Program Information
Robust Histogram-Based Analysis of Uniformity and Ghosting in MR Quality Control
M Siebert*, K Krugh , University of Toledo Medical Center, Toledo, OH
Presentations
SU-K-708-3 (Sunday, July 30, 2017) 4:00 PM - 6:00 PM Room: 708
Purpose: Current methods to analyze image uniformity and signal ghosting in MR quality control images can be highly labor intensive and yield results that vary significantly with size and location of the region-of-interest measurements. This study proposes use of histogram-based metrics to provide a robust automated measure of image uniformity and ghosting.
Methods: Eighteen phantom images acquired for MR coil quality control were analyzed using our automated histogram-based method and the conventional method as described by ACR. The images represented a variety of coil types and three different vendors. Our automated histogram-based method was incorporated into an ImageJ macro file that involved: 1) segmentation of the image into the phantom region, the background phase-encode region, and the background frequency-encode region, and 2) extraction of the signal histogram from each of the three regions. Using the phantom-region histogram the image uniformity was assessed by taking the signal values at a given percentile rank. The signal ghosting was assessed by taking the histogram intersection between the background phase- and frequency-encode histograms. These metrics were compared with the conventional ACR analysis method results.
Results: For image uniformity it was found that using the signal values at the 2nd and 98th percentile yielded the results closest to the ACR analysis method. When the 2nd/98th percentile uniformity values were normalized to the ACR uniformity values the average across the 18 images was 1.013 +/- 0.063. For signal ghosting the histogram intersection was found to have reasonable agreement with the ACR analysis values. The trend demonstrates a histogram intersection of 0.2 correlates to a 1.0% ghosting ratio.
Conclusion: The histogram-based analysis method produced robust results as demonstrated by close agreement to the ACR analysis results. The histogram-based method was easily automated and eliminates time intensiveness and user variability of results.
Contact Email: