Phase Oversampling — Fold-over Suppression / No Phase Wrap

MRIninja Knowledge Base | MRI Parameter Deep Dive Version 1.0 — May 2026

1. Introduction and General Purpose

Phase oversampling — known as fold-over suppression (Philips), No Phase Wrap (Siemens, GE), or anti-aliasing — is the acquisition parameter that prevents wraparound artefact (aliasing, fold-over) from appearing in the phase-encoding direction of an MRI image. It achieves this by extending the k-space sampling beyond the prescribed phase FOV, acquiring additional k-space lines that prevent anatomy outside the intended field of view from appearing inside the image.

Aliasing in MRI is not a failure — it is the mathematically inevitable consequence of the Nyquist theorem applied to discrete k-space sampling. When the MRI scanner samples k-space at a spatial frequency of 1/FOV_p (one line per FOV_p interval), any anatomy that extends further than FOV_p/2 from the centre of the FOV in the phase direction produces signals that are indistinguishable from anatomy at the opposite side of the FOV. The scanner "wraps" this anatomy back into the image. Phase oversampling is the universal solution: sample more k-space lines beyond the FOV boundary, extending the Nyquist coverage to encompass the full anatomical extent, then discard the extra k-space data before display.

This parameter is ubiquitous — it is present in every clinical MRI protocol and is one of the most operationally impactful decisions a technologist makes for image quality. Incorrect phase oversampling is the most common source of artefacts that simulate or obscure pathology in routine clinical MRI. Too little: anatomy wraps across and overlies the target organ. Too much: unnecessary acquisition time is wasted. The balance requires understanding both the anatomy and the physical mechanism.

Historical context: fold-over artefacts appeared in the very first clinical MRI images (early 1980s) whenever the phase FOV was smaller than the patient's dimensions. Early solutions included: saturating the anatomy outside the FOV with presaturation bands; extending the FOV to cover all anatomy (simple but resolution-reducing); or accepting the artefact and repositioning phase encoding. Dedicated phase oversampling as an independent scanner parameter was standardised across all vendors by the mid-1990s.

2. Physical Foundations

2.1 The Nyquist Theorem and Aliasing in MRI

In MRI, the phase-encoding gradient encodes spatial position in the phase direction by imposing a linearly increasing phase across the object. The discrete sampling of k-space means that the sampling frequency in phase is 1/FOV_p — one k-space line per FOV_p interval. The Nyquist theorem requires that the sampling rate be at least twice the highest spatial frequency in the object:

f_sampling ≥ 2 × f_max

For a finite object, the "highest spatial frequency" is determined by the object's spatial extent in the phase direction. If the object extends further than FOV_p/2 on either side of the isocentre, signals from outside the FOV have spatial frequencies that exceed the Nyquist limit of the sampling → aliasing.

The aliasing mechanism in phase encoding:

A proton at position y (outside the FOV) accumulates a phase:

φ(y) = γ × G_p × Δt × N_p × y

where G_p = phase gradient amplitude, Δt = gradient duration, N_p = phase-encoding step number. For a proton at position y + FOV_p (one FOV-width outside the FOV), the accumulated phase is:

φ(y + FOV_p) = φ(y) + 2π × N_p

Since the Fourier transform is periodic with period 2π, this phase is indistinguishable from φ(y). The scanner assigns this signal to position y — the position of the correctly imaged anatomy — producing a ghosted superimposition at the wrong location.

Clinical translation: anatomy at position y + FOV_p (outside the FOV) appears superimposed on anatomy at position y (inside the FOV). For a small FOV brain axial acquisition where the FOV extends from -100 mm to +100 mm (FOV_p = 200 mm), any patient with head dimensions > 200 mm will have anatomy beyond the FOV wrap to the opposite side.

2.2 Mathematical Foundations

Phase oversampling mechanism

Phase oversampling extends the k-space sampling from N_y lines (the required matrix) to N_y_total = N_y × (1 + OS/100) lines, where OS is the oversampling percentage:

N_y_total = N_y × (1 + OS/100)

This increases the effective Nyquist coverage:

Extended FOV_p = FOV_p × (1 + OS/100)

Anatomy within the extended FOV_p is correctly encoded without aliasing. After k-space acquisition, the scanner discards the additional lines (or applies a view filter) and displays only the originally prescribed FOV_p — producing an image without fold-over artefact and without any visible change in the displayed FOV.

