FLAIR (Fluid-Attenuated Inversion Recovery) Sequence

FLAIR (Fluid-Attenuated Inversion Recovery) — Physics, Parameters, and Clinical Applications

MRIninja Knowledge Base | Sequence Child Page Parent page: 9003-mri-sequences-overview-classification Parent sequence page: 9020-inversion-recovery-sequence Companion page: STIR sequence Version 1.0 — May 2026

1. Introduction: Historical Evolution and Clinical Purpose

FLAIR — Fluid-Attenuated Inversion Recovery — is the single most important brain MRI sequence in clinical neuroradiology. It is the standard for white matter lesion detection across virtually every neurological indication: multiple sclerosis, cerebrovascular disease, encephalitis, toxic-metabolic encephalopathy, cortical and juxtacortical lesions, and subarachnoid pathology. Its clinical dominance stems from solving a problem that defeated conventional T2-weighted imaging for two decades: the T2-bright cerebrospinal fluid (CSF) that fills the ventricles and sulci produces a uniformly bright background that obscures adjacent periventricular and cortical lesions.

FLAIR was introduced clinically by Graeme Bydder and colleagues at Hammersmith Hospital in the early 1990s, building on the IR sequence family established with STIR in 1985. The key innovation was recognising that a long TI — long enough to wait for CSF to reach its null point during T1 recovery — would suppress CSF throughout the brain simultaneously, regardless of the chemical composition of the CSF, because the suppression is T1-based not T2-based. At the CSF null point (TI ≈ 2200–2400 ms at 1.5T), the entire ventricular system and subarachnoid space appears dark, and any lesion with T2 signal longer than white matter — whether periventricular demyelination, an acute infarct, cortical dysplasia, or leptomeningeal pathology — is conspicuously bright against a now-dark CSF background.

The clinical problem FLAIR was designed to solve: the false-negative T2 scan for periventricular MS lesions. On conventional T2, periventricular plaques lie directly adjacent to the T2-bright CSF in the lateral ventricles. The lesion and the ventricular CSF produce similar T2 signal, making small periventricular lesions invisible or severely underestimated. FLAIR nulls the CSF signal, converting the background from bright to dark and rendering these lesions conspicuously bright. This single contrast improvement transformed the sensitivity and reliability of MS diagnosis.

The current clinical role of FLAIR:

• Primary white matter lesion detection sequence in all neurological indications
• Cortical and juxtacortical lesion detection (superior to T2 for cortical lesions)
• Subarachnoid haemorrhage detection (subacute phase; complementary to CT for delayed presentation)
• Leptomeningeal disease assessment (post-contrast FLAIR for meningeal enhancement)
• Encephalitis pattern recognition (FLAIR bright cortex/subcortex)
• Standard component of MS protocols (mandatory per MAGNIMS-CMSC-NAIMS 2021 [1])
• Standard component of epilepsy MRI (HARNESS protocol [2])
• Mandatory in all neurological MRI protocols at essentially every clinical centre worldwide

2. Physical Foundations

2.1 Pulse Sequence Logic

The complete physics of inversion recovery is described in the parent IR sequence page. This section focuses on the FLAIR-specific physics.

FLAIR uses the same 180° inversion → TI recovery → readout structure as STIR, but with a dramatically longer TI chosen to null CSF rather than fat:

TI_FLAIR = T1_CSF × ln(2) ≈ 0.693 × T1_CSF

At 1.5T: T1_CSF ≈ 3600 ms → TI_FLAIR ≈ 2490 ms (clinical range: 2200–2400 ms) At 3T: T1_CSF ≈ 4000 ms → TI_FLAIR_ideal ≈ 2770 ms

However, at 3T, the clinical TI is typically set to 1700–1900 ms — shorter than the theoretical null point. This apparent paradox has a specific physical explanation: with the long TR required for FLAIR (TR ≥ 5000–9000 ms), the steady-state Mz of CSF before each inversion pulse is not +M₀ but a partially recovered value. The effective null TI must account for this incomplete recovery, and the formula becomes:

TI_null = −T1_CSF × ln[(1 − e^(−TR/T1_CSF)) / (1 + e^(−TR/T1_CSF))]

This modified equation produces shorter effective null TI values when TR is finite — explaining why 3T FLAIR TI (1700–1900 ms) is paradoxically shorter than 1.5T FLAIR TI (2200–2400 ms) despite CSF T1 being longer at 3T. The difference arises because at 3T the longer CSF T1 means CSF recovers even less between TR repetitions, shifting the effective null point.

FLAIR pulse sequence timing diagram

2.2 Signal at the Echo

The FLAIR signal equation is identical to STIR (see parent IR page), with CSF nulled at TI:

S(TE) ∝ M₀ × |1 − 2·e^(−TI/T1)| × e^(−TE/T2)

At the long TE used for FLAIR (typically 100–150 ms), the T2 contrast is substantial:

• White matter (T2 ≈ 80 ms): retains ~37% signal at TE=100 ms
• Grey matter (T2 ≈ 90 ms): retains ~33% signal at TE=100 ms
• CSF (T2 ≈ 1800 ms): theoretically bright (long T2), but nulled by TI selection

The combination of CSF null + long TE creates the characteristic FLAIR appearance: white and grey matter are displayed at intermediate signal with high contrast relative to each other and relative to pathological tissue (which typically has longer T2 than normal parenchyma and thus appears bright on FLAIR).

2.3 The B1 Inhomogeneity Problem at 3T

At 3T, the FLAIR sequence faces a unique challenge absent at 1.5T: B1+ field inhomogeneity. The 180° inversion pulse must flip all spins by exactly 180° throughout the imaging volume to achieve complete CSF nulling. At 3T, the RF wavelength (~12 cm in tissue) is comparable to the dimensions of the brain, producing standing wave patterns that cause the B1+ field to vary by 20–40% across the imaging volume. In regions where B1+ is below the target value, the effective inversion flip angle is less than 180°, and the affected CSF does not fully invert — leaving residual CSF signal that appears as bright sulcal or ventricular signal on FLAIR.

This artefact is clinically significant because it mimics:

• Subarachnoid haemorrhage (bright CSF in sulci)
• Leptomeningeal enhancement (bright post-contrast sulcal FLAIR)
• Cortical spreading depression
• Encephalitis

The solution is adiabatic inversion pulses, which use frequency-swept RF pulses that achieve complete inversion independent of local B1+ amplitude. All major vendors implement adiabatic inversion as the standard for 3T FLAIR: Siemens uses an adiabatic half-passage variant; GE uses adiabatic frequency-modulated (AFC) pulses; Philips uses MultiTransmit B1 correction combined with standard inversion. At 3T, adiabatic inversion is not optional — it is the minimum technical requirement for diagnostic-quality FLAIR.

3. Key Parameters and Their Clinical Meaning

3.1 Parameter Table

3.2 Parameter Interdependence: The TR-TI Relationship

The TR and TI are tightly coupled in FLAIR. At shorter TR (e.g., TR = 6000 ms instead of 9000 ms), CSF does not fully recover to +M₀ before the next inversion pulse. The pre-inversion Mz of CSF is then < +M₀, which shifts the effective null point to a shorter TI. If the TI is not adjusted to compensate, incomplete CSF nulling results. All vendors implement automatic TR-TI coupling when TR is changed — the TI adjusts automatically. This coupling should not be overridden manually without recalculation of the effective null point.

