure 7). The facet surfaces were found to vary around the apex of the curve showing no correlation with the convexity or concavity (Figure 7). The remaining parameters that were found to be different statistically are presented in Table 3.

Figure 7
Figure 7:
Superior facets as seen from the back. Note the characteristic surface changes.
Table 3
Table 3:
List of Remaining Parameters With Statistically Significant Values (P < 0.05)


The study of idiopathic scoliosis and its three-dimensional regional deformity is not possible without a good understanding of the local vertebral changes. The objective of this study was to establish a deformity pattern for idiopathic scoliosis. Normal vertebral anatomy has been described in detail by several authors, 1,8–10,15 and recent work has focused on identifying characteristic changes occurring in pedicle size of patients with adolescent idiopathic scoliosis. 7 Although several authors have reported isolated observations made on rare anatomic specimens, this is the first time that such a large number of specimens have been studied so extensively.

Most of the studies on the normal morphology available in the literature used calipers and goniometers to measure different parameters on thoracic and lumbar vertebrae. 1,15 Panjabi et al used a three-dimensional morphometer to record the three-dimensional coordinates of a number of points of 12 complete specimens, calculating 64 different parameters from these sets of points. 9 This method was described as being more precise and more detailed, but again, the number of specimens was limited. Furthermore, the technique required identifying the position of the object before digitizing each point, making this a very tedious process. The Fastrack being internally calibrated and the source acting as the reference to which the exact position can be calculated make this solution much more precise and easier to use. Previous studies made on normal specimens were consistent with these findings. 6,16

The specimens used for this study were all part of two major osteologic sources. All specimens had complete dissection records as well as predissection photographs. These specimens are a rare source of anatomic specimens and provide an invaluable source of information. The relative drawback of using these specimens is that they represent a fixed deformity in time and provide relatively little information about the evolution of the disease. Furthermore, we selected the anatomic specimens after carefully examining them for any evidence of congenital cause of scoliosis or any pathology causing the deformity.

The changes observed were progressively more important as the curves increased in severity. Not all changes were observed in smaller curves, as most of the changes were isolated to the vertebral body. Pedicle changes as well as posterior element changes occurred in progressively larger curves. These findings raise some interesting questions as to the origin of the deformity observed. Are these changes secondary to external forces acting to remodel the vertebral body or are these changes the precursors of the regional deformity? Vertebral wedging could be the initiating factor creating a relative force imbalance, therefore creating a vicious circle with progressive wedging creating more imbalance. But this does not completely explain the changes observed in the posterior elements, especially at the level of the pedicles. Do neurocentral junctions play a role in the initiation of scoliosis by modifying the pedicles’ shape and forces acting through them during the growth process? Is the pedicle shape modified by external forces such as the head of the rib pushing against the costovertebral area or is this pedicle thinning secondary to the spinal cord beating relentlessly for years on the concavity of the scoliotic curve?

It is not possible with this study to provide information on the etiology and natural history of idiopathic scoliosis because the deformities are fixed in time. All radiographs available were done in a supine position after death, and we did not have radiographs available from the time the specimens were living. It was therefore not possible to correlate the changes observed with the Cobb angle or to try to monitor the evolution of the deformity over time. This study, however, should provide important information to refine the three-dimensional reconstruction methods currently available and provide the tools necessary to study the evolution of the scoliotic deformity.

This study is of utmost importance in understanding the deformity occurring in thoracic and lumbar vertebrae of patients affected with scoliosis. To our knowledge, this is the largest database of vertebrae from scoliotic and normal spines in the literature. It is also the most descriptive study of its kind with 145 parameters measured for each thoracic vertebra and 127 parameters measured for lumbar vertebra. For the first time, we were able to characterize objectively a statistically significant deformation pattern for vertebrae from scoliotic spines. Qualitative observations made by several authors have been confirmed statistically, such as pedicle size changes and vertebral wedging. With this deformation pattern now available, it should be possible to predict the posterior element morphology based on information gathered on simple anteroposterior and lateral radiographs from any patients with a scoliotic deformity. Our results are also consistent with the study by Liljenqvist et al, which showed decreased pedicular width on the concavity of thoracic curves toward the apex. 7 We therefore advocate caution with the use of pedicle screws in the thoracic spine, on the concave side of the curves, of patients with moderate to severe scoliotic curves.