Physical interpretation: phase oversampling samples more k-space lines to extend the Nyquist coverage without changing the image resolution (the in-plane voxel size Δy = FOV_p / N_y is unchanged), the displayed FOV, or the tissue contrast.

Time cost

Acquisition time scales with the total number of phase-encoding steps:

T_acq = TR × N_y_total × NSA / ETL = TR × N_y × (1 + OS/100) × NSA / ETL

The time penalty of phase oversampling is:

ΔT / T_base = OS / 100

For OS = 50% → 50% time increase. For OS = 25% and ETL = 16: ΔT = 25% × (T_base / 16) ≈ 1.6% of base time → negligible.

Key insight: the time penalty of phase oversampling is divided by ETL. For sequences with ETL = 16, even 100% phase oversampling adds only 100/16 = 6.25% to the acquisition time — negligible in most clinical contexts. This is why phase oversampling is almost always the correct choice for eliminating fold-over, as its time cost is very low for TSE protocols.

Phase oversampling in the frequency direction

In the frequency-encoding (readout) direction, aliasing is inherently prevented differently: the receiver bandwidth extends beyond the prescribed FOV in the readout direction by default (frequency oversampling, typically 100–200% extra in the readout direction). This adds no time cost (all frequency points are collected simultaneously during the readout) and is applied automatically by the scanner. The technologist does not need to set frequency oversampling separately — it is always active.

Phase oversampling and SNR

Phase oversampling does NOT improve SNR relative to the unoversampled acquisition. The additional k-space lines contain anatomy from outside the prescribed FOV — anatomy that does not contribute to the SNR of the displayed image. However, phase oversampling does not decrease SNR either: the displayed image is reconstructed from N_y_total lines but the SNR per voxel reflects N_y_total samples → SNR is marginally improved relative to an N_y-only reconstruction, since more data is used.

Practical rule: phase oversampling is SNR-neutral (slight benefit, negligible in practice) and nearly time-neutral (for high ETL) — making it a pure quality improvement with minimal cost.

3. Units, Terminology and Vendor Nomenclature

Phase oversampling is expressed as a percentage (%) of the prescribed phase FOV or phase matrix.

Siemens: "Phase oversampling" is set as a percentage (0–200%). This adds N_y × OS/100 extra phase-encoding steps distributed symmetrically around the phase direction centre. The display is cropped to the prescribed FOV_p.

Philips: "Fold-over suppression" is a Boolean (ON/OFF) with a slider for the percentage when ON. The Philips system explicitly shows the extended FOV that will be acquired vs the displayed FOV.

GE: "No Phase Wrap" is a Boolean toggle. When enabled, GE applies full phase oversampling (100%). There is no partial oversampling option in standard GE protocols — it is either off or fully on. This is the most restrictive of the vendor implementations because it doubles the phase-encoding steps when enabled.

Canon: Phase oversampling expressed as a percentage, similar to Siemens.

United Imaging: Standard percentage-based phase oversampling. The UIH scan planner highlights when anatomy extends beyond the prescribed FOV_p to alert the technologist to enable phase oversampling.

4. Typical Value Ranges

4.1 Phase Oversampling by Application and Anatomy

4.2 Phase Oversampling vs ETL — Time Impact

Key takeaway: at ETL ≥ 16 (standard for most TSE sequences), even 100% phase oversampling adds only 6% to acquisition time. The common clinical hesitation to apply phase oversampling ("it makes the scan longer") is almost always unfounded for TSE protocols.

5. Parameter Interaction Ecosystem

5.1 Parameter Relationships Matrix

5.2 The rFOV-OS Coupling

The most important clinical interaction of phase oversampling is with rectangular FOV (rFOV). When the phase FOV is reduced (rFOV applied to save time), the likelihood of anatomy outside the rFOV increases. The rule:

OS_required = max(0, (Anatomy_extent_in_phase / FOV_p − 1) × 100)

Example: Lumbar spine sagittal; FOV_p = 220 mm (A-P direction); patient's A-P depth = 280 mm.

• Without OS: anatomy at 280/2 = 140 mm from centre, but FOV_p/2 = 110 mm → 30 mm outside FOV → fold-over.
• Required OS = (280/220 − 1) × 100 = 27% → set OS ≥ 30% to ensure coverage.

