Quantitative magnetic resonance first-pass perfusion analysis: inter- and intraobserver agreement

Abstract

Magnetic resonance first-pass (MRFP) imaging awaits longitudinal clinical trials for quantification of myocardial perfusion. The purpose of this study was to assess inter- and intraobserver agreement of this method. Seventeen MRFP studies (14 rest and 3 under adenosine-induced hyperemia) from 14 patients were acquired. Two observers visually graded study quality. Each study was subdivided into eight regions. Both observers analyzed all 17 studies (8 x 17 = 136 regions) for interobserver agreement. Each observer then analyzed 10 of the 17 studies a second time (2 x 8 x 10 = 160 regions) for intraobserver agreement. Signal intensity curves were obtained with Argus software (Siemens, Iselin, NJ). The maximum amplitude of the impulse response function (Rmax) and the change of signal intensity (deltaSImax) of the contrast bolus were determined. Intraclass correlation coefficient was used to determine intra- and interobserver agreement. The quality was good or excellent in 14 studies. Intraobserver agreement of Rmax and deltaSImax were good (0.85 and 0.80, n = 160). Interobserver agreement of Rmax was fair (0.55, n = 136) but improved after exclusion of poor-quality studies (0.88, n = 112). Interobserver agreement of deltaSImax was good (0.73) and improved less than Rmax with study quality (0.83). Interobserver agreement for Rmax in individual myocardial regions before and after exclusion of studies with poor quality changed most markedly in lateral and posterior regions (0.69 and 0.65 vs. 0.97 and 0.94), where signal-to-noise ratios were reduced compared with anteroseptal regions (p < 0.01). Analysis of MRFP images provides good intraobserver agreement. Interobserver agreement of the quantitative perfusion analysis is good under the premise of good image quality.