Wozniczka, Jennifer K. MD^*; Ledonio, Charles G. T. MD^*; Polly, David W. Jr MD^*; Rosenstein, Benjamin E. BS^†; Nuckley, David J. PhD^‡

Scoliosis has been shown to have detrimental effects on pulmonary function, traditionally measured by pulmonary function tests (PFTs), which is theorized to be correlated to the distortion of the spine and thorax with deformity of the spine and ribs progressing to restrictive lung disease.¹ The changes in thoracic volume that occur with increasing coronal deformity as measured by the Cobb method, sagittal alignment, and surgical correction are not well understood and are difficult to assess using traditional 2-dimensional radiographic measurements. A direct link between thoracic volume and pulmonary function has not been well quantified.

Normal ventilation is dependent on a stable thorax that allows for expansion through the diaphragm and intercostal muscles of the ribs, airway resistance, lung volume, and lung tissue general health. Large curves in scoliosis have long been known to cause cardiopulmonary restriction, and restrictive lung disease is commonly seen in curves larger than 90 degrees, which results in higher rates of early death from cor pulmonale.² However, even patients with mild and moderate curves can have decreased lung function, and may demonstrate lower tolerance to maximal exercise.³ Spinal deformity, increased curve magnitude, levels of spine involved in the curve, age, rigidity of curve, duration of the deformity, and chest wall motion are all known causes of pulmonary impairment.⁴

The change in lung volume after surgical correction has not been quantified. Previous imaging studies using computed tomography (CT) have been used to measure thoracic volume,⁵ but limited data are available for analysis as patients with adolescent idiopathic scoliosis do not routinely have CT scans preoperatively and postoperatively. Computational modeling study does not require radiation exposure beyond routine radiographs, and allows 3-dimensional (3D) data to be obtained from retrospective 2-dimensional (2D) radiographs.⁶ This study uses this novel method to understand the 3D volume changes of the thorax with surgical correction of scoliosis, to gain further insight into an important component of lung function.

METHODS

This study is a level 3 retrospective comparison of 9 prospectively collected cases with Lenke type 1 adolescent idiopathic scoliosis with scoliotic curves as measured by the Cobb method of increasing magnitudes of severity. Patients who met criteria were selected through the Prospective Pediatric Scoliosis Study (PPSS) deidentified database. A representative pilot group of patients were chosen to demonstrate a range of moderate to severe spinal deformity. Patients were chosen with scoliotic curvature measurements of increasing value, with both neutral or hypokyphotic sagittal alignment, who had undergone surgical correction. All patients had 2-year postoperative anterior-posterior and lateral plane scoliosis radiographs.

All images were initially calibrated in pixels per centimeter using ImageJ software (version 1.46r. Open source: imagej.nih.gov; 2012). These calibrated images were imported into Blender software (version 2.63a. Open source: http://www.blender.org, Amsterdam, The Netherlands; 2012) to construct a 3D computational model of the spine from the 2D radiographs (Fig. 1). To accomplish this, the 3D thorax model (Fig. 2) developed by the spine research team was sized and deformed to match the radiographs (Figs. 3, 4) using Blender software. This was done via manipulations within Blender, including scaling, rotation, translocation, and simple deformation modifiers. Once the thorax model matched the sagittal and coronal radiographs, Blender software was used to obtain a corresponding 3D model of the thoracic volume (Fig. 5), by first exporting the deformed thorax model as a stereolithography (STL) file. A closed polyhedron icosphere mesh was inset, expanded, and bounded by the virtual thoracic cage. A shrinkwrap modifier was used so the virtual ribs and spine could be excluded from the volume measurement. This conformed icosphere was then exported from Blender as an STL file, where the isocahedron volume was analyzed using Mimics software (version 15; Materialise Inc., Leuven, Belgium). All thoracic modeling was performed by one person, and all volumes were obtained in a blinded manner by another individual.

