MRI Sequence & Protocol Optimiser
An interactive guide to tuning MR sequences. Start from a problem, explore what each parameter trades off, or tune a protocol toward a goal. Terminology adapts to your scanner vendor.
On this page
Problem‑solverParameter trade‑offsProtocol tunerAdvanced optimisation techniquesWhen to review images mid-protocol and the ideal sequence orderSources- To speed a scan, look first at TR, phase-encode steps, averages, parallel imaging (which always costs SNR) and, in TSE/FSE, echo-train length; frequency oversampling does not change scan time if TR is unchanged.
- SNR ∝ voxel volume × √(total sampling time), so doubling averages gives about 1.4× SNR but doubles scan time.
- Near metal, use TSE/FSE not gradient echo, high bandwidth and STIR not spectral fat-sat, escalating to a metal-artefact-reduction sequence; artefact strategy is not an implant safety clearance.
- Review each series while the next one runs: put decision-relevant and motion-sensitive sequences early, run the long 3D last as the review window, and never stop the table to look.
- Motion ghosts propagate along the phase-encode axis, so swapping the phase- and frequency-encode directions can move them off the anatomy of interest, at the risk of wrap.
Explain it simply
- To shorten a scan, change the repeat time, the number of data lines, repeats and echoes per repeat, or speed-up settings, which always add graininess. Extra sampling along each line usually adds no time.
- Bigger image blocks (voxels) and longer data collection make images less grainy. Doubling the repeats (averages) doubles the scan time but improves signal-to-noise only about 1.4 times.
- Near metal implants, staff choose scan types and settings that limit the distortion metal causes, and special metal-reduction scans if needed. This is not a safety clearance for the implant, which is checked separately.
- Staff check each set of images while the next one is being taken. Important and movement-sensitive scans go first, the long detailed scan goes last, and the scanner never sits idle while they look.
- Movement causes image faults called ghosts that spread along one direction of the image. Swapping the image directions can move them away from the area of interest, but may cause wrap-around.
A guide only: follow your departmental protocols and regional requirements.
Problem‑solver
What’s wrong with the image?
Expand a symptom to see fixes ranked from lowest-cost/highest-yield first, each with what it trades away.
Image looks grainy / low signal-to-noiseSNR
1Increase slice thickness
SNR scales roughly linearly with voxel volume, so a thicker slice raises SNR in proportion, with no added scan time in 2D.
Trades: through-plane resolution & more partial-volume averaging.
2Add signal averages (NEX/NSAAveragesNEXNSANAQ)
SNR rises with the square root of averages: doubling averages gives ~1.4× SNR.
Trades: scan time rises linearly (2× averages = 2× time).
3Lower receiver bandwidth
Narrower bandwidth samples signal for longer, improving SNR (∝ 1/√BW).
Trades: more chemical-shift artifact; longer minimum TE.
4Use a coil with more/closer elements
A dedicated surface/phased-array coil close to the anatomy dramatically improves SNR versus the body coil.
Trades: setup time; signal drop-off with depth; possible coil-shading.
5Increase field-of-view or reduce matrix
Larger voxels in-plane raise SNR.
Trades: in-plane spatial resolution, often the wrong compromise for small structures.
Blurring / ghosting from motionMotion
1Shorten the acquisition
Cut phase-encode steps, add parallel imaging, or use a faster sequence (single-shot/turbo).
Trades: SNR and/or resolution.
2Swap phase- and frequency-encode direction
Ghosts propagate along the phase-encode axis. Reorient so ghosts fall away from the anatomy of interest.
Trades: may introduce wrap/aliasing; sometimes lengthens scan.
3Respiratory/cardiac gating or triggering (navigators / triggeringPACE / navigatorNavigator / respiratory triggeringTrigger / navigatornavigator)
Synchronise acquisition to the respiratory or cardiac cycle for body/cardiac work.
Trades: longer, less predictable scan time; efficiency drops with irregular breathing.
4Radial / PROPELLER-type acquisition (radial bladesBLADEPROPELLERMultiVaneJET)
Oversampled central k-space with rotating blades averages out bulk motion and enables correction.
Trades: longer scans; its own blurring/streak artifacts.
5Saturation bands & flow compensation
Place saturation regions over moving/flowing tissue; add gradient moment nulling for flow ghosts.
Trades: extra RF (SAR), longer TR, possible saturation of wanted signal.
Signal void / distortion near metalMetal
1Increase receiver bandwidth
High bandwidth compresses the frequency-shift distortion from susceptibility. It is the first thing to reach for.