A total of 471 scoliotic and 510 normal vertebrae were measured. A typical deformation pattern was identified as being statistically significant which consisted of progressive vertebral wedging with the transitional vertebra showing no wedging and with the maximum deformity being found at the apex of the curve. Progressive thinning of the pedicle width was found on the concave side of the curve as well as variation of facet surfaces around the apex of the curve. The facet changes were not related to the concavity or convexity but did differ in size, especially with severe curvatures. These findings were more important in the thoracic spine. Finally, our results advocate caution with the use of pedicle screws in the thoracic spine on the concavity of scoliotic curves with moderate to severe deformity.

Key Points

  • Thirty scoliotic specimens from two major osteological sources were identified and measured.
  • A statistically significant deformation pattern was identified for idiopathic scoliosis.
  • Pedicle width is significantly smaller on the concavity of moderate to severe thoracic curves.
  • We advocate caution in the use of pedicle screws in the thoracic spine, on the concave side of the curves, of patients with moderate to severe scoliotic curves.


1. Berry JL, Moran JM, Berg WS, et al. A morphometric study of human lumbar and selected thoracic vertebrae. Spine 1987; 12: 362–7.
2. Enneking WF, Harrington P. Pathological changes in scoliosis. J Bone Joint Surg Am 1969; 51: 165–84.
3. Keim HA. Scoliosis. Clin Symp 1972; 24: 2–32.
4. Keim HA. Scoliosis. Clin Symp 1978; 30: 1–30.
5. Keim HA, Hensinger RN. Spinal deformities: scoliosis and kyphosis. Clin Symp 1989; 41: 3–32.
6. Laporte S, Mitton D, Ismael B, et al. Quantitative morphometric study of thoracic spine: a preliminary parameters statistical analysis. Eur J Orthop Surg Traumatol 2000; 10: 85–91.
7. Liljenqvist UR, Link TM, Halm HF. Morphometric analysis of thoracic and lumbar vertebrae in idiopathic scoliosis. Spine 2000; 25: 1247–53.
8. Panjabi MM, Duranceau J, Goel V, et al. Cervical human vertebrae: quantitative three-dimensional anatomy of the middle and lower regions. Spine 1991; 16: 861–9.
9. Panjabi MM, Goel V, Oxland T, et al. Human lumbar vertebrae: quantitative three-dimensional anatomy. Spine 1992; 17: 299–306.
10. Panjabi MM, Takata K, Goel V, et al. Thoracic human vertebrae: quantitative three-dimensional anatomy. Spine 1991; 16: 888–901.
11. Parent S, Mitulescu A, Seeman I, et al. A method for measuring scoliotic vertebrae: a preliminary study. Presented at: Canadian Orthopaedic Research Society Meeting; June 2000; Edmonton, Alberta, Canada.
12. Perdriolle R. La scoliose son étude tridimensionnelle. In: Maloine SA, ed. Paris: 1979.
13. Roaf R. Spinal Deformities. Philadelphia, PA: Lippincott; 1977.
14. Ronchetti PJ, Stokes IAF, Aronsson DD. Vertebral body and disc wedging in scoliosis. Res Spinal Deformities 1997; 1: 81–4.
15. Scoles PV, Linton AE, Latimer B, et al. Vertebral body and posterior element morphology: the normal spine in middle life. Spine 1988; 13: 1082–6.
16. Semaan I, Skalli W, Veron S, et al. Quantitative three-dimensional anatomy of lumbar spine. Rev Chir Reparatrice Appar Mot 2001; 87: 340–53.
17. Xiong B, Sevastik JA, Hedlund R, et al. Radiographic changes at the coronal plane in early scoliosis. Spine 1994; 19: 159–64.

scoliosis; thoracic vertebrae; lumbar vertebrae; biometry; comparative study; anthropometry; computer models]Spine 2002;27:2305–2311

© 2002 Lippincott Williams & Wilkins, Inc.