FIGURE 1

FIGURE 2

FIGURE 3

FIGURE 4

FIGURE 5

Previous validation of this technique has been completed at our laboratory by comparison of volumes obtained from CT data, the gold standard, with volume measurements obtained by the thorax deformation modeling method. The patient data set used for validation similarly had scoliosis radiographs, but also had CT data available for comparison. Percent error was computed as the total thoracic mimics volume from CT minus the computed volume divided by the total thoracic volume from CT then multiplied by 100. The patient-specific models of the thorax were compared with CT volumetric modeling, and revealed a mean error of 2% (range, 1.9% to 4.8%) from the gold standard.⁷

There was no correlation seen in this group between preoperative age, height, or weight and preoperative volume measurement.

RESULTS

Patient Population

Nine patients with adolescent idiopathic scoliosis with Lenke type 1 curves were identified with preoperative curve measurements assessed using the Cobb technique >50 degrees (range, 50 to 96 degrees). Preoperatively, 4 of the patients had neutral sagittal contour, with T5-T12 angle between 10 and 40 degrees, and 5 of the patients had hypokyphotic sagittal contour, with T5-T12 angle < 10 degrees. All patients underwent spinal fusion with 2-year postoperative radiographs that were used to compare thoracic volume differences.

The average age of the patients preoperatively was 14.0 years (range, 10 to 18.7 y) (Table 1). Average height was 154.6 cm (±15.3 cm) and average weight was 47.2 kg (±18.9 kg). Preoperative age, height, and weight did not correlate with preoperative volume measurement.

TABLE 1

Surgical Correction

Surgical intervention resulted in decreased curve measurements using the Cobb method, and increased mean thoracic volume, in all patients. Evaluation of postoperative data demonstrated that curve measurements were reduced and thoracic volume increased. Curve measurements were significantly reduced from a mean of 69 degrees preoperatively to 27 degrees postoperatively (P<0.001). Thoracic volume significantly increased by a mean of 567 mm³ (P<0.001) (range, 51 to 1095 mm³). The percent change in thoracic volume after surgical correction averaged 40% (range, 3% to 87%). The smaller the baseline preoperative volume, the greater the change in volume postoperatively (r=–0.86) (Fig. 6).

FIGURE 6

Postoperative curve measurements, sagittal contour, or change in spinal curvature after surgical intervention did not have significant correlations with postoperative volume or change in volume. With regard to preoperative radiographic parameters, only curve measurements as assessed using the Cobb technique had a weak correlation with postoperative volume (r=0.45), but this was not statistically significant.

Preoperative Data

With the preoperative volumes, curve measurements as determined using the Cobb method were the only measures found to correlate to the preoperative volume, and with an increase in curve measurement, there was a corresponding decrease in volume. For the neutral sagittal contour patients with T5-T12 angle between 10 and 40 degrees, there was a moderate inverse correlation between curve measurement and thoracic volume (r=−0.629) (Fig. 7). At a curve measurement of 50 degrees, volume was measured at 1862 mm³; at 60 degrees, volume was 1263 mm³; at 71 degrees, 1538 mm³; and at 80 degrees, 1421 mm³ (Table 2, Figs. 2, 3).

FIGURE 7

TABLE 2

In patients with hypokyphotic sagittal contour, where T5-T12 angle was <10 degrees, there was a trend toward decreased volumes in patients with larger curve measurements (Fig. 7). A weak inverse correlation was seen between curve measurement and thoracic volume (r=−0.458), but this finding was not statistically significant. At a curve measurement of 50 degrees, volume was measured at 1858 mm³; at 60 degrees, volume was 1538 mm³; at 71 degrees, 1483 mm³; at 86 degrees, 1604 mm³; and at 96 degrees, 1596 mm³ (Table 2, Figs. 2, 3).

In directly comparing the patients within these groups, a statistically significant correlation could not be found between preoperative sagittal angle and preoperative volume.

