Automatic detection of small bowel contraction frequencies in motility plots using lomb-scargle periodogram and sinus-fitting method—initial experience
Contraction frequencies are an important parameter for the analysis of bowel motility in MRI. The contraction curve can be rather noisy and the frequency-evaluation might be difficult. The aim was to evaluate manual calculations of small bowel contraction frequency in comparison with automatic calculations using two mathematically established methods.
The institutional-review-board approved study consisted of 48 segmental small-bowel motility-plots out of a previous study with 25 patients (15 men/10 women; mean 39 years, standard deviation ± 14.67) undergoing MRI (1.5 T, GE-Medical Systems; two-dimensional-Fiesta, pulse repetition time 2.91, echo time 1.25, flip-angle 45°, matrix 256 × 256, slice thickness 10 mm) were evaluated. Calculations of contraction frequency was performed either manually or using Lomb-Scargle-periodograms and Sinus-Fitting method. The results were compared using intraclass correlation coefficient, Friedman's-test and Wilcoxon-matched-paired-signed-rank-test (P < 0.05 considered statistically significant).
Mean contraction frequency was 5.69 (standard deviation ± 1.99) for manual calculations showing moderate interreader reliability (intraclass correlation coefficient 0.72; 95% confidence-interval: 0.59;0.82). No significant difference (P = 0.776) was found to Lomb-Scargle-periodograms (5.34 contractions-per-minute [cpm], standard deviation ± 2.47; P = 0.174) and to the sinus-fitting-method (5.47 cpm; standard deviation ± 2.57; P = 0.0779). For the latter calculations failed in 3/48 plots, while manual and Lomb-Scargle-periodograms could be performed successfully in all 48/48 plots.
Herein, we could show that Lomb-Scargle-periodograms and Sinus-Fitting are fast and reliable methods to automatically measure small-bowel contraction frequencies even in noisy small bowel motility plots. Magn Reson Med 71:628–634, 2014. © 2013 Wiley Periodicals, Inc.