Susceptibility Weighted Imaging (SWI) Sequence

Susceptibility Weighted Imaging (SWI) — Physics, Parameters, and Clinical Applications

MRIninja Knowledge Base | Sequence Child Page Parent page: 9003-mri-sequences-overview-classification Version 1.0 — May 2026

1. Introduction: Historical Evolution and Clinical Purpose

Susceptibility Weighted Imaging is a three-dimensional, fully velocity-compensated gradient echo sequence that uses both the magnitude and the phase of the MR signal to generate an image with exquisite sensitivity to magnetic susceptibility differences between tissues. It was developed to solve a specific diagnostic gap: the detection of iron-containing structures, deoxyhaemoglobin-rich venous blood, calcification, and haemosiderin deposits at spatial scales and susceptibility sensitivities beyond the reach of conventional T2*-weighted GRE sequences.

The sequence was introduced by E. Mark Haacke and colleagues at Wayne State University [7], initially described as high-resolution blood oxygen level-dependent (BOLD) venography in 1997, and formally published in its modern form as "susceptibility weighted imaging" in 2004 [6]. The fundamental innovation was the recognition that the phase image of a GRE acquisition — which conventional MRI discards or uses only for B0 shimming — contains quantitative information about local magnetic susceptibility differences that is not present in the magnitude image. By processing the phase image into a phase mask and multiplying it repeatedly into the magnitude image, Haacke demonstrated that small susceptibility differences (venous blood vs. brain parenchyma, haemosiderin vs. normal tissue, calcification vs. normal tissue) could be rendered with dramatically enhanced contrast.

The clinical problems SWI was designed to solve: 1. Detection of cortical microbleeds (previously invisible on standard T2* or T1) — relevant for cerebral amyloid angiopathy, hypertensive microangiopathy, traumatic axonal injury 2. Venous anatomy mapping — the deep medullary veins, cortical draining veins, and dural sinuses appear as a dark network on SWI against a bright brain parenchyma background 3. Haemorrhage staging — SWI is the most sensitive sequence for haemosiderin, deoxyhaemoglobin, and methaemoglobin, enabling detection of remote haemorrhage invisible on T1 and T2 4. Intracranial calcification — calcifications appear bright on phase images (paramagnetic compared with brain on standard phase) and dark on magnitude images, allowing calcium vs. iron differentiation that CT cannot reliably achieve from MRI data alone

SWI is now a standard component of:

• Brain trauma protocols (microbleed detection, diffuse axonal injury)
• Stroke protocols (haemorrhagic transformation, thrombus detection on mIP)
• MS protocols (central vein sign — CVS — for lesion characterisation) [1, 2]
• Paediatric brain MRI (vascular malformations, calcifications, metabolic iron deposition)
• Cavernoma detection and mapping [5]

2. Physical Foundations

2.1 Pulse Sequence Logic

SWI pulse sequence timing diagram

SWI uses a 3D spoiled gradient echo (3D FLASH/SPGR) sequence — no 180° refocusing pulse. The absence of a refocusing pulse is fundamental: without a refocusing pulse, T2 (not T2) governs transverse magnetisation decay. T2 includes T2 relaxation AND contributions from local magnetic field inhomogeneities (susceptibility differences, iron, calcium, deoxyhaemoglobin). Tissues that produce local B0 field perturbations decay faster in T2* and appear darker on the magnitude image.

The sequence acquisitions parameters are specifically chosen to maximise susceptibility contrast:

• Long TE: at TE ≈ T2*_tissue, the susceptibility contrast is maximised. For brain at 3T, optimal TE ≈ 20–25 ms; at 1.5T, TE ≈ 35–40 ms. This long TE would be too slow for a single-echo acquisition but is achievable with 3D acquisition and parallel imaging.
• Small flip angle: typically 15–20°, producing partial T1 saturation; optimised for T2* contrast rather than T1.
• Full velocity compensation on all gradient axes: flow compensation (gradient moment nulling) is applied to the readout, phase, and slice-selection gradients. This eliminates signal dephasing from flowing blood (venous blood in small veins retains signal on the magnitude image, making veins visible with their susceptibility contrast rather than being nulled by flow).
• 3D acquisition: isotropic or near-isotropic voxels (0.5–1 mm) with full brain coverage enabling minimum-intensity projections (mIP) for visualisation of the venous network.

2.2 Signal Formation: Magnitude and Phase

At echo time TE, the complex MR signal contains:

Magnitude signal: S_mag ∝ M₀ · sin(α) · e^(−TE/T2*) · (1 − e^(−TR/T1)) / (1 − cos(α) · e^(−TR/T1))

The magnitude signal reflects T2 contrast — tissues with short T2 (deoxyhaemoglobin, haemosiderin, iron, calcification) appear dark.

