Related Experiment Videos
Identifying sources of variability in scoliosis classification using a rule-based automated algorithm.
Ian A F Stokes1, David D Aronsson
1Department of Orthopaedics and Rehabilitation, University of Vermont, Burlington, Vermont 05405, USA. Ian.Stokes@uvm.edu
Spine
|December 18, 2002
Summary
A rule-based algorithm was developed to improve the reliability of spinal deformity classification in adolescent idiopathic scoliosis. While it reduces some errors, classification remains challenging at boundary conditions.
Area of Science:
- Orthopedics
- Biomedical Engineering
- Radiology
Background:
- Classification of thoracic idiopathic scoliosis for surgical planning has known reliability issues.
- Sources of unreliability in current classification schemes are poorly understood.
Purpose of the Study:
- To determine if unambiguous, computer-encoded rules can ensure reliable radiographic classification of spinal deformities.
- To identify sources of variability in radiographic classification using an automated algorithm.
Main Methods:
- Developed and tested an objective classification methodology based on the King et al. (1983) scheme.
- Digitized vertebral corner positions for automated evaluation of spinal shape parameters.
- Utilized a rule-based algorithm for classification after resolving ambiguities in the original scheme.
Main Results:
- The automated algorithm encountered reliability issues in specific cases: near-midline lumbar curves, equal thoracic/lumbar curves, ambiguous thoracolumbar apex, Cobb angles near 10 degrees, and flexibility index near unity.
- Identified specific scenarios where the automated classification system faced challenges.
Conclusions:
- Objective measurements and rule-based algorithms can mitigate interobserver and intraobserver errors in spinal deformity classification.
- Reliability challenges persist when classification parameters approach decision boundaries.