ABMP MRI Physics common mistakes
ABMP MRI Physics · Understand the traps and how to avoid them.
Explaining T1 and T2 relaxation by saying the magnetization 'grows back' or 'decays away' without distinguishing longitudinal recovery from transverse dephasing, and treating T2* as interchangeable with T2.
The two processes are taught side by side and both are exponential, so candidates collapse them into one 'relaxation' concept and forget that T1 concerns Mz recovery toward equilibrium while T2 concerns loss of Mxy phase coherence from spin-spin interactions.
What to do instead
For every relaxation question, state the component (Mz or Mxy), the time constant, and the mechanism. Reserve T2* for combined spin-spin plus static field inhomogeneity dephasing and note that T2* <= T2.
Computing scan time or predicting contrast from TR and TE without first identifying the pulse sequence type and the number of phase-encoding steps, then applying a formula blindly.
Candidates memorize 2D spin-echo time as TR x phase-encoding steps x averages and then misapply it to gradient-echo, fast spin-echo with echo train length, or 3D acquisitions where the formula differs.
What to do instead
Before calculating, write down the sequence, dimensionality, and whether multiple lines are acquired per TR. For FSE divide the phase-encode count by echo train length; for 3D multiply by the number of partitions.
Describing spatial encoding as gradients 'selecting' a location in the image rather than mapping position to precessional frequency or phase, and confusing the roles of the slice-select, phase-encode, and readout gradients.
The three gradients are introduced together and the shorthand 'the gradient picks the slice' obscures that the readout gradient encodes frequency during acquisition, the phase-encode gradient imposes position-dependent phase before readout, and the slice gradient selects by frequency during RF excitation.
What to do instead
For each gradient axis, state what it does, when it is applied relative to RF and readout, and what physical quantity (frequency or phase) encodes position. Tie slice thickness to RF bandwidth and gradient amplitude.
Treating SNR as improved by any coil with more elements or a higher field, without accounting for coil loading, fill factor, receive bandwidth, and the difference between quadrature and phased-array reception.
Surface coils give high SNR only near the coil, phased arrays trade per-element coverage for combined SNR, and increasing receive bandwidth lowers SNR. Candidates remember 'more elements equals better' and ignore geometry and noise.
What to do instead
Evaluate SNR questions by naming the noise source, the coil-to-sample proximity, and the bandwidth. State that surface coils have high SNR near the coil with rapid falloff, and that phased arrays combine element signals for wider coverage.
Confusing aliasing (wrap-around) with chemical shift, ghosting, and motion artifact when reading an artifact scenario, and proposing a fix that addresses the wrong mechanism.
All four appear as displaced or repeated structures, and candidates pattern-match on appearance rather than on the acquisition parameter (FOV, bandwidth, phase direction, motion) that produces each one.
What to do instead
Classify the artifact by its signature: wrap-around repeats the anatomy outside the FOV, chemical shift displaces fat relative to water along the frequency direction, ghosting repeats along phase-encode from periodic motion, and zipper artifacts come from RF interference. Then match the remedy to the mechanism.
Assuming parallel imaging or compressed sensing always shortens scan time without noting the SNR penalty, the need for coil sensitivity calibration, and the g-factor and acceleration limits.
Acceleration factor is presented as a free speedup, so candidates forget that undersampling reduces SNR by roughly the square root of the acceleration factor and that geometry factor worsens with higher acceleration.
What to do instead
For every fast-imaging question, state the acceleration factor, the SNR penalty, the calibration requirement, and the artifact risk (residual aliasing). Note that compressed sensing relies on sparsity and incoherence rather than coil sensitivity alone.
Calculating SAR or RF heating without distinguishing whole-body, head, and local SAR, and ignoring that SAR scales with the square of B1 and inversely with duty cycle and patient size.
Candidates memorize a single SAR limit and apply it to all coils and body regions, missing that local transmit coils, high duty cycle sequences, and larger patients change the relevant limit and heating pattern.
What to do instead
Identify the transmit configuration, the applicable SAR averaging (whole-body vs local), and the sequence duty cycle. State that SAR rises with B1 squared, flip angle, and duty cycle, and that local SAR governs surface transmit coils.
Treating the MR safety zones, the static-field projectile hazard, and the RF burn hazard as one undifferentiated 'safety' topic, and misstating which implants or conditions are absolute contraindications.
Safety content is often studied as a checklist, so candidates blur zone boundaries, the difference between MR-safe, MR-conditional, and MR-unsafe labeling, and the specific mechanisms (torque, heating, induced currents) behind each hazard.
What to do instead
For each question, name the hazard mechanism, the zone or condition it applies to, and the device labeling category. Distinguish static-field (projectile, torque), gradient (induced currents, acoustic noise, PNS), and RF (heating, burns) hazards explicitly.
Accepting a phantom or QA result as 'passing' without knowing the tolerance, the reference baseline, or which parameter the test actually measures (e.g., treating a uniformity scan as a geometric accuracy test).
QA tables are memorized as pass/fail lists, so candidates cannot link a specific test (SNR, uniformity, geometric distortion, slice thickness, ghosting) to the parameter it validates or to the acceptance criterion.
What to do instead
For each QA test, state what is measured, the phantom used, and the typical tolerance. Tie each test to the failure it would catch (e.g., geometric accuracy to gradient calibration, uniformity to coil or B1 problems).
Reporting quantitative values such as T1, T2, ADC, or perfusion without stating the acquisition model, fitting assumptions, or confounds like B1 inhomogeneity, motion, and partial volume.
Quantitative methods are taught as formulas, so candidates quote numbers without recognizing that the estimate depends on the sequence, fit model, and corrections applied.
What to do instead
For quantitative questions, name the acquisition (e.g., inversion recovery, multi-echo, diffusion-weighted), the fitting model, and the main confounds. State that B1 mapping, motion correction, and partial-volume effects must be addressed for reliable values.