IEC 62464-2: Magnetic Resonance Equipment — Image Quality Measurement — Part 2

IEC 62464-2, published in 2010, is the second part of the IEC 62464 series that establishes standardized methods for measuring key image quality parameters of magnetic resonance imaging (MRI) systems used in clinical applications. While Part 1 of this series defines the essential image quality parameters themselves, Part 2 provides the specific measurement protocols, phantom requirements, and data analysis procedures that enable reproducible and comparable image quality assessments across different MRI systems, sites, and time points. As MRI technology has evolved from 1.5 Tesla to 3 T, 7 T, and beyond, and with the proliferation of advanced imaging techniques such as parallel imaging, compressed sensing, and artificial intelligence-based reconstruction, standardized image quality assessment has become increasingly critical for ensuring consistent diagnostic performance.

The standard addresses a fundamental challenge in MRI quality assurance: while image quality in MRI depends on a complex interplay of static magnetic field strength, gradient performance, radiofrequency coil characteristics, pulse sequence parameters, and reconstruction algorithms, there must be objective, quantitative methods to verify that a system is performing within its specifications. IEC 62464-2 provides these methods, enabling clinical engineers, medical physicists, and service engineers to perform consistent acceptance testing, constancy testing, and cross-site comparisons using standardized phantoms and analysis workflows.

Key Image Quality Parameters and Measurement Methods

The standard defines precise measurement protocols for six fundamental image quality parameters. Signal-to-noise ratio (SNR) is measured using both the difference method (subtracting two identically acquired images to eliminate structured noise) and the single-image region-of-interest method for comparison. The measurement uses a uniform phantom filled with doped water (typically 1.25 g NiSO₄ x 6H₂O + 5 g NaCl per 1000 g H₂O) to provide a stable, homogeneous signal source with T1 and T2 relaxation times representative of clinical tissue. SNR is defined as the mean signal intensity in a central region-of-interest divided by the standard deviation of the background noise, corrected for the Rayleigh distribution of magnitude noise.

Image uniformity measures the spatial variation of signal intensity across the field of view in the absence of any structure. The standard specifies two measurement approaches: integral uniformity (the ratio of maximum to minimum pixel values across the entire image) and the more robust root-mean-square uniformity calculated using the NEMA standard approach. For a properly tuned 1.5 T system with a quadrature body coil, integral uniformity of better than 80% is expected over the central 80% of the field of view, while surface coil images will inherently show lower uniformity due to their sensitivity profile.

Spatial resolution is characterized using a line-pair phantom or by measuring the modulation transfer function from a sharp edge. The standard distinguishes between in-plane resolution (determined primarily by the image matrix size, field of view, and k-space trajectory) and through-plane resolution (determined by the slice selection gradient and RF pulse bandwidth). For clinical imaging, the standard recommends verifying that the limiting spatial resolution meets the manufacturer specification, typically 0.5-1.0 mm FWHM for in-plane resolution on modern 1.5 T and 3 T systems. Ghosting artifacts, caused by patient motion, physiological motion, or system instabilities, are quantified as the ratio of the mean ghost signal intensity to the mean object signal intensity. The measurement is performed using a uniform phantom positioned off-center to maximize the effect of gradient-related instabilities.

Phantom Specifications and Data Analysis

The standard specifies detailed requirements for test phantoms to ensure reproducible measurements across different sites. The uniform phantom must have an inner diameter of at least 200 mm (for head coils) or 400 mm (for body coils), with wall thickness less than 3 mm to minimize susceptibility artifacts. The phantom filling material must have T1 relaxation time between 200-600 ms and T2 relaxation time between 100-400 ms at the measurement field strength, mimicking the relaxation properties of human brain tissue. The standard provides specific recipes for doped water solutions that achieve these relaxation characteristics at both 1.5 T and 3 T field strengths.

Data analysis protocols are specified in detail to minimize operator-dependent variability. The SNR measurement using the difference method requires two identical acquisitions with the same receiver gain, bandwidth, and reconstruction parameters. The subtraction image eliminates anatomical or phantom structure, leaving only random noise from which the standard deviation is measured. A correction factor of 0.655 (1/√(2-π/2)) is applied to account for the Rayleigh distribution of magnitude noise in the subtraction image. The standard requires that at least five independent ROI measurements be performed and averaged to obtain a reliable SNR estimate.

Engineering Design Insights for MRI Quality Assurance

From an engineering perspective, the measurement protocols in IEC 62464-2 serve not only as acceptance criteria but as diagnostic tools for identifying specific system faults. For instance, a decrease in SNR without change in uniformity typically indicates preamplifier noise figure degradation or receiver chain attenuation, while reduced SNR accompanied by decreased uniformity often points to RF coil element failure or cable connection issues. Asymmetric uniformity loss localized to one quadrant of the image strongly suggests a specific coil element fault in array coils, enabling targeted service intervention without exhaustive troubleshooting.

Ghosting ratio measurements are particularly valuable for assessing system stability over the timescale of a single acquisition. Elevated ghosting can arise from gradient amplifier nonlinearities, main field (B0) drift during the scan, or eddy current effects from rapidly switched gradients. By analyzing the spatial distribution of ghosting (whether it appears as a central line through the image or as displaced replicas), experienced engineers can differentiate between B0 drift (central ghost) and gradient timing errors (displaced ghost). This diagnostic capability is increasingly important for high-performance imaging applications such as diffusion tensor imaging and functional MRI, where ghosting levels below 1% are typically required for reliable quantitative analysis.

Geometric distortion measurement is becoming more critical with the growing use of MRI for radiotherapy treatment planning and image-guided interventions. The standard requires that geometric accuracy be verified at least at nine positions across the field of view, with measurements along both frequency-encoding and phase-encoding directions to identify gradient scaling errors and B0 inhomogeneity effects. For hybrid PET/MR systems, geometric accuracy requirements are particularly stringent, as MR-based attenuation correction maps must be geometrically accurate to prevent misregistration and quantification errors in the PET reconstruction.

The standard also addresses the growing importance of parallel imaging techniques (such as GRAPPA and SENSE). IEC 62464-2 recommends that SNR and ghosting measurements be performed both with and without parallel imaging acceleration to separately evaluate the base system performance and the parallel imaging reconstruction quality. The g-factor (geometry factor), which quantifies the SNR penalty incurred by parallel imaging, should be measured for each acceleration factor and compared to the design value. An unexpectedly high g-factor may indicate suboptimal coil element geometry, improper reference scan data, or reconstruction algorithm issues that need engineering attention.