3.3 Temporal Magnetisation Diagrams

4. Tissue Contrast Profiles

Key Interpretation Pitfalls

Post-contrast FLAIR artefact (sulcal bright FLAIR): when gadolinium enters the CSF — through blood-brain barrier disruption, ventricular administration, or physiological leakage — it shortens CSF T1 below the null point TI, causing bright sulcal FLAIR signal. This artefact is strongest 10–25 minutes after injection and appears as diffuse sulcal brightening. It mimics: SAH, meningitis, leptomeningeal carcinomatosis. The differential is timing: if the FLAIR was acquired in the 10–25 minute post-injection window, gadolinium artefact is likely. If acquired pre-contrast or > 30 minutes post-injection, the finding is more likely to be pathological.

Incomplete CSF nulling at 3T (B1 artefact): without adiabatic inversion, CSF in the posterior fossa, temporal horns, and sulcal spaces appears bright on 3T FLAIR due to B1 undershoot. This is the most common FLAIR quality problem at 3T in non-optimised scanners.

Fat bright on FLAIR: subcutaneous fat, orbital fat, and scalp fat appear T1-bright and are not near the FLAIR null. These appear bright on FLAIR and should not be confused with cortical or sulcal pathology. Verify against T1 — structures that are bright on T1 and bright on FLAIR are fat, not oedema.

Slow-flow bright FLAIR: slow venous flow in cortical veins and dural sinuses can produce bright signal on FLAIR due to time-of-flight effects and saturation effects. This is typically linear and follows venous anatomy.

5. Vendor Implementations

Implementation Differences

3D FLAIR contrast stability: 3D FLAIR (SPACE/CUBE/VISTA) uses VFA echo trains of 50–200 echoes with adiabatic inversion. The VFA schedule affects the grey-white contrast and the T2-lesion conspicuity — different vendor VFA implementations produce subtly different contrast profiles at identical nominal TI/TE. The MAGNIMS-CMSC-NAIMS 2021 consensus [1] accepts all major vendor 3D FLAIR implementations as equivalent for clinical MS assessment but recommends standardisation within longitudinal follow-up series on the same platform.

Post-contrast 3D FLAIR: Siemens and GE offer 3D FLAIR optimised for post-contrast acquisition (with slightly adjusted TI to account for general T1 shortening from intravascular Gd), marketed as the standard leptomeningeal enhancement detection sequence. Philips and Canon implementations are equivalent in principle.

6. Clinical Applications Overview

When FLAIR is NOT the Right Choice

Infratentorial lesions (posterior fossa, brainstem): FLAIR is notoriously unreliable for posterior fossa lesions. CSF nulling in the posterior fossa is frequently incomplete due to CSF pulsation, inflow effects, and B1 inhomogeneity at 3T. Small brainstem lesions are better assessed on T2 TSE. This is a key limitation that all neuroimaging guidelines acknowledge.

Neonatal and infant brain: in neonates and infants < 18 months, myelin water content and T1/T2 values differ substantially from adults, making the standard FLAIR TI and TR values suboptimal. FLAIR is less reliable in the unmyelinated brain and should be supplemented with T1 and T2 sequences.

Spinal cord: STIR (not FLAIR) is the standard fat-suppressed T2 sequence for spinal cord and paraspinal assessment. FLAIR can be used for cord lesion detection but its long TR and specific TI make it impractical for routine spine imaging.

7. Artefacts

8. Advanced Technical Parameters

Adiabatic Inversion at 3T: Why It Matters

At 1.5T, the RF wavelength in tissue (~25 cm) is larger than the brain dimensions, producing near-uniform B1+ distribution. A standard rectangular 180° inversion pulse achieves near-complete CSF inversion throughout the brain. At 3T, the RF wavelength shrinks to ~12 cm — similar to brain dimensions — creating standing wave interference patterns that cause B1+ to vary by 20–40% across the imaging volume. A standard rectangular 180° pulse in a region where B1+ = 0.75 × target produces only 135° effective inversion, leaving 1 − 2·e^(−TI/T1_CSF) × sin(135°/2) ≠ 0 at TI_null, meaning CSF is not completely nulled.