This calculation should be performed for every rFOV application. In practice, 50% OS is a safe default for A-P phase direction in the spine and body.

6. Effects on Image Appearance

6.1 With Adequate Phase Oversampling (OS ≥ anatomy extent)

• No fold-over: anatomy is cleanly confined to the prescribed FOV; no spurious signals appear from outside
• Image appearance unchanged relative to a larger-FOV acquisition: the displayed image looks exactly as intended — no evidence of oversampling in the final image
• Lesion conspicuity preserved: lesions near the image margin are not obscured by fold-over from opposite anatomy
• No change in resolution, contrast, or SNR perceptible to the reader

6.2 With Insufficient Phase Oversampling

Fold-over artefact appears: anatomy from outside the FOV_p is superimposed on anatomy inside the FOV_p at the mirror-symmetric position across the FOV centre.

Specific appearances:

• Brain axial, small FOV, R-L phase: opposite side of skull wraps over the contralateral hemisphere → simulates extra tissue adjacent to the brain
• Spine sagittal, A-P phase: anterior abdominal wall and subcutaneous fat wraps posteriorly onto the spine → fat signal overlies the posterior elements and spinal canal → masks disc herniation; simulates epidural fat
• Neck axial, A-P phase: posterior cervical fat wraps anteriorly → soft tissue signal in the anterior triangle → simulates lymph node or mass
• Breast axial, R-L phase: the opposite breast or arm tissue wraps into the image → signal superimposition across the field
• WB-MRI coronal, R-L phase: patient's arms wrap into the bilateral nodal chains on both sides → obscures lymph nodes

6.3 Partial Oversampling

When OS is set between 0% and the required minimum, aliasing is partially reduced. The fold-over anatomy appears at the edge of the image (not the centre), potentially overlapping less critical regions. This is occasionally acceptable (artefact at the image periphery, far from the target anatomy) but should be explicitly verified for each protocol.

7. Effects on Acquisition Time

7.1 Direct Effect

ΔT = T_base × (OS/100) / ETL

This is the only direct effect. The extra k-space lines (N_y × OS/100 lines) must be acquired, each requiring one TR period (divided by ETL).

7.2 Practical Significance

For standard clinical TSE (ETL = 12–20):

• OS = 50%: ΔT = 50/12 to 50/20 = 2.5–4.2% extra time
• OS = 100%: ΔT = 100/12 to 100/20 = 5–8.3% extra time

For a 4-minute T2 brain with ETL=16 and OS=50%: ΔT = 4 × 0.031 = 7.5 seconds → negligible.

For SE (ETL=1):

• OS = 50%: ΔT = 50% extra time — significant
• OS = 100%: ΔT = 100% extra time — doubles scan time

This is why phase oversampling is rarely used in pure SE sequences (extremely long scans). In practice, SE is replaced by TSE for almost all clinical applications, making the time cost of OS clinically irrelevant.

7.3 Special Case: GE "No Phase Wrap"

GE applies 100% phase oversampling when "No Phase Wrap" is enabled — no partial option. For GE sequences with ETL=1 (rare in clinical practice), this doubles scan time. For standard GE TSE applications (ETL=8–20), the time impact is 5–12.5%.

8. Effects on SNR and CNR

8.1 SNR Effect

Phase oversampling is effectively SNR-neutral. The additional k-space lines do contain data (signals from anatomy outside the prescribed FOV) that contribute to the reconstruction, providing a marginal SNR benefit. Theoretical SNR_OS / SNR_no-OS = √(1 + OS/100) — for OS=100%: √2 = 1.41× theoretical SNR improvement. However, this improvement is practically unnoticeable because: (a) the anatomy contributing to the extra k-space is not of diagnostic interest; (b) other noise sources dominate; (c) the display is cropped to the original FOV_p.

Practical rule: treat phase oversampling as SNR-neutral.

8.2 CNR Effect — Diagnostic Importance

Phase oversampling has a critically important indirect CNR effect. Fold-over artefact can:

• Simulate pathology: a fat-containing fold-over on a fat-suppressed sequence may simulate a T1-bright lesion; a fold-over of a cyst may simulate an enhancing mass; fold-over of the anterior chest wall onto the spine may simulate an epidural mass
• Obscure pathology: fold-over artefact superimposed on the target anatomy reduces the effective CNR between the lesion and its background

In both cases, adequate phase oversampling restores the true diagnostic CNR by removing the confounding fold-over signal.