DISCUSSION

Computational Modeling

A variety of methods have attempted to directly assess thoracic volume in patients with scoliosis.^2,5,8,9 Early studies used radiographs to determine lung volume,^10,11 but these methods did not account for the 3D deformity of the spine. Gollogly et al⁵ used CT scan data to assess lung volumes in children who underwent expansion thoracoplasty for spinal deformity and found an increase in lung volumes of 25% to 90% after surgical intervention. This method works well in patients with severe spinal deformities that require CT for preoperative planning. The majority of patients with adolescent idiopathic scoliosis do not have preoperative or postoperative CT scans as this exposes the patient to additional radiation with its own risks.¹² Our method of lung volume computation minimizes radiation exposure by using radiographs that are used to follow all patients with scoliosis. Furthermore, our method goes beyond previous radiographic studies and assesses the 3D changes in the thorax. Previous studies have used this computational technique and found volume measurement maximum error of 4.8% with this technique compared with CT scan measurements.¹³ This study is an assessment of this modeling technique, which can be used in patients when CT data are not available to begin to understand volume changes with surgical correction.

Pulmonary Function

PFTs have traditionally been used to study the effect of surgical correction on patients with scoliosis.^14–16 Using PFTs to determine lung function can be problematic, as other factors such as patient cooperation or intrinsic lung disease may affect results. Patients may have a significant decrease in their PFT values before they drop out of the normal range.⁸ Calculations for predicted lung values are based on height, and due to spine curvature, the predicted values in patients with scoliosis can therefore underestimate expected values.^17,18

Radiographic measurements, such as spinal curvature measurements as determined by Cobb technique, have not been shown to strongly correlate with pulmonary function.¹⁹ Isolating larger thoracic curves, Johnston et al¹ found significant correlation with poor preoperative PFTs, but many AIS curves have more moderate deformities. Newton et al¹⁸ used preoperative PFT on over 600 patients with adolescent idiopathic scoliosis, and the authors acknowledged that radiographic findings could not explain all the variability in PFTs. With surgical correction, Kim et al²⁰ noted no correlation between curvature measurement correction and significant clinical improvement in PFTs of 31 patients.

Surgical Correction of Thoracic Volume

Surgical correction and fusion in adolescent idiopathic scoliosis prevents progression of the spinal deformity, and traditionally, surgical indications have largely been based on coronal deformity as measured by the Cobb technique, progression, and remaining growth. In particular, curve angle is associated with thoracic deformity as the increased angle generally results in a decreased height of the hemithorax on the concave side of the curve.⁸ Because of a paucity of literature on the subject, the relationship between postsurgical thoracic volume changes and lung function has not been well quantified.^20–24 Surgical correction through a posterior approach has not been shown to change PFTs.²⁵ Studies using static thoracic dimensional measurements are only weakly correlated predictors of pulmonary function outcome, including the recent 2014 study by Glotzbecker et al.²⁶ This study employed traditional 2D measurements, and concluded that further investigations using 3D measurements need to be developed. Our computational model is 3D, and investigates preoperative and postoperative thoracic volume change. Rosenstein and colleagues demonstrated that preoperative thoracic volume was diminished in patients with the lowest PFT values, and these same patients had both an increase in thoracic volume and improvement of PFTs postoperatively,²⁷ but this may not apply in more moderate curves usually seen in adolescent idiopathic scoliosis. The confounder in all of this is that although thoracic volume may increase with surgical correction, chest wall stiffness may also increase, resulting in a net effect on PFTs that is variable.

