# Predicting pulmonary function using thoracic deformity parameters in early onset scoliosis patients

## TL;DR

This study shows how thoracic deformity measurements from X-rays can predict lung function in children with early onset scoliosis, avoiding the need for difficult pulmonary tests.

## Contribution

The study introduces the most precise models to date for predicting lung function using radiographic deformity parameters in early onset scoliosis patients.

## Key findings

- MLR models predicted %FVC with R² = 0.54 and %FEV1 with R² = 0.59 in EOS patients.
- Validation showed no significant differences in prediction error magnitudes.
- Key deformity parameters were identified to guide surgical treatment decisions.

## Abstract

Thoracospinal deformities in early onset scoliosis (EOS) patients often lead to thoracic insufficiency syndrome, in which respiration or lung growth is impaired. Pulmonary function tests (PFTs) are used to assess pulmonary deficits but are challenging to comply with for EOS patients, who typically are between 5 and 10 years old. Thus, the objective was to predict PFT values in EOS patients directly from deformity parameters measured on routine radiographs.

Corresponding preoperative radiographs and PFT values were retrospectively obtained from 47 EOS patients (13M/34F; mean age: 9.8 ± 3.0 years), and 19 literature-based deformity parameters were measured. Multiple linear regression (MLR) analyses using an exhaustive search feature selection method were used to estimate percent predicted forced vital capacity (%FVC) and forced expiratory volume in one second (%FEV1). Ten percent of the dataset was set aside to validate the predictive accuracy of the MLR models.

The additive contributions of multiple thoracospinal deformity parameters successfully yielded significant (p < 0.001) MLR models that predicted %FVC (R2 = 0.54) and %FEV1 (R2 = 0.59) in EOS patients. For the validation test, no significant differences (p > 0.05) in prediction error magnitudes were found.

The developed MLR models provide the highest reported precision for predicting PFT values in EOS patients from radiographic deformity parameters. Additionally, a key subset of deformity parameters was identified, and their relative contributions to predicting PFT values provide quantitative metrics to guide surgical treatment.