9. Artefacts Associated with Phase Oversampling

10. Behaviour Across Sequence Families

Spin Echo (SE)

OS adds one full TR per extra k-space line → time cost proportional to OS%. For SE at TR=2000 ms and N_y=256: each 1% OS adds 2000/100 × 1 = 20 ms per line × (256/100) lines per 1% = 5120 ms per 1% OS. OS=50% adds 256 s = over 4 min — impractical. Standard SE uses minimal OS (< 10%) or selects phase direction to minimise the need.

Turbo Spin Echo (TSE/FSE)

OS is almost free (divided by ETL). The standard approach: always apply OS ≥ 25–50% for body and spine TSE; it costs < 3% scan time and eliminates the most common clinical artefact source.

Gradient Echo (GRE/FLASH)

At short TR (5–10 ms), OS time cost = N_y × OS/100 × TR. For TR=5 ms and OS=50% at N_y=256: 256 × 0.5 × 5 = 640 ms → negligible for long scans; relevant for breath-hold sequences where total acquisition time must be < 20 s. For DCE arterial phase (TR=4 ms, N_y=200, ETL=1-equiv, breath-hold 20 s): OS=50% adds 200×0.5×4=400ms — acceptable.

Inversion Recovery (STIR, FLAIR, MPRAGE)

Same principles as TSE (all IR sequences in clinical practice use TSE readout). OS is ETL-divided and affordable. For MPRAGE (3D GRE): OS in the 3D z-direction prevents z-aliasing at slab edges; standard is 15–25%.

EPI (DWI, DSC, fMRI)

EPI acquires all N_y lines in a single shot — there is no meaningful per-line TR cost. Phase oversampling in EPI changes the single-shot echo train length. Adding OS to EPI phase direction extends the echo train → longer readout → more geometric distortion and T2* signal decay. For standard brain EPI, OS is typically 0% (frequency oversampling handles the frequency direction). For body EPI, the aliasing risk in the phase direction is managed by reduced-FOV techniques (ZOOMit/iZOOM) rather than conventional OS. EPI is the sequence family where phase oversampling is most carefully limited.

Dixon

Underlying sequence family determines OS behaviour. Dixon GRE or TSE: same as the respective family. The Dixon fat-water separation process is unaffected by phase oversampling.

DCE (Dynamic Contrast Enhanced)

For 3D GRE DCE (breast, liver): OS is applied in the 3D z-direction (15–25% slice oversampling) to prevent z-aliasing. In the in-plane phase direction, standard OS (25–50%) is applied. The temporal resolution constraint limits how much OS can be applied — every extra k-space line costs time that must fit within the dynamic phase window.

ASL

3D readout ASL: OS in the z-direction standard (15–25%). In-plane phase OS same as for any 3D readout. The very low ASL SNR makes any unnecessary scan time extension potentially harmful (fewer averages available in the same scan time) — OS should be set to the minimum required.

bSSFP

Standard OS applies. For cardiac cine bSSFP: the phase direction is typically A-P (axial images); OS prevents anterior chest wall fold-over. For 3D CISS: z-oversampling prevents slab-end wrap.

Spectroscopy

In MRSI, phase oversampling prevents spatial aliasing in the spectral map. For SVS (single-voxel), spatial oversampling is irrelevant (the voxel is a single point).

11. Field Strength Behaviour

Phase oversampling is a k-space sampling strategy and is field-strength independent in its mechanism. The time cost, the aliasing physics, and the remedy are identical at any field strength. However, some field-strength-related aspects are worth noting:

At 0.55T, SNR is low. Any unnecessary scan time (from excessive OS) represents a lost opportunity to use that time for more NSA. OS should be set to the minimum required — not defaulted to 100%.

At 3T in 3D protocols, the additional SAR from extra phase-encoding steps (OS in z for 3D acquisitions) contributes to total SAR load. When 3T protocols are SAR-limited, reducing 3D z-oversampling from 25% to 15% may allow a higher flip angle or shorter TR elsewhere.

12. Vendor-Specific Implementation

Siemens

Phase oversampling is set as a percentage (0–200%) in the contrast/sequence tab. The system displays the total number of phase-encoding steps (N_y_total = N_y × (1 + OS/100)) and the resulting scan time. A Siemens-specific feature: the "iPAT" (integrated parallel imaging techniques) with GRAPPA interacts with phase oversampling — the auto-calibration signal (ACS) lines at the centre of k-space are always acquired regardless of the phase direction oversampling setting, ensuring that parallel imaging calibration is not compromised.

