Reliability of Manually Segmenting T1ρ Magnetic Resonance Sequences of Talar Articular Cartilage

Abstract

Context: Quantifying early posttraumatic ankle osteoarthritis pathogenesis using compositional magnetic resonance (MR) imaging sequences is becoming more common. These MR sequences are often manually segmented to isolate the cartilage of interest before cartilage compositional values (eg, T1ρ or T2) are quantified. However, limited information is available regarding the reliability and reproducibility of manual segmentation for the entire talar dome. Objective: The purpose of this study was to determine the intraobserver and interobserver reliability of manually segmenting T1ρ MR sequences of the entire talar dome and 4 subregions of interest. Design: Descriptive observational study. Setting: Laboratory. Patients or Other Participants: Ten uninjured healthy individuals (4M and 6F: 21.40 [3.03] y, 170.00 [7.93] cm, 71.03 [14.97] kg) participated. Intervention: None. Main Outcome Measures: Two investigators manually segmented 10 T1ρ ankle MR sequences using ITK-SNAP software to calculate T1ρ mean relaxation times and cartilage volumes. Each observer repeated the segmentation twice, with segmentations separated by 1 month. Intraobserver and interobserver reliability was determined using intraclass correlation coefficients (ICCs) with 95% confidence intervals and root mean square coefficient of variations (RMSCVs). Results: For T1ρ relaxation time, intraobserver (ICC = .994–.997, RMSCV = 1.31%–1.51%) and interobserver reliability (ICC = .990, RMSCV = 2.36%) was excellent for the overall talar dome. Excellent intraobserver (ICC = .975–.980, RMSCV = 3.88%–4.59%) and excellent interobserver reliability (ICC = .970, RMSCV = 5.13%) was noted for overall talar cartilage volume. Conclusions: The results demonstrate that manual segmentation of the entire talar dome from a T1ρ MR is reliable and repeatable.