Trades: lowers SNR (∝1/√BW).
2Prefer fast/turbo spin-echoTSEFSETSEFSE over gradient-echo
The 180° refocusing pulse recovers susceptibility-induced dephasing that GRE cannot. Avoid GRE near metal.
Trades: higher SAR; less T2* sensitivity (usually desirable here).
3Use a dedicated metal-artifact-reduction sequence (VAT + slice-encodingWARP (VAT + SEMAC)MAVRIC SLO-MAR / MSVAT-SPACEmetal-reduction package)
Slice-encoding (SEMAC) and multi-spectral (MAVRIC) methods correct both in-plane and through-plane distortion.
Trades: substantially longer scan times.
4Use STIR instead of spectral fat-sat
Field inhomogeneity near metal breaks chemical-shift-selective fat suppression; STIR’s inversion-based suppression is far more robust.
Trades: lower SNR; suppresses any short-T1 tissue (not fat-specific).
5Thinner slices & smaller voxels; align frequency-encode along the implant’s long axis
Smaller voxels reduce through-plane dephasing; orientation choices minimise the distorted region.
Trades: SNR and scan time.
Scan is too longTime
1Parallel imaging (SENSE/GRAPPAGRAPPA / mSENSEASSET / ARCSENSESPEEDER)
Acceleration factor R cuts phase-encode steps and time by ~R.
Trades: SNR drops by 1/√R plus a spatially-varying g-factor penalty.
2Raise the turbo/echo-train length (ETL / turbo factorTurbo factorETLTSE factorETL)
More echoes per TR fills k-space faster in TSE/FSE.
Trades: T2 blurring, altered contrast, higher SAR.
3Reduce phase-encode steps / use partial Fourier
Rectangular FOV, lower phase matrix, or partial-Fourier (half-scanHalf FourierFractional NEXHalf scanhalf scan) all trim acquisition.
Trades: resolution and/or SNR; some blurring with partial Fourier.
4Shorten TR (for T1/PD) or drop averages
TR is a direct multiplier on scan time.
Trades: less T2 weighting, fewer slices per TR, lower SNR.
5Deep-learning reconstruction / simultaneous multi-slice
DL-recon (DL reconDeep ResolveAIR Recon DLSmartSpeed / Compressed SENSEAiCE) and simultaneous multi-slice (SMSSimultaneous Multi-SliceHyperBandMB-SENSEmultiband) give large time savings while defending SNR.
Trades: availability/licensing; recon behaviour to validate before clinical reliance.
Poor tissue contrast / lesion not conspicuousContrast
1Match TR/TE to the weighting you want
T1: short TR/short TE. T2: long TR/long TE. PD: long TR/short TE. Drifted contrast is usually a TR/TE mismatch.
Trades: longer TR = more time; longer TE = less SNR.
2Add fat suppression appropriate to the field
Spectral fat-sat for conspicuity where B0 is homogeneous; STIR or Dixon where it isn’t.
Trades: SNR (STIR), time (Dixon), SAR.
3Use an inversion-recovery prep to null a tissue
FLAIR nulls CSF; STIR nulls fat; choose TI to null the unwanted signal (TI ≈ 0.69×T1 when TR is much longer than T1; a shorter TR needs a shorter TI).
Trades: longer TR, lower SNR, more time.
4Optimise flip angle (GRE) toward the Ernst angle
For a given TR and tissue T1, the Ernst angle maximises signal; raise FA for more T1 weighting, lower it for less.
Trades: SAR; contrast between tissues shifts.
5Add a dedicated contrast mechanism
DWI/ADC for cellularity, SWI/T2* for blood/mineral, dynamic post-contrast for perfusion/enhancement kinetics.
Trades: time; DWI adds distortion/SNR cost; gadolinium considerations.
Parameter trade‑offs
If I change this parameter, what happens?