Phase signal: φ(TE) = γ · ΔB₀ · TE

where γ is the gyromagnetic ratio and ΔB₀ is the local field variation due to susceptibility differences. The phase accumulates linearly with TE and with the local susceptibility difference. At TE = 25 ms at 3T: iron-rich venous blood (deoxyhaemoglobin χ ≈ 0.8 ppm vs. oxygenated tissue) accumulates approximately 0.25 rad of phase per ppm of susceptibility difference per metre of static field.

2.3 The Phase Mask and SWI Processing

The SWI image is not simply the magnitude GRE image. It is produced by a multi-step post-processing algorithm:

1. Phase unwrapping and high-pass filtering: the raw phase image contains both local susceptibility information and large-scale B0 field variations (shimming residuals, tissue interfaces). A high-pass filter (typically homodyne or Gaussian) removes the slowly varying background phase, leaving only the local susceptibility-driven phase variations.

2. Phase mask generation: the filtered phase is converted into a phase mask with values between 0 and 1:

• For diamagnetic substances (calcium, most normal tissues): phase < 0 → assigned mask value > 0
• For paramagnetic substances (deoxyhaemoglobin, iron, haemosiderin): phase > 0 → assigned mask value near 0

3. Phase mask multiplication: the phase mask is multiplied into the magnitude image N times (typically N = 4 times). Each multiplication amplifies the susceptibility contrast by making regions with high phase shift darker. The final SWI image thus has contrast from both T2* (magnitude contribution) and local susceptibility (phase mask contribution).

4. Minimum intensity projection (mIP): the SWI volume is processed with mIP over 5–15 mm slabs to display the three-dimensional venous network as a continuous structure.

2.4 Diamagnetic vs. Paramagnetic: The Phase Sign Convention

This is the most clinically important concept in SWI interpretation and the source of most diagnostic errors:

Paramagnetic substances (deoxyhaemoglobin, haemosiderin, iron, manganese, gadolinium): produce local B0 field increase → positive phase shift → darker on standard phase images → darker on SWI magnitude × mask product.

Diamagnetic substances (calcium, most crystalline materials): produce local B0 field decrease → negative phase shift → appear bright on standard phase images relative to surrounding tissue.

Practical consequence: calcium appears BRIGHT on phase images and relatively dark on magnitude images (T2* effect). Iron/deoxyhaemoglobin appears DARK on both phase and magnitude images. This phase sign difference allows MRI to distinguish calcium from iron — a capability CT provides based on Hounsfield units, which MRI could not reliably achieve before SWI phase imaging.

However, the phase sign convention varies between vendors: some define positive phase as positive on the image (default), others invert the display. The radiologist must know which convention is used on their system before interpreting phase images for calcium vs. iron differentiation.

3. Key Parameters and Their Clinical Meaning

3.1 Parameter Table

3.2 Parameter Interdependence

TE and voxel size are the two most clinically critical parameters. At longer TE, susceptibility-driven dephasing accumulates more phase — the susceptibility effect scales linearly with TE and with B0. At 3T (double the field of 1.5T), the susceptibility effect is twice as large for the same TE. Therefore, the optimal TE at 3T is approximately half that at 1.5T to achieve comparable susceptibility contrast per unit SNR.

The 3D isotropic voxel is a fundamental requirement for SWI — not an optional optimisation. Partial volume effects in 2D acquisitions or thick anisotropic 3D slices average the susceptibility signal of small structures (microbleeds, small veins) with adjacent normal tissue, reducing the susceptibility contrast below detection threshold.

4. Tissue Contrast Profiles

Major Interpretation Pitfalls

Phase image display convention: calcium appears bright on phase images on some vendor implementations (standard phase) and dark on others (inverted phase). Always verify the phase convention on your specific scanner before interpreting phase images for calcium vs. iron differentiation. If the convention is unknown, use CT correlation.

Blooming artefact overestimation: the apparent size of SWI-dark lesions (haemosiderin, microbleeds, calcifications) is systematically larger than the true lesion because the susceptibility effect extends beyond the physical boundaries of the lesion through the surrounding tissue. Never estimate the size of haemorrhagic lesions from SWI magnitude images alone.

Venous blood vs. haemosiderin: both appear dark on SWI magnitude. Veins are linear and follow vascular anatomy; haemosiderin deposits are rounded and remain dark on all subsequent sequences. The clinical context and the morphology distinguish the two.