Frequency oversampling (Siemens "frequency oversampling"): separately settable (default: automatically applied in the readout direction at 100% → prevents frequency-direction aliasing at zero time cost). Technologist does not need to set this.

GE

"No Phase Wrap" is Boolean (ON/OFF). When ON: 100% oversampling is applied (no partial option). This is the most conservative approach — maximum artefact suppression, maximum time cost per line. For sequences with low ETL, this can meaningfully extend scan time. For TSE, the impact is small. GE technologists must understand that "No Phase Wrap" doubles the number of phase-encoding lines collected (but acquisition time increase is ETL-divided).

Philips

"Fold-over suppression" control: percentage slider (0–100%) shown together with the extended FOV graphically on the planning screen. The Philips interface is the most visually intuitive: the technologist can see exactly how much anatomy falls outside the primary FOV and adjust the fold-over suppression percentage to cover it exactly. The graphic representation of the extended vs prescribed FOV makes optimal OS setting straightforward.

Canon

Percentage-based, similar to Siemens. The Canon system provides automated FOV coverage suggestions based on patient body width measurement from localiser images — reducing the need for manual OS estimation.

United Imaging

Percentage-based. The UIH planning system shows anatomy width in the phase direction and highlights when anatomy extends beyond the prescribed FOV_p, prompting the technologist to adjust OS.

Universal hidden behaviour: on all platforms, when phase oversampling extends the k-space acquisition, the final reconstruction crops the image to the originally prescribed FOV_p. The extended data is used during reconstruction (improving Nyquist coverage) but is not displayed. This is transparent to the radiologist — the reported image dimensions reflect the prescribed FOV, not the acquisition FOV.

13. Practical Optimisation Strategies

13.1 Clinical Optimisation Recipes

13.2 Phase Direction Selection and OS Interaction

The choice of phase encoding direction and the OS setting are inseparable decisions. The phase direction should be chosen to:

1. Place the longer anatomical dimension in the frequency direction (no aliasing risk)
2. Place any swallowing/motion artefacts away from the target anatomy
3. Minimise the anatomy extent in the phase direction (reducing OS needed)

When the phase direction cannot be chosen to fully avoid fold-over risk, OS compensates. The two strategies together define the artefact management approach for every protocol.

14. Parameter Extremes

14.1 Zero Phase Oversampling

With OS = 0%, any anatomy extending beyond FOV_p/2 from the FOV centre in the phase direction produces fold-over. This is appropriate only when:

• The FOV_p is confirmed to be larger than the patient's anatomical extent in the phase direction (e.g., large-FOV whole-body coronal T1 where FOV > patient width)
• The fold-over anatomy is confirmed to fall outside the target diagnostic region
• Scan time is critical and any extra lines are unacceptable (e.g., arterial phase DCE at the tightest breath-hold timing)

Routine OS=0% without these confirmations is a common protocol error.

14.2 100–200% Phase Oversampling

Full or double oversampling (100–200%) is rarely required but is appropriate for:

• WB-MRI coronal (arms alongside body, extending well beyond the FOV): 100% OS ensures the full arm-to-arm width is within the Nyquist coverage
• Small FOV targeted acquisitions where the surrounding anatomy (e.g., orbits: FOV=80 mm, surrounding head tissue extends 150 mm beyond centre in the phase direction) would produce massive fold-over without OS

At 200% OS, the extended FOV is 3× the prescribed FOV_p — the scanner acquires N_y × 3 phase-encoding lines. At ETL=16, the time cost is 200/16 = 12.5%. Acceptable for most clinical applications; may push breath-hold body acquisitions beyond tolerance.

15. Common Optimisation Errors

16. MRI Technologist Pearls

The localiser is the OS planning tool: before setting OS, check the localiser in the phase-encoding direction. Measure (approximately) how far the patient's anatomy extends beyond the prescribed FOV_p boundary. That excess, divided by FOV_p, expressed as a percentage, is the minimum OS needed. For example: patient width in the phase direction = 320 mm; FOV_p = 220 mm → excess per side = (320−220)/2 = 50 mm → OS_min = (50/110) × 100 = 45% → set 50%.