The results using our computational modeling technique demonstrated that surgical intervention with correction and fusion resulted in decreased curve measurement as determined by the Cobb technique and increased postoperative thoracic volume as expected. Evaluation of postoperative data demonstrated that curve measurements were significantly reduced from a mean of 69 degrees (range, 50 to 96 degrees) preoperatively to 27 degrees (range, 13 to 33 degrees) postoperatively (P<0.001). Thoracic volume significantly increased by a mean of 567 mm³ (P<0.001). The smaller the baseline volume, the greater the change in volume postoperatively (r=−0.86), as these patients had the greatest opportunity for improvement with surgical correction. The percent change in thoracic volume after surgical correction averaged 40% (median 41%), but there was a large range of 3% to 87%. The variation in volume change on a case-by-case basis will require further investigation, as the differences in preoperative deformity do not fully explain the range.

Improvement of thoracic volume is a necessary, but not sufficient condition for potentially improving lung function; however, it is the one parameter that can be changed surgically. Thoracic volume is only one piece in understanding respiratory function in scoliosis patients, but discrete measurements of volume through this computational technique can provide us with knowledge on how spinal correction results in variable improvements in thoracic volume.

Curve Measurement, Sagittal Contour

There was a weak, not statistically significant, correlation between preoperative curve measurements as assessed using the Cobb technique and change in volume postoperatively. Postoperative curve measurement, sagittal contour, or change in curve measurement after surgical intervention all did not have a statistically significant correlation with postoperative volume or the change in volume. Because of our small sample size and patient heterogeneity, we have limited ability to analyze the effect of sagittal contour on thoracic volume and lung volume. Previous studies that have tried to correlate preoperative curve measurement and sagittal contour with lung function have varied findings.^1,19,25 Johnston et al¹ found that patients with preoperative curves >70 degrees and T5-T12<10 degrees had significantly lower FEV₁ and FVC preoperatively. Newton et al¹⁸ have found trends showing increasingly abnormal PFTs with larger curve magnitude, as measured by curve length and major curve angle, and patients with hypokyphosis were the most likely in the cohort to have moderate or severe pulmonary impairment. Yet, Redding and Mayer¹⁹ investigated children with early-onset scoliosis and found that major curve angle had a poor correlation with lung function preoperatively and with changes in major curve angles postoperatively.

Preoperative Curve Measurement, Sagittal Contour, and Preoperative Thoracic Volume

Preoperative results did demonstrate a moderate inverse correlation between preoperative curve measurements as determined by the Cobb method and preoperative thoracic volume; whereby, as curve measurements increased, thoracic volume decreased. These results agree with the consensus that increasing curve measurements generally result in a decreased height of the hemithorax on the concave side of the curve,⁸ which can decrease space available for the lungs.

There was not a statistically significant correlation identified between sagittal contour angle and thoracic volume preoperatively or postoperatively. This was an unexpected finding and may be due to a number of possible reasons. First, there are a small number of patients included in this study. Also, wide variations in coronal deformity were used, so it may be difficult to discern the effects of the smaller variations in sagittal contour in this limited study. In addition, sagittal measurement contour depends on the quality of lateral radiographs, and modeling therefore can be more difficult than in the coronal plane.

Limitations

There are limitations to the modeling method and study. These data analyzing spinal deformity, involves a small number of patients. Also thoracic volume data were not normalized to height, age, or weight, but volumes were compared preoperatively and postoperatively for the same patient. PFT data were not available for comparison to thoracic volume. In addition, changes in structure were only derived from 1 time point preoperatively. This method also relies on radiographs where quality of radiographs and rotation of the patient may cause small errors in the volume measurements. Validation of this technique has shown differences of thoracic volumes between CT and this method to have a maximum error of 4.8%.¹³ Additional investigation with computational modeling can provide us with an understanding of how thoracic volume changes as a result of spinal deformity and with surgical correction.

CONCLUSIONS

This pilot study demonstrates methodologic plausibility for measuring 3D changes in thoracic volumes using 2D imaging. This is an assessment of the novel modeling technique, to be used in larger future studies to assess clinical significance.

ACKNOWLEDGEMENTS

The authors wish to acknowledge the Prospective Pediatric Scoliosis Study (PPSS) for providing the deidentified images and data which were utilized in this study.

REFERENCES