ABMP MRI for Radiation Therapy common mistakes
ABMP MRI for Radiation Therapy · Understand the traps and how to avoid them.
Treating the 0.35 T on-board treatment MRI as if it followed the same image-quality and safety rules as a 1.5 T or 3 T diagnostic scanner, e.g. assuming high SAR is the main constraint or that chemical-shift artifacts scale the same way.
Candidates memorize diagnostic-field physics and apply it unchanged; low-field systems have different T1/T2 relaxation behavior, reduced SAR, different SNR and fat-water separation behavior, and the hybrid system has unique constraints (split gradient coils, linac interference, attenuation of the beam by the system).
What to do instead
Study the physics of low-field imaging separately: how SNR, T1/T2 contrast, SAR, and susceptibility scale with B0, and how the integrated MR-linac hardware changes acquisition and safety limits. Practice explaining why a sequence chosen at 1.5 T may need redesign at 0.35 T.
Confusing the MR safety zones (I-IV) and who is authorized in each, or assuming MR-safe means MR-conditional without checking the specific conditions.
Zone definitions and the safe/conditional distinction are tested as discrete facts, and candidates blur 'MR safe,' 'MR conditional,' and 'MR unsafe' or forget that conditions include field strength, SAR limits, and specific absorption conditions.
What to do instead
Memorize the four-zone model and the ASTM terminology triad with examples. For any implanted device question, state the field strength and conditions explicitly before declaring it acceptable.
Describing geometric distortion correction as a single step, without distinguishing gradient nonlinearity distortion from susceptibility-induced distortion and without stating how each is corrected or mitigated.
Both appear as 'distortion' on images, so candidates give a generic answer. Gradient nonlinearity is system-related and corrected by vendor distortion maps/spherical harmonic correction; susceptibility distortion is patient/tissue-related and addressed by sequence choice, bandwidth, and shim, not by the same correction.
What to do instead
Build a two-column comparison: cause, spatial pattern, whether it is correctable by system calibration, and the practical mitigation for treatment planning (e.g. large-FOV distortion phantom QA, avoiding echo-planar readouts for planning scans).
Ignoring the difference between scanner coordinates and treatment-room coordinates, and describing registration accuracy only as 'visual check' or 'mutual information is accurate.'
Candidates conflate image registration with geometric alignment to the treatment isocenter; they forget that the MR-to-CT and MR-to-treatment coordinate transforms are separate, and that registration accuracy must be verified with a physical phantom or fiducials in room coordinates.
What to do instead
Draw the transform chain explicitly: MR patient coordinates to CT (registration) to planning isocenter to treatment-room coordinates (couch shifts, on-board imaging). State which step each QA test validates and the tolerance associated with it.
Assuming rigid registration is always adequate for MR-CT fusion in the pelvis or thorax, or applying deformable registration without a validation strategy.
Rigid registration is easy to defend and candidates default to it, but organ motion, bladder/rectal filling, and breathing cause non-rigid differences; conversely, DIR can introduce implausible deformations that are hard to detect.
What to do instead
For each anatomical site, state whether rigid or deformable registration is appropriate and why, and describe the validation method (fiducial or landmark agreement, Jacobian inspection, visual review of contours on both datasets, independent second observer).
Treating MR-derived electron density as equivalent to CT Hounsfield units, or failing to state how dose calculation handles the absence of electron density in MRI.
MRI signal is not a function of electron density; candidates either assume a bulk-density override is universally acceptable or assume a synthetic CT is always accurate, without discussing conversion methods and their validation.
What to do instead
Know the options: CT-MR fusion with CT-based dose calculation, bulk density assignment by tissue class, atlas- or deep-learning-based synthetic CT, and MR-only workflows. For each, state the required commissioning/validation and the tissues where errors are largest (bone, air cavities, lung).
Answering QA questions with 'follow the manufacturer's protocol' or 'daily QA' without naming the specific test, phantom, frequency, and tolerance.
QA is often studied passively; candidates know QA exists but not the test-tolerance-frequency triad, so they give generic answers that earn no credit.
What to do instead
Build a table of tests (geometric accuracy/distortion, SNR, uniformity, ghosting, slice thickness, center frequency, laser alignment, couch position) with phantom, frequency, and typical tolerance for both diagnostic and MR-sim/on-board systems. Be able to justify each tolerance.
Describing RF safety using only whole-body SAR, ignoring local SAR, B1+rms, implant heating, and the effect of transmit coil and patient positioning.
Whole-body SAR is the most familiar number on the console, so candidates stop there and miss that local SAR and B1+rms can exceed limits while whole-body SAR appears normal, and that implants create local heating risk not captured by SAR at all.
What to do instead
State the hierarchy: whole-body SAR, local SAR (per mass of tissue), B1+rms as the sequence-level metric, and implant-specific heating. For implant cases, describe the role of MR-conditional conditions, transmit/receive coil choice, and the need to consult device labeling.
Overlooking the effect of patient setup and immobilization on MR image quality and geometric fidelity - e.g. using the treatment immobilization device in the scanner without checking whether it fits the bore or causes susceptibility/SNR issues, or forgetting that immobilization reproducibility must be verified in the treatment position.
Candidates focus on sequence parameters and forget that the MR-sim workflow must reproduce treatment positioning; coil placement, immobilization, and table overlay affect both image quality and the validity of the planning scan.
What to do instead
Walk through the MR-sim workflow end to end: immobilization compatibility with the bore and coils, setup reproducibility, laser/tattoo alignment, scan extent and coverage of the target plus margin, and documentation of position. Identify where errors can enter and how each is checked.
Giving a vague answer on motion management in on-board MRI - saying 'use gating' without specifying how the MR signal is used as a surrogate, how latency is handled, and how the beam gate is synchronized with the imaging.
The MR-linac workflow is unfamiliar and candidates know the concept of gating but not the practical chain: cine imaging, surrogate selection, latency, beam-on/off timing, and verification of target coverage during delivery.
What to do instead
Describe the full loop: fast 2D cine for surrogate tracking, the surrogate-to-target relationship established at simulation, system latency and its dosimetric consequence, and how delivery is verified (e.g. dosimetric reconstruction, end-of-treatment imaging). Be able to discuss when gating is preferred over breath-hold or tracking.