OS is nearly free for TSE — always apply it: at ETL=16, 50% OS costs 3% extra time. The benefit — eliminating the most common artefact in clinical body MRI — is always worth this cost. Default all spine and body protocols to OS ≥ 50%.

The mirror rule for identifying fold-over: any structure that appears as a ghost in the phase direction at a position that is exactly mirrored from a known anatomical structure (across the FOV centre) is a fold-over, not pathology. If you see an unexpected soft tissue density in the anterior spine on a sagittal sequence: look for its mirror counterpart at the corresponding anterior distance from the centre. If present, it is fold-over. Apply OS and rescan.

Phase direction and OS are planned together: the phase direction choice is always made together with the OS decision. When selecting A-P as the phase direction for the neck (to manage swallowing artefacts), simultaneously verify that the A-P patient depth is covered by FOV_p + OS_extended_coverage. If the patient is large (A-P > 280 mm) and the FOV_p is 220 mm, OS must be ≥ 27%.

OS in 3D is z-direction oversampling (different control, same concept): check that the "Slice oversampling" (Siemens) or equivalent is set to ≥ 15% for every 3D protocol. The consequence of forgetting this is z-wrap at the slab edges — a subtle artefact that simulates pathology at the brain vertex or inferior organ pole.

Do not change the phase direction mid-protocol without reassessing OS: if the phase direction is changed from the default (e.g., switching a spine axial from A-P to R-L), immediately reassess whether the existing OS setting is adequate for the new direction. A 50% OS for A-P (adequate for spine depth) may be completely insufficient for R-L (width of the patient plus arms).

17. Real Clinical Examples

Example 1: Lumbar Spine Sagittal — Anterior Abdominal Wall Fold-Over

Clinical scenario: low back pain; lumbar spine MRI sagittal T2 TSE. FOV_p = 220 mm (A-P), rFOV; OS = 0%.

Problem: the patient's A-P depth is 290 mm. The anterior abdominal wall is at 145 mm anterior from the FOV centre; the FOV extends to only 110 mm → 35 mm of anterior fat outside the FOV → folds posteriorly onto the posterior vertebral elements and spinal canal. Bright fat signal from the anterior subcutaneous tissue appears over the L3–L5 posterior elements, partially obscuring the view of the posterior disc margin and simulating epidural fat.

Solution: OS = 50% → extended FOV = 330 mm → covers the full 290 mm A-P depth → fold-over eliminated.

Time cost: TR=3500/ETL=16: ΔT = 3500 × 256 × 0.5 / 16 = 28,000 ms ≈ 28 s extra → total scan time increases from ~3.0 min to ~3.5 min.

Diagnosis: with fold-over removed, the L4–L5 central disc herniation is now clearly visible without the confounding anterior fat signal.

Example 2: WB-MRI Myeloma — Arms Fold-Over on Coronal STIR

Clinical scenario: staging WB-MRI for multiple myeloma; coronal STIR, R-L phase encoding, FOV = 420 mm; OS = 25%.

Problem: patient's arm-to-arm width (arms alongside body) = 580 mm. The FOV of 420 mm covers the central trunk but not the arms (580 > 420). With OS = 25%: extended FOV = 525 mm — still less than 580 mm. Fold-over: the lateral arms appear bilaterally over the mid-thorax nodal chains → the level II and III nodes are partially obscured by fold-over arm signal on the coronal STIR.

Solution: OS = 100% → extended FOV = 840 mm → covers 580 mm arm-to-arm width completely.

Time cost: STIR ETL=20; ΔT = TR × N_y × 1.0 / 20 = TR × N_y / 20 ≈ 25% additional lines ÷ ETL=20 = 5% time increase per station. At 4.5 min per station (×4 stations): total extra time ≈ 54 seconds across the full WB-MRI — clinically acceptable.

Diagnosis: thoracic level II and III nodes now clearly visible; a 12 mm nodal metastasis at left level III was obscured on the OS=25% acquisition.

Example 3: Small FOV Brain — Skull Fold-Over

Clinical scenario: targeted brain MRI for hippocampal assessment; axial T2 TSE, FOV_p = 180 mm (R-L), OS = 0%.

Problem: patient's head width = 164 mm → within the FOV (no aliasing expected). However, the technologist noticed a thin bright band at the left lateral edge of the image — not an expected finding.

