Comparison of clinical-CT segmentation techniques for measuring subchondral bone cyst volume in glenohumeral osteoarthritis

Abstract

Purpose: This study aimed to assess the accuracy and reproducibility of four common segmentation techniques measuring subchondral bone cyst volume in clinical-CT scans of glenohumeral OA patients. Methods: Ten humeral head osteotomies collected from cystic OA patients, having undergone total shoulder arthroplasty, were scanned within a micro-CT scanner, and corresponding preoperative clinical-CT scans were gathered. Cyst volumes were measured manually in micro-CT and served as a reference standard (n = 13). Respective cyst volumes were measured on the clinical-CT scans by two independent graders using four segmentation techniques: Qualitative, Edge Detection, Region Growing, and Thresholding. Cyst volume measured in micro-CT was compared to the different clinical-CT techniques using linear regression and Bland–Altman analysis. Reproducibility of each technique was assessed using intraclass correlation coefficient (ICC). Results: Each technique outputted lower volumes on average than the reference standard (-0.24 to -3.99 mm3). All linear regression slopes and intercepts were not significantly different than 1 and 0, respectively (p < 0.05). Cyst volumes measured using Qualitative and Edge Detection techniques had the highest overall agreement with reference micro-CT volumes (mean discrepancy: 0.24, 0.92 mm3). These techniques showed good to excellent reproducibility between graders. Conclusions: Qualitative and Edge Detection techniques were found to accurately and reproducibly measure subchondral cyst volume in clinical-CT. These findings provide evidence that clinical-CT may accurately gauge glenohumeral cystic presence, which may be useful for disease monitoring and preoperative planning. Level of evidence: Retrospective cohort Level 3 study.