ABMP MRI for Radiation Therapy cheat sheet

Larmor frequency: f0 = (γ/2π)·B0; for 1H, γ/2π ≈ 42.58 MHz/T, so 1.5 T → 63.9 MHz and 3 T → 127.7 MHz.

T1 relaxation is spin-lattice recovery (exponential regrowth of longitudinal magnetization); T2 is spin-spin decay of transverse magnetization; T2* adds static field inhomogeneity to T2 decay.

Signal-to-noise ratio scales approximately linearly with B0; chemical shift in Hz scales linearly with B0, so fat-water separation in pixels doubles from 1.5 T to 3 T for a fixed bandwidth.

Gradient coils encode spatial position: slice selection uses a gradient plus RF bandwidth, phase encoding uses amplitude steps, frequency encoding uses readout gradient; FOV = 1/(Δk) and Δx = FOV/N.

RF energy deposition (SAR) rises roughly with the square of B0 and with flip angle squared; SAR is reported in W/kg and is a primary constraint at 3 T.

A spin echo refocuses static off-resonance at TE = 2·(180° pulse spacing), while a gradient echo does not refocus static field inhomogeneity, producing T2* weighting.

Parallel imaging (SENSE/GRAPPA) reduces phase-encoding steps and readout time at the cost of SNR loss proportional to the square root of the acceleration factor and possible aliasing.

Shim coils correct B0 inhomogeneity; active shimming is required for large-FOV body imaging and for off-isocenter or large-patient geometry.

MR conditional labeling applies only to the specific field strength, coil configuration, and conditions in the manufacturer's labeling; device status outside those conditions is unknown.

Static field hazards include translational force, torque, and projectile risk; the 5-gauss line marks the fringe field boundary and must be controlled.

RF-induced heating risk scales with SAR and patient size; implanted leads and elongated conductors are the highest-risk configurations for RF burns.

Time-varying gradient fields can induce peripheral nerve stimulation (PNS); dB/dt limits and the vendor PNS monitor constrain fast sequences.

Acoustic noise is quantified by dB(A) and is highest with high-gradient-duty sequences such as EPI and 3D GRE; hearing protection is required.

Pregnancy: no confirmed harm at 1.5 T/3 T, but gadolinium contrast is generally avoided; screening for metallic implants and prior occupational exposure is mandatory before entering Zone III/IV.

Quench: superconducting magnet loses field rapidly; cryogen asphyxiation and cold-burn risk require oxygen monitoring and immediate evacuation, not re-entry.

Gadolinium in renal impairment: NSF risk is linked to older linear agents; macrocyclic agents are preferred, and eGFR screening applies to high-risk patients per institutional policy.

SNR ∝ voxel volume × sqrt(NEX) × sqrt(N_phase) / sqrt(bandwidth); doubling slice thickness or FOV increases SNR at the cost of resolution.

Chemical shift artifact at fat-water interfaces is proportional to B0 and inversely proportional to receive bandwidth; it appears as bright/dark bands along the frequency-encoding direction.

Susceptibility artifact from metal or air-tissue interfaces grows with TE, field strength, and echo-planar readout duration; spin echo and short TE reduce it.

Motion artifact: phase-encoding direction shows ghosting; swapping phase and frequency directions moves ghosts, and respiratory gating or averaging reduces them.

Geometric distortion from gradient nonlinearity is worst at the FOV periphery and off-isocenter; distortion correction is required for RT geometry.

QA program: daily SNR/uniformity/ghosting checks, monthly geometric accuracy and slice position, annual ACR-style phantom testing of spatial resolution, slice thickness, and low-contrast detectability.

Aliasing (wrap-around) occurs when the object exceeds the FOV in the phase-encoding direction; oversampling the frequency direction avoids it without time penalty.

B1 inhomogeneity at 3 T causes shading and flip-angle error, especially in body imaging; dielectric pads and B1 shimming mitigate it.

RT simulation requires a flat tabletop, external laser alignment, and an MR-compatible immobilization device indexed to the table for reproducible setup.

