Interobserver Variation in Cerebellar Tonsillar Tip Localization

Autor: Andreea G. Moore, Diana Gomez-Hassan, Hugh J. L. Garton, Douglas J. Quint, Cormac O. Maher, Ashok Srinivasan
Rok vydání: 2016
Předmět:
Zdroj: Journal of Computer Assisted Tomography. 40:34-38
ISSN: 0363-8715
DOI: 10.1097/rct.0000000000000326
Popis: PURPOSE The aim was to evaluate the interobserver agreement in the assessment of cerebellar tonsil position on sagittal magnetic resonance imaging using 3 different osseous landmarks. MATERIALS AND METHODS This retrospective study consisted of brain magnetic resonance imagings performed at our institution in patients with and without Chiari I malformation between January 2010 and 2012. Sagittal T1-weighted images were reviewed by 2 senior board-certified neuroradiologists (blinded to underlying clinical diagnosis) with measurement of both cerebellar tonsillar positions based on lines drawn perpendicular from the tonsillar tip to the foramen magnum [FM] line, C1 line, and C2 line. Spearman correlation coefficients were calculated. Interobserver variation between the readers was assessed using Bland-Altman analysis and intraclass correlation coefficient. RESULTS A total of 320 cerebellar tonsils on 160 patients, 50 with Chiari I malformations, and 110 control subjects without Chiari I malformation were evaluated. The Spearman correlation coefficients for the entire cohort were 0.86 (FM), 0.94 (C1), and 0.90 (C2). Bland-Altman analysis for the entire cohort showed the best interobserver agreement for C1 line (-0.3 mm bias) and the least for C2 line (4.6 mm bias). The Intraclass correlation coefficients for all patients were 0.84 (FM), 0.92 (C1), and 0.54 (C2). The least bias and highest correlation coefficients were also seen individually in the Chiari and non-Chiari cohorts with the C1 technique. CONCLUSIONS Determination of cerebellar tonsillar position using a C1 arch landmark may be superior to the currently more commonly used FM-based landmark with lesser interobserver variability and higher interobserver correlation.
Databáze: OpenAIRE