Purpose: To determine the ‘best’ method for measuring RF coil SNR as part of the annual MRI Performance Evaluation.
Methods: Since April, 2014, as part of over 800 MRI yearly performance evaluations of systems from every major vendor, we have collected roughly 25,000 measurements of RF Coil SNR data. All calculations were performed using four methods of estimating the background noise level, sigma, by measuring 1) the Standard Deviation (SD), 2) the Mean, or 3) the Root Mean Square (RMS) of the background ‘air’ pixels or by 4) using the NEMA subtraction method. To compare the four methods, we looked at 10 combinations of coils, phantoms and planes chosen for the ease of reproducibly setting up from year to year. The phantoms were all standardly available with each system. A total of 1513 measurements obtained from 196 individual coils. Studies where the coil was identified as having a problem were excluded. The data was analyzed both by aggregating all of the measurements for on coil/phantom class and by looking at how each individual coil changed from year to year.
Results: The average of the standard deviations across the 10 combinations when the data was aggregated were 21.8%, 8.0%, 6.6% and 6.3% for the NEMA, S.D., RMS and Mean methods respectively. When looking at each individual coil over time the average of the SDs were 18.1%, 4.3%, 2.91% and 2.90% in the same order as above. The smallest range of values was obtained with the Head coils and the ACR phantom while the largest was with the long cylinder in the Body 18 coils.
Conclusion: The least reliable method, by far, of measuring SNR is to use the NEMA subtraction method. Measuring the mean of the noise is on average about 30% better than measuring the S.D. of the background noise.
Funding Support, Disclosures, and Conflict of Interest: Simply Physics provides MRI QC and ACR accreditation services. All of the data presented here were acquired as part of that service.
Not Applicable / None Entered.