Patient position for simulation (supine, arms up, head-first) should match the treatment position; couch and coil geometry must be recorded and reproduced.

Scan protocols must produce geometrically accurate images: large FOV, thin slices, no distortion-prone EPI, and vendor distortion correction enabled.

DICOM export must carry correct slice spacing, orientation, and patient coordinates; the DICOM coordinate system origin is the scanner isocenter.

Multi-sequence acquisition (T1, T2, DWI, DCE) is fused to define GTV/CTV and OARs; a dedicated planning series is selected as the primary geometry.

Motion management at simulation: 4D-MRI or respiratory-correlated acquisition for thoracic/abdominal targets; breath-hold instructions should mirror treatment.

MR simulation workflow must reconcile MR coordinates with CT simulation coordinates through a common table position or fiducial-based registration.

Rigid registration uses 6 degrees of freedom (3 translation, 3 rotation); affine adds 3 scaling/shear terms; deformable registration allows local nonrigid warping.

Mutual information is the standard similarity metric for CT-MR fusion because it does not assume a linear intensity relationship between modalities.

Registration accuracy is assessed by target registration error (TRE) using landmarks; fiducial-based registration generally outperforms intrinsic registration for rigid fusion.

CT-MR fusion relies on bony landmarks or fiducials for the rigid component; soft-tissue deformation between scans is unmodeled and must be evaluated.

Deformable registration for MR-guided RT requires validation with a known deformation phantom and contour propagation review by the physician.

Registration errors propagate to dose: a 2 mm systematic error in the PTV margin can be clinically significant; margins must account for registration uncertainty.

DICOM registration objects (Spatial Registration, Deformable Registration) store the transform and should be archived with the plan.

MR lacks electron density; dose calculation requires either a CT for heterogeneity correction or a synthetic CT generated from MR (bulk-density or deep-learning methods).

MR excels at soft-tissue contrast for brain, prostate, and head-and-neck targets, improving GTV delineation and reducing interobserver variability.

Geometric distortion must be corrected before contouring; residual distortion should be quantified (typically <1-2 mm within the imaging FOV) and included in margins.

Contours drawn on MR must be transferred to the planning CT through validated registration; the contouring series should be the primary geometry.

Bone and air appear dark on most MR sequences; synthetic CT must accurately represent bone for dose calculation in heterogeneous regions.

MRI-guided planning for prostate and spine uses T2-weighted sequences; DWI aids target definition but has higher distortion and is generally not used for geometry.

PTV margins must include MR distortion, registration error, and setup error in addition to target motion.

MR-linac systems (e.g., 1.5 T or 0.35 T) integrate a linac with an MR scanner; the magnetic field deflects secondary electrons, altering surface and interface dose (electron return effect).

Daily MR-guidance allows online adaptive radiotherapy: recontouring and replanning based on the anatomy of the day, with plan QA before delivery.

The electron return effect increases dose at tissue-air interfaces and requires dose calculation algorithms that model magnetic field effects (Monte Carlo or explicit Lorentz correction).

Cine MR enables real-time target tracking for gating and beam-hold delivery; latency between imaging and beam control must be characterized.

MR-linac QA includes image quality, geometric accuracy, and dosimetric verification with the magnetic field present; standard linac QA phantoms may not be MR-compatible.

Treatment delivery with the magnetic field requires MR-safe immobilization and coils that do not attenuate or perturb the beam.

ALARA: occupational dose limits are 20 mSv/y averaged over 5 years (50 mSv in any single year) for radiation workers; public limit is 1 mSv/y.

MRI staff are not exposed to ionizing radiation from the scanner but may be exposed from the linac in combined MR-linac suites; dosimetry badges are required.

The 5-gauss line defines the controlled MR environment; access control and signage (Zone I-IV) are required per ACR guidance.

Time, distance, and shielding govern radiation protection: dose rate from a point source falls as 1/r².

Professional practice: treatment plans and adaptive replans must be reviewed and approved by a radiation oncologist and physics before delivery.

Documentation: registration transforms, contouring series, and adaptive replans must be archived in the record-and-verify system for auditability.