Investigation: the fold-over is from the right-side scalp/skull (at 82 mm from centre), which at FOV/2 = 90 mm is within the FOV — no fold-over expected. However, the coil element sensitivity extends further than 90 mm laterally, picking up signal from the right ear (at 95 mm from centre) → this 5 mm excess folds onto the left edge.

Solution: OS = 10% → extended FOV = 198 mm → covers the full 190 mm including both ear regions.

Time cost: TR=5000/ETL=16: ΔT = 5000 × 256 × 0.1 / 16 = 8000 ms ≈ 8 s → negligible.

Lesson: even when the anatomical extent appears within the FOV from the localiser, the coil's extended sensitivity region may exceed the FOV. Always apply at least 10–15% OS for all brain protocols.

Example 4: Breast MRI — Arm Fold-Over in DCE

Clinical scenario: breast DCE for high-risk screening; bilateral axial FOV = 360 mm; R-L phase encoding; OS = 0%.

Problem: the patient's shoulder-to-shoulder width (arms alongside body) = 490 mm. FOV_p = 360 mm → each arm is at approximately ±245 mm from centre, which is > 180 mm (FOV/2) → bilateral arm fold-over.

Effect: both arms wrap into the breast image, producing bilateral linear enhancement artefacts at the medial breast margins that simulate enhancing masses.

Solution: OS = 50% → extended FOV = 540 mm → covers 490 mm arm-to-arm width.

Time cost: 3D VIBE ETL-equivalent ≈ single echo per TR effectively → at TR=4 ms, N_y=288, OS=50%: ΔT = 288 × 0.5 × 4 ms = 576 ms per 3D phase → within the breath-hold tolerance.

Important caveat: for the arterial phase (< 20 s total), even 576 ms extra may push the limit. Check total phase time after setting OS; reduce N_y slightly if needed to maintain temporal resolution. Alternatively, position the patient with arms above the head (reduces arm-to-arm width within the coil to ≈ 380 mm) — eliminating the OS requirement.

18. Visual Educational Material

18.1 The Aliasing Mechanism Diagram

PHASE DIRECTION →

FOV = 200 mm
Centre at position 0

Anatomy positions:
  Head tissue at +80 mm (within FOV ✓)
  Subcutaneous fat at +115 mm (OUTSIDE FOV: 115 > 100)
  
WITHOUT PHASE OVERSAMPLING:
  The fat at +115 mm appears at: +115 - 200 = -85 mm
  → Folded OVER the brain at -85 mm from centre
  
  [ skull | brain | brain | brain | (FOLDED FAT) | skull ]
    -100    -85     0      +85      ← +115 folds here

WITH 30% PHASE OVERSAMPLING:
  Extended FOV = 200 × 1.30 = 260 mm
  Coverage: ±130 mm → covers fat at +115 mm ✓
  No aliasing. Display crops to ±100 mm.
  
  [ skull | brain | brain | brain | skull ]  (clean image)
    -100    -85     0      +85    +100

18.2 OS Setting Decision Tree

IS ANATOMY EXTENT IN PHASE DIRECTION > FOV_p?

Measure: Anatomy_width_phase (from localiser)
Compare to: FOV_p (prescribed)

Anatomy_width > FOV_p?
│
├── YES → OS_minimum = (Anatomy_width / FOV_p − 1) × 100%
│         Set OS ≥ OS_minimum + 10% safety margin
│         
│         ETL ≥ 8? → Time cost is negligible (< 5%) → apply OS
│         ETL = 1 (SE)? → Consider phase direction change first;
│                         OS may double scan time
│
└── NO → OS = 0% theoretically adequate
         BUT: consider 10–15% safety margin for coil sensitivity
         beyond the FOV boundary
         
SPECIAL CASES:
  - WB-MRI with arms alongside: always OS ≥ 100% (R-L phase)
  - 3D slab: always z-oversampling ≥ 15%
  - EPI: use reduced-FOV technique, NOT conventional OS

18.3 Time Penalty vs OS and ETL

       ETL:   1     4     8    16    32    64
OS 25%:      25%   6.3%  3.1%  1.6%  0.8%  0.4%
OS 50%:      50%  12.5%  6.3%  3.1%  1.6%  0.8%
OS 100%:    100%   25%  12.5%  6.3%  3.1%  1.6%

→ At ETL ≥ 8:  100% OS costs ≤ 12.5% time
→ At ETL ≥ 16: 100% OS costs ≤ 6.3% time
→ CONCLUSION: OS is nearly always affordable for TSE protocols

19. Evidence Gaps and Ongoing Debate

Optimal OS for specific body applications: the OS values recommended in vendor default protocols are based on conservative estimates of patient anatomy extent, not prospective studies of aliasing prevalence at different OS levels. The minimum OS required to reliably prevent clinically significant fold-over for specific patient populations (obese, paediatric, different ethnicities with different body dimensions) has not been formally studied.

