Reassessing vertebral endplate references in parametric spinal analysis: Validation of a centroid-based method for...

The vertebral endplate, the primary anatomical reference in spine measurements, is highly susceptible to observer variability. Alternatively, the geometric centroid model offers a more reproducible reference. This model is constructed by fitting a curve through geometric centroids derived from the four annotated vertices of each vertebral body.

To introduce vertebral slope measurement using the normal to the centroid-based spinal model.

Three evaluators annotated thirty-two standing lateral radiographs. Dedicated software computed geometric centroids, interpolated a mathematical spinal curve, and measured vertebral inclination relative to the curve's normal line. For comparison, vertebral slope was measured using the classical endplate-based method. Vertebral slope difference between adjacent vertebrae was also analyzed. Inter and intraobserver variability was evaluated using intraclass correlation coefficients (ICC). Paired t-tests were applied to compare methods at each vertebral level.

Vertebral slope reflects the absolute spatial orientation of each vertebra, and the vertebral slope difference captures local curvature. Mean intraobserver ICCs were 0.88 for the endplate method and 0.97 for the centroid method, while interobserver ICCs were 0.85 and 0.96, respectively. The centroid method demonstrated significantly lower variability in vertebral inclination across most levels (C4 to L4, p < 0.05). Differences at L5 (p = 0.160) and S1 (p = 0.397) were not statistically significant.

Vertebral slope and vertebral slope difference offer an individualized analysis of the spine. Compared to the endplate method, the geometric centroid model substantially reduced observer variability. While the endplate proved suboptimal overall, the S1 endplate reference remains necessary due to sacral anatomy.