Validity and reliability of active shape models for the estimation of cobb angles in adolescent idiopathic scoliosis

Choosing the most suitable treatment for scoliosis relies heavily on accurate and reproducible Cobb angle measurement from successive radiographs. The objective is to reduce variability of Cobb angle measurement by reducing user intervention and bias. Custom software to automate Cobb angle measurement from posteroanterior radiographs was developed using active shape models. Validity and reliability of the automated system against a manual and semi-automated measurement method was conducted by two examiners each performing measurements on 3 occasions from a test set (N=22). A training set (N=47) of radiographs representative of curves seen in a scoliosis clinic was used to train the software to recognize vertebrae from T4 to L4. Images with a maximum Cobb angle between 20 degrees and 50 degrees, excluding surgical cases, were selected for training and test sets. Automated Cobb angles were calculated using best-fit slopes of the detected vertebrae endplates. Intra-class correlation coefficient (ICC) and standard error of measurement (SEM) showed high intra-examiner (ICC > 0.90, SEM 2-3 degrees) and inter-examiner (ICC > 0.82, SEM 2-4 degrees), but poor inter-method reliability (ICC=0.30, SEM 8-9 degrees). The automated method underestimated large curves. The reliability improved (ICC = 0.70, SEM 4-5 degrees) with exclusion of the 4 largest curves (> 40 degrees) in the test set. The automated method was reliable for moderate sized curves, but did not properly detect vertebrae in larger curves. Optimization of constraints on scaling, rotation, translation, and iteration may improve reliability with larger curves.