AI-guided OS recommendation: scanner AI platforms (GE SmartExam, Siemens AIAlign) are increasingly able to detect patient anatomy extent from localiser images and automatically propose OS settings. The clinical accuracy and safety of AI-proposed OS recommendations — particularly for large patients or unusual positioning — has not been prospectively validated.

EPI OS alternatives: for EPI-DWI, the standard recommendation is to use reduced-FOV techniques (ZOOMit/iZOOM/FOCUS/REVEAL) rather than conventional phase oversampling, to avoid extending the EPI echo train. Whether AI-based geometric distortion correction could replace the need for FOV restriction in EPI (enabling conventional OS without distortion penalty) is an area of active investigation.

Phase oversampling in parallel imaging: GRAPPA and SENSE both require that the aliasing pattern be predictable from the coil geometry. Large phase oversampling changes the aliasing pattern in undersampled parallel imaging acquisitions in ways that may affect reconstruction accuracy. The interaction between OS percentage and parallel imaging g-factor for specific coil configurations has not been comprehensively characterised.

20. Miscellaneous and Future Directions

Historical context: fold-over artefact was recognised as the primary image quality problem in early clinical MRI (1981–1985). The first solution was simply to increase the FOV to encompass all anatomy — practical but resolution-reducing. Phase oversampling as an independent parameter (separate from FOV change) was implemented on commercial scanners starting around 1986–1990. The trade-off between OS time cost and artefact suppression was first explicitly described in Listerud et al. (1992) in the context of fast SE imaging.

Partial Fourier interaction with phase oversampling: partial Fourier (acquiring only 60–75% of k-space in the phase direction) interacts with phase oversampling in non-trivial ways. Combining partial Fourier with OS means acquiring 60% of an extended k-space → the effective coverage is 60% × (1+OS/100) × N_y lines. The reconstruction algorithm must account for both the partial Fourier asymmetry and the oversampling extension simultaneously. This is handled automatically by the scanner but the technologist should verify that partial Fourier + OS combinations produce acceptable image quality for the specific application.

Compressed sensing and aliasing: CS acquisitions deliberately introduce aliasing (incoherent undersampling) as part of the reconstruction strategy. The concept of phase oversampling (which prevents aliasing) is incompatible with CS-style deliberate aliasing. For CS protocols, aliasing suppression is achieved by the iterative CS reconstruction algorithm, not by OS. Technologists using CS protocols should understand that "phase oversampling" in the conventional sense does not apply — the CS algorithm handles the Nyquist problem differently.

21. Evidence-Based References

A. Guidelines / Consensus / Society Recommendations

(Phase oversampling is a universal technical standard not addressed by specific clinical guidelines. The Nyquist theorem and its application to MRI sampling are foundational physics referenced in all standard MRI textbooks.)

B. Systematic Reviews / Meta-analyses

(No dedicated systematic reviews address phase oversampling optimisation as a primary subject.)

C. Important Prospective / Original Studies

(Phase oversampling is a fundamental acquisition technique implemented uniformly across vendors; clinical validation studies address specific applications rather than the parameter itself.)

D. Technical MRI Papers

E. Landmark Historical References

(References [1] and [2] above serve as the landmark historical references for this parameter.)

End of document — Phase Oversampling / Fold-over Suppression / No Phase Wrap — MRIninja v1.0 — May 2026 Parent page: MRI Parameters — Overview and Classification (9501) Related child pages: FOV — Field of View · Acquisition Matrix · 2D vs 3D Acquisition · Partial Fourier · Parallel Imaging (R) · EPI and Geometric Distortion

Related parameter deep dive: Parallel Imaging explains acceleration factor, g-factor, SENSE/GRAPPA reconstruction and SMS/Multiband trade-offs relevant to this parameter.