Susceptibility artefact from metallic implants: metallic clips, coils, and aneurysm clips produce massive susceptibility artefacts on SWI (blooming extending centimetres in all directions). Near metallic implants, SWI is non-interpretable and should not be included in the clinical protocol.

5. Vendor Implementations

Key Implementation Differences

Phase processing algorithm: the specific homodyne high-pass filter parameters, the phase mask generation function, and the number of mask multiplications differ between vendors. The Siemens and GE implementations use similar Haacke-derived algorithms; Philips and Canon use proprietary variants. The resulting SWI images from different vendors for the same patient at the same field strength may differ in contrast weight and small lesion detection sensitivity.

Phase image display: on Siemens, the standard phase image is displayed with paramagnetic structures as negative (dark) and diamagnetic as positive (bright). On some GE implementations, the convention may differ. All platforms provide the option to invert the phase display. This must be verified during scanner commissioning.

rSWI / QSM-assisted SWI: newer implementations (Siemens rSWI) use quantitative susceptibility mapping (QSM) processing applied to the SWI phase to generate improved images with reduced orientation artefacts. QSM processing is described in the advanced technical section.

6. Clinical Applications Overview

7. Artefacts

8. Advanced Technical Parameters

Why 3D Acquisition is Mandatory for SWI

SWI must be acquired as a 3D volume — 2D SWI (single slices) is not a valid clinical alternative. The reasons:

1. Isotropic voxels: mIP post-processing requires isotropic or near-isotropic voxels. Anisotropic voxels produce mIP images where vessels parallel to the slice plane appear thicker than those perpendicular — misleading the venographic display.

2. SNR efficiency: 3D acquisition provides the SNR advantage of √N_slices over 2D (by encoding all slices in the phase direction simultaneously), which is needed to maintain adequate SNR at the long TE and thin slices required for microbleed detection.

3. Phase coherence: phase images require spatially coherent phase across the entire brain volume. Per-slice 2D acquisitions have slice-dependent phase offsets that cannot be consistently high-pass filtered across slices.

Flow Compensation: Technical Requirements

Full flow compensation (gradient moment nulling, GMN) on all three gradient axes is the defining technical requirement of SWI vs. standard T2* GRE. GMN zeroes the first moment of each gradient waveform, ensuring that spins moving at constant velocity accumulate zero additional phase compared with stationary spins. Without GMN, flowing blood in veins (velocity typically 3–10 cm/s) accumulates sufficient flow-induced phase to partially or fully null the venous signal on both magnitude and phase images — defeating the venographic purpose of SWI.

The practical consequence: SWI has longer minimum TE than an equivalent T2* GRE without flow compensation, because the flow-compensated gradient waveforms require more time. At 3T, minimum achievable TE with full 3D flow compensation is approximately 18–22 ms. This is compatible with the desired TE for brain SWI at 3T.

Quantitative Susceptibility Mapping (QSM)

QSM is an advanced post-processing technique that uses the SWI phase data to compute a quantitative susceptibility map expressed in parts per million (ppm). Unlike the qualitative SWI image (which mixes T2* magnitude and phase), QSM provides:

• Orientation-independent susceptibility quantification (unlike SWI, where the effect of elongated structures depends on their angle to B0)
• Separation of diamagnetic (calcium) and paramagnetic (iron) contributions
• Quantitative iron load measurements for neurodegeneration research

QSM requires SWI acquisition data (the raw phase images) plus post-processing software. It is available as a research or advanced clinical tool on all major platforms. At the time of writing, QSM is not universally standardised for clinical reporting but is increasingly included in research protocols for neurodegeneration imaging.

Parallel Imaging and Compressed Sensing

The long 3D SWI acquisition (8–12 minutes at full resolution without acceleration) is the primary barrier to routine clinical use. Parallel imaging (GRAPPA R=2–3) reduces acquisition time to 4–6 minutes. Compressed sensing (HyperSense/CompSense) further reduces to 3–4 minutes. Both approaches are validated for brain SWI at 3T without clinically significant loss of microbleed detection sensitivity. At 1.5T, the lower baseline SNR limits the acceleration factor to R=2 before SNR becomes a limiting factor for small-lesion detection.

Partial Fourier (Zero-Filling)

Partial Fourier acquisition (5/8 or 6/8 of k-space) reduces scan time by 30–40% with minor Gibbs ringing penalties. For the phase images, partial Fourier introduces asymmetric phase errors that require homodyne processing for correction. Most SWI implementations use partial Fourier as a standard efficiency strategy, with the vendor's phase correction algorithm compensating for the asymmetric k-space.