Each row is an increase in the parameter. Read across for the effect on SNR, resolution, scan time and artifacts.
| Increase ↓ | SNR | Scan time | Resolution | Notes / artifacts |
|---|---|---|---|---|
| TR | ↑ | ↑ | — | More T1 recovery (less T1 weighting), more slices per acquisition |
| TE | ↓ | — | — | More T2 weighting; longer TE = more susceptibility/T2* signal loss |
| Flip angle (GRE) | ↑* | — | — | Peaks at Ernst angle; more T1 weighting & SAR beyond it |
| Slice thickness | ↑ | — | ↓ | More partial-volume averaging; fewer slices for coverage |
| Matrix (phase) | ↓ | ↑ | ↑ | Smaller voxels; time rises with phase-encode steps |
| FOV | ↑ | — | ↓ | Bigger voxels at fixed matrix; too-small FOV = wrap/aliasing |
| Averages | ↑ | ↑ | — | SNR ∝ √NEX; also averages out some motion |
| Receiver bandwidth | ↓ | — | — | Higher BW → less chemical-shift & shorter TE, but less SNR |
| Parallel imaging factor R | ↓ | ↓ | — | Time ÷ R; SNR ÷ √R × g-factor; residual unfolding |
| Echo-train length | — | ↓ | ↓† | Faster, but T2 blurring, contrast shift & higher SAR |
*Flip-angle SNR maximises at the Ernst angle (cos α = e−TR/T1). †Effective resolution loss from T2 decay across the echo train, not matrix.
The core trade-off triangle
You can usually optimise two of these three at once; the third pays for it.
| Want more… | Cheapest lever | What you give up |
|---|---|---|
| SNR | Thicker slice, more averages, lower BW, better coil | Resolution, time, or chemical-shift |
| Resolution | Larger matrix, thinner slice, smaller FOV | SNR and time |
| Speed | Parallel imaging, longer ETL, partial Fourier, SMS/DL-recon | SNR, and sometimes contrast/blurring |
Anchor: SNR ∝ voxel volume × √(total sampling time). Almost every parameter above is one term of that relationship in disguise.1
Protocol tuner
Tune a protocol toward a goal
Pick the situation closest to yours; each gives a concrete starting recipe and the first parameter to move.
Patient can’t hold still: prioritise speed & robustness
Recipe
Single-shot or short-ETL TSE; add parallel imaging (R 2–3); consider radial (radialBLADEPROPELLERMultiVaneJET) for T2/FLAIR; SMS for DWI; drop averages to 1.
First lever if still too long: raise parallel-imaging factor before sacrificing resolution.
Small structure / need high spatial resolution
Recipe
Small FOV + high matrix; thin slices or 3D isotropic (3D TSESPACECUBEVISTA3D MVOX); dedicated coil; recover SNR with more averages or lower BW.
First lever if too noisy: lower receiver bandwidth, then add an average, before thickening the slice.
Deep / small-signal anatomy: prioritise SNR
Recipe
Thicker slices, more averages, lower bandwidth, closest-fitting coil, minimise parallel-imaging factor (R 1–2). Consider higher field where available.
First lever if too slow: accept modest parallel imaging (R 2); the √R SNR cost is usually worth the time saved.
Metal hardware in the field: prioritise artifact reduction
Recipe
TSE/FSE only (no GRE); high bandwidth; STIR (not spectral fat-sat); thin slices; frequency-encode along the implant’s long axis; escalate to a MAR sequence (VAT+SEMAC/MAVRICWARPMAVRIC SLO-MARmetal package) if distortion persists.
First lever if distortion remains: increase bandwidth further, then add SEMAC/VAT encoding steps.
Lesion detection: prioritise contrast / conspicuity
Recipe
Set TR/TE for the target weighting; add fat suppression (spectral if B0 homogeneous, STIR/Dixon if not); consider an IR prep (FLAIR/STIR) to null background; include DWI where cellularity matters; post-contrast T1 with fat-sat for enhancement.
First lever if conspicuity is poor: add or switch fat-suppression method before touching TR/TE.
Use with the other tools: pair a tuned protocol here with the Protocols & Scan-Parameters Explorer for full pathology-specific recipes, and the Sequence Physics Explainer for the underlying mechanisms.
Advanced optimisation techniques
Where scan time comes from, and the levers that buy speed, SNR or contrast without wrecking the others. Start from the time equation; it explains why some FOV changes cost nothing and others cost minutes.
Scan time = TR × Nphase × Naverages ÷ (ETL × acceleration factor)
ETL = 1 for sequences without an echo train.
In 2D, TR, phase-direction parameters, averages, echo-train length and acceleration move this number; 3D also multiplies it by the number of partitions, and slices that do not fit in one TR add concatenations or extra acquisitions. Frequency-direction changes (readout FOV, frequency matrix, frequency oversampling) do not change scan time as long as TR does not change, because they’re sampled within the same readout. A longer readout (larger frequency matrix or lower bandwidth) can, however, force a longer minimum TE, TR or echo spacing, or fewer slices per TR.
