Effects of Normalization, Filtration, and Distortion Correction Options On MRI ACR Phantom Images
N Busse*, L Page, P Hou, D Reeve, UT MD Anderson Cancer Center, Houston, TXSU-E-I-67 Sunday 3:00PM - 6:00PM Room: Exhibit Hall
Purpose: Toptimization of clinical MR image quality through protocol design on modern scanners requires understanding a wide variety of image acquisition options. Using ACR phantom analysis results for comparison, we sought to understand the effect of these options on the resulting image quality measurements.
Methods: Twenty-six series were acquired with different combinations of normalization, filtration, distortion correction, radiofrequency mode, and gradient mode using a Siemens Aera (Erlangen, Germany) 1.5T scanner with 20-channel head/neck phased-array coil. The ACR T1-weighted phantom protocol (transverse, TR=500 ms, TE=20 ms, 256X256 matrix, 5mm thickness and gap, 25 cm FOV) was used to scan a large ACR MRI phantom (J.M. Specialty Parts, San Diego, California). All images were acquired in the same session without repositioning the phantom. Automated ACR analysis was implemented in MATLAB (Mathworks, Natlick, Massachusetts) for most tests. The software results were compared to manual analysis for five series. High-contrast resolution, slice thickness, and low-contrast detectability (LCD) were evaluated manually for all scans following the recommendations in the ACR's Phantom Test Guidance manual (http://www.acr.org/~/media/ACR/Documents/Accreditation/MRI/LargePhantomGuidance.pdf).
Results: Scans without normalization failed ACR uniformity (1.5T) and slice thickness requirements and had lower LCD scores. Filtering with the Raw Filter degraded resolution to 1.1 mm and increased the ghosting ratio but resulted in decreased noise. On this scanner 2D (in-plane) distortion correction can be applied prospectively or retrospectively. Results showed that prospective correction provided better geometric accuracy performance. Images acquired using whisper and fast gradient modes, as well as fast and low SAR options, resulted in poorer LCD scores compared to normal gradient and radiofrequency modes.
Conclusion: MR protocol optimization requires compromise to achieve relative balance between signal, noise, uniformity, resolution, geometric accuracy and other image metrics. Phantom studies with quantitative metrics provide a robust method of determining how imaging options affect these image parameters.