Accuracy on the preoperative assessment of patients with adolescent idiopathic scoliosis using biplanar low-dose stereoradiography: a comparison with computed tomography

We prospectively recruited all consecutive patients with AIS who were needed surgical treatment in our institution between June 2015 and January 2017. Children with other spinal pathologies, such as early onset scoliosis, previous spinal surgery, neurological symptoms or neural axis abnormalities, syndromes associated with disorders of growth, or atypical left convex thoracic curves or right convex (thoraco) lumbar curves, were excluded. Patients were further included when they had undergone preoperative EOS and CT scans at our institution on the same day. Written informed consent was obtained from the subjects or the parents of minor subjects before participating in this study. Ethical approval was obtained from the ethics review board of the joint NTEC-CUHK clinical research ethics committee. All study procedures were conducted in accordance to the guidelines approved by the ethics committee and the Declaration of Helsinki.

All subjects underwent whole body biplanar stereographs (EOS imaging, Paris, France) with a standardized radiographic protocol by a team of experienced radiographer. Subjects were instructed to stand in a comfortable position with hips and knees extended and with hands on a support. Upright EOS captured simultaneously the radiographs (Fig. 1a) of frontal and lateral views with two pairs of X-ray sources which positioned perpendicular to each other. The average scan speed used for EOS acquisition was 19.16 s. EOS images were then analyzed by the sterEOS workstation (Surgimap, Nemaris Inc., New York, NY) to generate the 3D surface reconstructions of each vertebra at each global axial rotation offset angle. Two trained observers, consisting of a senior consultant (B.K.W.N.) and an orthopaedics specialists (T.P.L.) with more than 30 years of experience in diagnosing AIS, reviewed the EOS scans. The measurements were performed twice for intra-rater reliability assessment. However, if there is an uncertainty between the two observers, a medical scientific officer (G.C.W.M.), with 14 years of experience in the field of scoliosis research, will helped with the diagnosis.

CT imaging (slice thickness of 0.625 mm, in-plane resolution of 0.352 mm/pixel, 64 Slice Multi-detector CT scanner, GE Healthcare, Chalfont, St. Giles, UK) was acquired in prone position, that was the standard workup in our medical centre during the inclusion period. The acquisition parameters for the CT imaging and 3D reconstruction is the following: 120 kV; 170 mA; rotation time = 0.8 s. The scan coverage in each case was from C7 to S1. Two trained observers, consisting of a Professor of Radiology (W.C.C.W) and a Professor of Orthopaedics and Traumatology (J.C.Y.C.) with more than 20 years of experience in diagnosing AIS, reviewed the CT scans. The measurements were performed twice for intra-rater reliability assessment. However, if there is an uncertainty between the two observers, a medical scientific officer (G.C.W.M.), with 14 years of experience in the field of scoliosis research, will helped with the diagnosis. In addition, a previously validated software and a semiautomatic image processing technique for CT scans of the scoliotic spine (ScoliosisAnalysis 4.1, Image Sciences Institute, Utrecht, The Netherlands, develop using MeVisLab, MeVis Medical Solutions AG, Bremen, Germany) were used to provide complete 3D coordinate systems of the individual structures of the spine [3]. By this method, the exact height of the osseous (anterior and posterior sides of the vertebral bodies, the laminae and the spinous processes) and non-osseous structures (anterior and posterior sides of the intervertebral discs, interlaminar spaces, and interspinous spaces) in the midsagittal plane were measured, while correcting for rotation and tilt in 3D (Fig. 1b). In contrast to the anatomical midsagittal plane of the patient, this complete 3D analysis method enabled the observer to reconstruct the midsagittal plane of each structure by taking account of axial rotation and coronal and sagittal tilt, as well as torsion (internal rotation) of each individual structure.

All the EOS and CT images were collected from the Picture Archiving and Communications Systems (PACS) workstation (Carestream solution working station, Carestream Health, Version 11.0, Rochester, New York, USA). The calculated parameters were divided into 6 categories. Each category refers to global (whole spine), regional (scoliotic segment), and local (vertebra) descriptors. Vertebra centroid is understood as the halfway point between the centers of the 2 endplates of the vertebra. The local vertebra axis system is defined by the SRS 3D terminology group as follows (Fig. 2a – f):

The height and the width were the distance in a coronal plane between superior and inferior endplates at the centre of the vertebra, and between left and right sides at the centre of vertebra which was perpendicular to the height line measured in a coronal plane from T1-L5 vertebrae, respectively.

This is the measured rotation between two adjacent vertebrae in axial plane at apical, upper and lower curves, and thoracolumbar junction (T12-L1) vertebrae.

Mean of the sum of intervertebral axial rotation (measured according to the local referential of the inferior vertebrae) of the 2 hemicurvatures of the curve (between upper end vertebra and apex and between lower end vertebra and apex)

This is the measured angle between superior and inferior endplate of apical vertebra, between the disks and apical vertebra, between the disks and upper curve vertebra, and between the disks and lower curve vertebra.

Cobb angle is defined as the most tilted vertebrae above apex and below apex of a curve on coronal plane. Also, measured the angles on sagittal plane for thoracic kyphosis (T4-T12) and lumbar lordosis (L1-L5).

This was measured in the coronal plane from the angle between the centre of vertebra T1 to the centre of apex, and the centre of apex to the centre of vertebra L5.

All of the parameter results were compared between the prone (CT-generated DRR) and upright (EOS) positions.

Using descriptive statistics to compute the means and standard deviations. A paired t-test was performed and compared the parameters between two scans. A Wilcoxon signed ranks test was used for the parameter with non-normal distributed data. The agreement between the two positions was tested according to the Bland-Altman plot; first, the one sample t test showed if there was a significant difference between the measurements; second, if there was no significant difference, the regression analysis showed if there was agreement between the measurements [15]. The correlations between two scans in the measurements were analyzed with a Pearson correlation test. The intra- and inter-observer reliability were obtained as intraclass correlation coefficients (ICC). P-value less than 0.05 was considered statistically significant for all analyses. All analyses were conducted with the SPSS software (Version 25.0; SPSS, Chicago, IL, USA).

A total of 33 pre-operative AIS patients were recruited during the study period. This consisted of 26 females and 7 males with a mean age of 18.4 ± 4.2 years (range, 13–31 years) and mean Cobb angle of 62.9 ± 9.3°. There were 29 thoracic, 2 thoracolumbar and 2 lumbar curves. Most of the curves were classified as Lenke type 1 and 2 of the severe AIS patients. All descriptive statistics are shown in Table 1.

The intraclass correlation coefficient (ICC) of intra-rater reliabilities for all scoliotic parameters at prone and upright positions were ≥ 0.913 and ≥ 0.878, respectively. The ICC of inter-rater reliabilities for all scoliotic parameters at prone and upright positions were ≥ 0.930 and ≥ 0.706, respectively. Overall, the ICC data suggested excellent measurement consistency and reliability in both the prone and upright positions (Table 2).

Based on the coronal plane, no significant difference (P > 0.05) in slenderness was found between the prone CT and upright EOS (Table 3, Fig. 3). Likewise, the Cobb angle was significantly larger in upright EOS (62.9 ± 9.3°) than in prone CT (47.3 ± 10.0°, P < 0.01) positions as well as the spinal deformity (Table 3, Fig. 3). Although there was no observable difference in vertebral wedging being found, however, there was a significant difference in intervertebral wedging at apical, upper and lower end levels (P < 0.01) being found (Table 3, Fig. 3). And according to the Bland-Altman method, there was significant difference between Cobb angle, spinal deformity, vertebral wedging and intervertebral wedging in the prone (CT) and upright (EOS) positions (Table 3). In addition, all these parameters were significantly correlated between prone (CT) and upright (EOS) positions (R² ≥ 0.75, P < 0.01; Fig. 4).

On the sagittal plane, no significant difference in kyphosis was found between prone CT (20.0 ± 14.2°) and upright EOS (18.8 ± 10.3°) (Table 3, Fig. 3). On the other hand, the prone LL (27.9 ± 11.4°) was significantly lower than upright lordosis (48.8 ± 12.4°) (P < 0.01) (Table 3, Fig. 3). And from the Bland-Altman analysis, kyphosis was in agreement between the prone and upright positions (Table 3). However, lordosis was also found to be significantly lowered in prone (CT) position than when measured at upright (EOS) position (Table 3). No significant correlation was found in both kyphosis and lordosis between two positions (Fig. 4).

While for the axial profile, there was no significant difference in intervertebral axial rotation at apical, upper and lower end levels, and the thoracolumbar (T12-L1) level being observed between prone (CT) and upright (EOS) (Table 3, Fig. 3). Additionally, no significant difference was found on torsion (P > 0.05) when compared between prone (CT) and upright (EOS) (Table 3, Fig. 3). According to the Bland-Altman method, there was agreement between the intervertebral rotation at apical level and torsion in the prone and upright positions (Table 3). However, there was significant correlation in intervertebral rotation at the apical level (R² = 0.834; P < 0.01) between the prone CT and upright EOS (Fig. 4).