FOV & phase FOV, and why FOV “doesn’t cost time”
| Change | Effect on time | Why |
|---|---|---|
| Readout (frequency) FOV ↓, matrix fixed | No change | Readout duration & number of samples are unchanged; you’re just narrowing displayed frequency field. (Resolution changes, time doesn’t.) |
| Frequency oversampling | No change (free) | Extra readout samples are collected within the same echo. It removes readout-direction wrap at no time cost. Usually on by default. |
| Phase FOV ↓ (rectangular FOV) | Faster ↓ | Fewer phase-encode steps for the same resolution → fewer TR repetitions. Put the shorter anatomical dimension in phase to exploit this. Risk: wrap if anatomy exceeds phase FOV. |
| Phase oversampling (no-phase-wrap / foldover) | Slower ↑ | Adds phase-encode lines beyond the displayed FOV to stop aliasing; time and SNR effects vary by vendor: it usually adds time, but some systems (for example GE No Phase Wrap) reduce averages automatically to hold scan time. Use the minimum % that clears the wrap. |
| Phase resolution ↑ (finer phase matrix) | Slower ↑ | More phase steps = more TRs. Trade against averages / acceleration. |
Rule: to speed a scan, look first at TR, phase steps (matrix, phase FOV, phase resolution), averages and parallel-imaging factor, not the readout side.
Driven equilibrium (DRB / RESTORE / DRIVE / FRFSE) with fast T2 / HASTE
A driven-equilibrium / restore pulse (Siemens RESTORE / DRB, Philips DRIVE, GE FRFSE / fast recovery) applies a −90° pulse at the end of the echo train to flip residual transverse magnetisation back to the +z (longitudinal) axis, rather than letting it recover naturally.
- Longitudinal magnetisation recovers faster → you can use a shorter TR for the same T2/fluid contrast → shorter scan.
- Boosts signal from long-T2 fluid (CSF, bile, synovial fluid) → brighter, higher-contrast heavily-T2 images.
- Classic uses: MRCP, cisternography/CISS-type fluid imaging, T2 cartilage & spine, and heavily T2-weighted TSE where you want fluid bright without a punishingly long TR.
- Cost: extra RF → higher SAR and slightly longer echo train; can reduce slices per TR.
Note: single-shot HASTE/SSFSE already behaves as if TR is very long per slice, so restore benefits multi-shot TSE most; it’s still paired with fast T2 fluid imaging to maximise fluid contrast at reduced TR.
Flip angle: nodes, modes & types
| Concept | What it is | When you change it & why |
|---|---|---|
| Excitation flip | Angle of the initial RF tip. | GRE: low flip (near Ernst angle) maximises SNR for a given TR & T1; high flip adds T1 weighting. bSSFP: higher flip (~35–70°) boosts T2/T1 contrast up to an optimum then SAR-limited. |
| Refocusing flip (TSE) | Angle of the 180° (or lower) refocusing pulses in the echo train. | Dropping refocusing below 180° cuts SAR and reduces blurring; enables long echo trains, but lowers signal. Basis of variable-flip trains. |
| Variable / hyperecho & VRFA | Refocusing angle varies along the echo train (very high early, tapering). | Used in 3D long-ETL TSE (SPACE / CUBE / VISTA) to hold signal, slash SAR and limit blurring over hundreds of echoes, which makes isotropic 3D TSE feasible. |
| Ernst angle | θ where cos θ = e−TR/T1, the flip giving maximum GRE signal. | Pick it when SNR-limited in fast spoiled GRE; shorter TR → smaller optimal flip. |
| Low flip (SAR / speed) | Reducing all flips. | Cuts SAR (helps at 3 T & with implants), enables more slices per TR; costs signal/contrast. |
“Flip angle mode” on some consoles selects constant vs variable (hyperecho/VRFA) refocusing trains. Choose variable for 3D/long-ETL to control SAR & blurring, constant for maximal signal on short trains.
Putting it together: quick decision cues
- Need speed? Cut phase FOV to the anatomy, raise parallel-imaging factor, drop averages, then consider TR. Check SNR & wrap after each.
- Fluid too dark / TR too long? Add driven equilibrium (RESTORE/DRIVE/FRFSE) and shorten TR.
- Wrap in phase? Add phase oversampling (costs time) or a sat band, or swap PE to the fitting axis (free).
- SAR limit hit (3 T, implants, long ETL)? Lower/vary refocusing flip, reduce excitation flip, fewer sat pulses.
Vendor terminology differs (RESTORE/DRB · DRIVE · FRFSE; SPACE · CUBE · VISTA). Educational content. Always work within your scanner’s validated protocols and SAR limits.
