Quadratic Spline Fitting for Robust Measurement of Thoracic Kyphosis Using Key Vertebral Landmarks.

Area of Science:

• Orthopedics and Sports Medicine
• Radiology
• Biomedical Engineering

Background:

• Traditional kyphosis measurement methods like the Cobb angle have limitations in accuracy and continuous curvature representation.
• Accurate assessment of thoracic spine curvature is crucial for diagnosing and managing various spinal conditions.

Purpose of the Study:

• To introduce and validate a novel kyphosis measurement technique utilizing quadratic spline fitting through T12, T8, and T4 vertebral landmarks.
• To compare the accuracy and reliability of the spline method against traditional circle fitting and the reference Cobb angle.

Main Methods:

• Retrospective analysis of 560 lateral thoracic spine radiographs across diverse kyphosis types.
• Independent annotation by two raters to compute kyphosis angles using quadratic spline fitting and least-squares circle fitting.
• Evaluation of agreement with Cobb angles via Pearson correlation, MAE, RMSE, ROC analysis, and Bland-Altman plots; reliability assessed using ICC.

Main Results:

• Both spline and circle methods demonstrated excellent intra- and inter-rater reliability (ICC ≥ 0.967).
• The spline method exhibited superior performance with lower MAE (5.81°), lower RMSE (8.94°), and reduced bias compared to the circle method.
• Both methods showed strong correlation with Cobb angles (r ≥ 0.851) and high classification accuracy (AUC > 0.950).

Conclusions:

• Quadratic spline fitting offers an accurate, reliable, and robust method for kyphosis measurement, particularly effective for severe spinal deformities.
• This automated and reproducible technique has the potential to improve clinical evaluation of spinal curvature, integrating with AI-driven image analysis.