Improved precision of syndesmophyte measurement for the evaluation of ankylosing spondylitis using CT: a phantom and patient study

• Published in: Physics in medicine & biology Vol. 57; no. 14; pp. 4683 - 4704
• Main Authors: Tan, Sovira, Yao, Jianhua, Yao, Lawrence, Ward, Michael M
• Format: Journal Article
• Language: English
• Published: England IOP Publishing 21.07.2012
• Subjects: Algorithms; biomedical image processing; computer-aided diagnostic; Humans; medical image analysis; Patient Positioning; Phantoms, Imaging; precision/reliability; Spine - diagnostic imaging; Spine - physiopathology; Spondylitis, Ankylosing - diagnostic imaging; Spondylitis, Ankylosing - physiopathology; Tomography, X-Ray Computed - instrumentation; volumetry
• ISSN: 0031-9155; 1361-6560; 1361-6560
• DOI: 10.1088/0031-9155/57/14/4683

Ankylosing spondylitis is a disease characterized by abnormal bone formation (syndesmophyte) at the margins of inter-vertebral disc spaces. Syndesmophyte growth is currently typically monitored by the visual inspection of radiographs. The limitations inherent to the modality (2D projection of a 3D object) and rater (qualitative human judgment) may compromise sensitivity. With newly available treatments, more precise measures of syndesmophytes are needed to determine whether treatment can slow rates of syndesmophyte growth. We previously presented a computer algorithm measuring syndesmophyte volumes and heights in the 3D space of CT scans. In this study, we present improvements to the original algorithm and evaluate the gain in precision as applied to an anthropomorphic vertebral phantom and patients. Each patient was scanned twice in one day, thus providing two syndesmophyte volume and height measures. The difference between those two measures (ideally zero) determines our algorithm's precision. The technical improvements to the algorithm decreased the mean volume difference (standard deviation) between scans from 3.01% (2.83%) to 1.31% (0.95%) and the mean height difference between scans from 3.16% (2.99%) to 1.56% (1.13%). The high precision of the improved algorithm holds promise for application to longitudinal clinical studies.