When to review images mid-protocol and the ideal sequence order
The goal: catch pathology and quality problems early enough to add or repeat sequences, without ever leaving the scanner idle. Reviewing and scanning should overlap: the scanner keeps acquiring while you look at what’s already reconstructed.
Show the full section
The core principle: review in the shadow of the next acquisition
Every sequence reconstructs and lands on the console the moment it finishes. That is your review window, while the next sequence is already running. Never stop the table to “have a look.” Structure the protocol so the decision-relevant series finish early, giving you the whole rest of the scan to review them and still add sequences before the patient is off the table.
- Front-load the diagnostic anchor. Put the sequence that most often drives a protocol change (e.g. post-contrast timing, DWI for stroke/abscess, a key T2) early enough that its result can change what you do next.
- Review during the longest sequence. The high-resolution 3D or long TSE at the end is your best uninterrupted review window; it runs for minutes while you check everything before it.
- Decide add-ons before the last long sequence starts, so a supplementary series slots in without dead time.
Ideal ordering logic (generic, adapt per vendor/protocol)
- Localiser / calibration: review coil position & coverage immediately; the fastest possible feedback loop.
- A fast survey sequence of the region (e.g. a quick T2/HASTE or 2D localiser-plus): confirms anatomy is covered and gross pathology, so you can adjust FOV/slab before committing minutes.
- Motion- and timing-sensitive sequences early: DWI, breath-holds, first-pass or dynamic contrast, anything that fails if the patient tires. Do these while the patient is freshest.
- Contrast timing anchor: if the study is contrast-dependent, structure pre-contrast series so they finish before injection, then run post-contrast without a gap. Review the pre-contrast set during the post-contrast run.
- Long high-resolution / 3D isotropic last: this is both the highest-yield anatomical series and your review window: while it runs (often 4–8 min) you review the entire prior stack and confirm nothing needs repeating.
- Add-ons / repeats: slot decided during step 5; if nothing needed, the scan ends on the 3D and the patient is released with no idle time.
Review checkpoints vs what’s running
| Checkpoint | What you review | What is running meanwhile | Decision it enables |
|---|---|---|---|
| After localiser | Coverage, coil, positioning | First survey sequence | Re-landmark / re-plan slab before committing time |
| After survey/first T2 | Gross pathology, artefact, motion | DWI / next weighting | Add fat-sat, swap plane, tighten FOV |
| After DWI / dynamic | Restriction, enhancement pattern | Pre-contrast structural series | Decide extra sequence or extra station |
| During post-contrast | The whole pre-contrast stack | Post-contrast series | Add delayed/extra post-Gd plane before table-off |
| During final 3D | Entire study for completeness | 3D isotropic (longest run) | Repeat any degraded series or release patient |
See also the protocol tuner and parameter trade-off modes above for how to shorten individual sequences without losing the diagnostic content you’re reviewing for.
You’ve just optimised a protocol for free. Pro is for doing it at scale.
The optimiser, the advanced techniques and the mid-protocol review guide above stay free. When Pro opens, it adds:
- Save and share optimised protocols across your department PRO
- Vendor-specific parameter cards (Siemens, GE, Philips, Canon) as printable console sheets PRO
- The protocol library with scan-time and SAR trade-off notes PRO
Sources
1. Elmaoglu M, Celik A. MRI Handbook: MR Physics, Patient Positioning, and Protocols. Springer, 2012.
2. Westbrook C, Talbot J. MRI in Practice. 5th ed. Wiley-Blackwell, 2018.
3. Bushberg JT et al. The Essential Physics of Medical Imaging. 4th ed. Wolters Kluwer, 2020.
4. Hargreaves BA et al. Metal-induced artifacts in MRI. AJR 2011;197(3):547–555. doi:10.2214/AJR.11.7364
5. Deshmane A et al. Parallel MR imaging. JMRI 2012;36(1):55–72. doi:10.1002/jmri.23639
6. Pipe JG. Motion correction with PROPELLER MRI. MRM 1999;42(5):963–969.
7. Barth M et al. Simultaneous multislice (SMS) imaging techniques. MRM 2016;75(1):63–81. doi:10.1002/mrm.25897
8. Questions and Answers in MRI (Elster AD), mriquestions.com.
Educational reference for imaging professionals. Optimisation choices must be validated on your own scanner and adapted to clinical indication, field strength, and local protocol. Vendor feature names are trademarks of their respective manufacturers and listed for cross-reference only. © High Yield Imaging.