9. Comparison with Alternative Sequences

SWI vs. T2 GRE (2D or 3D): T2 GRE without flow compensation is faster and simpler but inferior to SWI for venous blood detection, microbleed sensitivity, and phase-based calcium vs. iron differentiation. T2* GRE remains appropriate for rapid haemorrhage screening when SWI acquisition time is prohibitive (e.g., acute stroke where speed is paramount). SWI is the superior choice for all other susceptibility-related indications.

SWI vs. DWI for stroke haemorrhage detection: DWI detects acute ischaemia. SWI detects haemorrhagic transformation as petechial dark lesions before they become visible on CT or T2. The combination of DWI + SWI in stroke protocols provides both ischaemia mapping and haemorrhage detection simultaneously.

SWI vs. CT for calcium: CT remains the gold standard for calcium detection and quantification (Hounsfield units). SWI can detect calcium based on phase sign (diamagnetic = positive phase) but is not quantitative and can be confused with susceptibility from other sources. For calcium quantification (e.g., basal ganglia calcification grading), CT is preferred.

SWI vs. QSM: QSM is quantitative and orientation-independent; SWI is qualitative and orientation-dependent. QSM provides higher lesion specificity for research applications. For routine clinical brain MRI, SWI remains the clinical standard because QSM requires additional post-processing time and standardisation.

10. Evidence Gaps and Ongoing Debate

Central vein sign threshold standardisation: the North American Imaging in MS (NAIMS) consortium defines the CVS threshold at ≥ 3 CVS-positive lesions with high specificity for MS vs. mimics (sensitivity 0.87, specificity 0.75 per Taber et al. 2025 [2]). However, the required SWI acquisition parameters for reliable CVS detection are not fully standardised — voxel size, TE, and post-processing differences between centres affect CVS detectability. The 3T CVS has been validated; 1.5T CVS performance is less well-defined.

QSM clinical standardisation: despite two decades of research, QSM has not achieved a universally standardised acquisition protocol, post-processing algorithm, or reference values. The MRI physics community continues to debate optimal phase processing approaches (SHARP, PDF, COSMOS, MEDI). Clinical validation studies use different QSM implementations, making cross-study comparison difficult.

Microbleed quantification thresholds: the definition of cerebral microbleed (CMB) on SWI — a dark rounded lesion < 10 mm — requires careful distinction from other susceptibility sources (small veins, calcifications, flow artefacts). No universally validated automated CMB quantification algorithm exists, and manual rater concordance for small CMBs (< 3 mm) is moderate at best.

SWI at ultra-high field (7T): 7T SWI provides superior spatial resolution and susceptibility sensitivity. Clinical applications are emerging but regulatory and standardisation barriers limit widespread deployment. The expanded susceptibility artefact from metallic implants at 7T further restricts patient eligibility.

11. Miscellaneous and Related Topics

SWI in Paediatric Brain MRI

SWI is particularly valuable in paediatric neuroimaging where:

• Metabolic disorders with iron deposition (PKAN, neurodegeneration with brain iron accumulation — NBIA) show characteristic deep grey nuclei iron patterns
• Non-accidental trauma (shaken baby syndrome) detection of retinal haemorrhage analogues and cortical microhaemorrhages
• Vascular malformations in paediatric epilepsy

Paediatric SWI requires the same technical parameters as adult SWI but with careful attention to patient motion (shorter acquisition times preferred; consider sedation protocols).

SWI for Body Applications

While SWI was developed and validated for brain imaging, the principle of 3D flow-compensated GRE with phase post-processing is increasingly applied to body imaging:

• Hepatic iron quantification (liver siderosis in haemochromatosis, transfusion-related iron overload)
• Hepatic calcification
• Hepatocellular carcinoma susceptibility (tumour microvasculature)

Body SWI requires adaptation of the echo time (shorter T2* in liver), fat suppression (chemical shift interference with phase), and respiratory triggering or breath-hold acquisition.

BOLD Venography and Functional MRI Connection

SWI is directly related to BOLD fMRI through the deoxyhaemoglobin susceptibility effect. The BOLD fMRI signal derives from the same physics (T2* change from local deoxyHb concentration change during neuronal activation). SWI exploits the same deoxyHb susceptibility for static venous anatomy mapping. High-resolution SWI of the brain at 7T provides submillimetre cortical vein mapping that enables vessel-based fMRI registration.

12. Bibliography

A. Guidelines / Consensus / Society Recommendations

B. Systematic Reviews / Meta-analyses

C. Important Prospective / Original Studies

D. Technical MRI Papers

End of document — SWI Sequence Page — MRIninja v1.0 — May 2026

Pulse diagram (2.1) and magnetisation diagrams (3.2) are rendered as interactive educational widgets on the MRIninja page.