Early brace treatment for idiopathic scoliosis may change the paradigm to improve curves

Deep learning classification of pediatric spinal radiographs for use in large scale imaging registries

PURPOSE

The purpose of this study is to develop and apply an algorithm that automatically classifies spine radiographs of pediatric scoliosis patients.

METHODS

Anterior-posterior (AP) and lateral spine radiographs were extracted from the institutional picture archive for patients with scoliosis. Overall, there were 7777 AP images and 5621 lateral images. Radiographs were manually classified into ten categories: two preoperative and three postoperative categories each for AP and lateral images. The images were split into training, validation, and testing sets (70:15:15 proportional split). A deep learning classifier using the EfficientNet B6 architecture was trained on the spine training set. Hyperparameters and model architecture were tuned against the performance of the models in the validation set.

RESULTS

The trained classifiers had an overall accuracy on the test set of 1.00 on 1166 AP images and 1.00 on 843 lateral images. Precision ranged from 0.98 to 1.00 in the AP images, and from 0.91 to 1.00 on the lateral images. Lower performance was observed on classes with fewer than 100 images in the dataset. Final performance metrics were calculated on the assigned test set, including accuracy, precision, recall, and F1 score (the harmonic mean of precision and recall).

CONCLUSIONS

A deep learning convolutional neural network classifier was trained to a high degree of accuracy to distinguish between 10 categories pre- and postoperative spine radiographs of patients with scoliosis. Observed performance was higher in more prevalent categories. These models represent an important step in developing an automatic system for data ingestion into large, labeled imaging registries.

Correlation of transverse rotation of the spine using surface topography and 3D reconstructive radiography in children with idiopathic scoliosis

PURPOSE

The relationship between axial surface rotation (ASR) measured by surface topography (ST) and axial vertebral rotation (AVR) measured by radiography in the transverse plane is not well defined. This study aimed to: (1) quantify ASR and AVR patterns and their magnitudes from T1 to L5; (2) determine the correlation or agreement between the ASR and AVR; and (3) investigate the relationship between axial rotation differences (ASR-AVR) and major Cobb angle.

METHODS

This is a retrospective study evaluating patients (age 8-18) with IS or spinal asymmetry with both radiographic and ST measurements. Demographics, descriptive analysis, and correlations and agreements between ASR and AVR were evaluated. A piecewise linear regression model was further created to relate rotational differences to Cobb angle.

RESULTS

Fifty-two subjects met inclusion criteria. Mean age was 14.1 ± 1.7 and 39 (75%) were female. Looking at patterns, AVR had maximal rotation at T8, while ASR had maximal rotation at T11 (r = 0.35, P = .006). Cobb angle was 24.1° ± 13.3° with AVR of - 1° ± 4.6° and scoliotic angle was 20.9° ± 11.5° with ASR of - 2.3° ± 6.6°. (ASR-AVR) vs Cobb angle was found to be very weakly correlated with a curve of less than 38.8° (r = 0.15, P = .001).

CONCLUSION

Our preliminary findings support that ASR measured by ST has a weak correlation with estimation of AVR by 3D radiographic reconstruction. This correlation may further help us to understand the application of transverse rotation in some clinical scenarios such as specific casting manipulation, padding mechanism in brace, and surgical correction of rib deformity.

Outcomes for nighttime bracing in adolescent idiopathic scoliosis based on brace wear adherence

BACKGROUND

This study determined brace wear adherence for patients treated with nighttime braces and evaluated the effect of brace adherence on curve progression.

METHODS

One hundred twenty-two patients with AIS ages 10-16 years, Risser stages 0-2, major curves 20°-40° treated with Providence nighttime braces prescribed to be worn at least 8 h per night were prospectively enrolled and followed until skeletal maturity or surgery. Brace adherence was measured using iButton temperature sensors after 3 months of brace initiation and at brace discharge.

RESULTS

Curve types were single thoracolumbar/lumbar (62%, n = 76), double (36%, n = 44), and single thoracic (2%, n = 2). Brace adherence averaged 7.8 ± 2.3 h after 3 months (98% adherence) and 6.7 ± 2.6 h at brace discharge (84% adherence). Curves that progressed ≥ 6° had decreased brace adherence than non-progressive curves after 3 months (7.0 h vs. 8.1 h, p = 0.010) and at brace discharge (5.9 h vs. 7.1 h, p = 0.017). Multivariate logistic regression analysis showed that increased hours of brace wear [odds ratio (OR) 1.23, 95% confidence interval (CI) 1.06-1.46], single curves (OR 3.11, 95% CI 1.35-7.53), and curves < 25° (OR 2.61, 95% CI 1.12-6.44) were associated with non-progression at brace discharge.

CONCLUSIONS

Patients treated with nighttime bracing have a high rate of brace adherence. Lack of curve progression is associated with increased brace wear. Nighttime bracing is effective at limiting curve progression in AIS single thoracolumbar/lumbar and double curves.

LEVEL OF EVIDENCE

Prognostic Level 2.