Adiabatic pulses use a frequency-swept RF waveform (hyperbolic secant, Silver-Hoult) that drives spins through inversion adiabatically — as long as the local B1 exceeds a minimum threshold (~30–40% of target), the inversion is complete and B1-independent. This makes adiabatic inversion robust throughout the brain including the posterior fossa, temporal poles, and cerebellum where B1+ is lowest at 3T.

2D vs 3D FLAIR

2D FLAIR (standard for routine brain imaging, still used at many centres): per-slice acquisition; independent TI optimisation; robust to between-slice motion; no inter-slice overlap (gaps if used); acquisition time 3–6 minutes; isotropic MPR not possible. Acceptable for most clinical brain indications.

3D FLAIR (SPACE/CUBE/VISTA): isotropic 1–1.3 mm voxels; full MPR; no inter-slice gaps; detection of small cortical lesions (which may fall between 3 mm 2D slices) improved; preferred by MAGNIMS 2021 [1] for MS and by HARNESS [2] for epilepsy; acquisition time 6–12 minutes (or 4–7 min with CS or SMS acceleration); more sensitive to bulk motion during acquisition.

The case for 3D FLAIR over 2D FLAIR is now substantial: a Cochrane-level systematic review equivalent (MAGNIMS consensus [1]) mandates 3D FLAIR for MS protocols. The practical adoption at many centres has been limited by acquisition time; compressed sensing and SMS can reduce 3D FLAIR acquisition to 4–5 minutes, removing the primary barrier.

SAR at 3T for FLAIR

FLAIR at 3T has high SAR because: (1) adiabatic inversion pulses are energy-intensive; (2) long TSE readout with multiple 180° refocusing pulses; (3) long TR provides limited averaging per unit time but requires sustained RF deposition. At TR = 9000 ms and ETL = 20, the SAR per TR is well within limits, but the scanner may still flag SAR concerns when combined with fast acquisition modes (short TR, multiple stacks). VFA readout in 3D FLAIR substantially reduces SAR compared with fixed 180° refocusing — this is the primary SAR management strategy.

Parallel Imaging and Compressed Sensing

For 2D FLAIR: R=2 (GRAPPA/SENSE) reduces acquisition time by half with acceptable SNR penalty. The long TR of FLAIR means the actual scan time with R=2 is still 4–6 minutes for adequate brain coverage.

For 3D FLAIR: SMS (simultaneous multi-slice, also called multiband) acceleration in the slice direction combined with CAIPIRINHA in the phase direction achieves combined acceleration factors of 4–6× without the g-factor penalties of standard parallel imaging. This is the current state-of-the-art for sub-5-minute whole-brain 3D FLAIR. Available on Siemens (SMS SPACE), GE (CUBE with HyperBand), Philips (VISTA with MB-SENSE).

9. Comparison with Alternative Sequences

FLAIR vs T2 TSE for brain lesion detection: T2 TSE shows periventricular white matter lesions against a bright CSF background — the classic "poor contrast" problem for periventricular MS plaques. FLAIR nulls CSF, transforming the background from bright to dark and making periventricular lesions conspicuously bright. For periventricular, juxtacortical, and cortical lesions, FLAIR is definitively superior. For posterior fossa lesions, T2 TSE is superior because FLAIR is unreliable in the posterior fossa.

FLAIR vs DIR (Double Inversion Recovery): DIR uses two sequential IR pulses to null both CSF and white matter, leaving only grey matter visible. This provides exceptional cortical lesion conspicuity that is superior to FLAIR for intracortical lesions in MS. However, DIR has lower SNR, longer acquisition time, and is available as 3D only on some platforms. The MAGNIMS 2021 protocol [1] recommends 3D FLAIR as the primary sequence and DIR as an optional addition for cortical lesion assessment.

FLAIR vs post-contrast T1 for meningeal pathology: Standard post-contrast T1 shows enhancing meningeal lesions. Post-contrast 3D FLAIR (acquired in the first 5 minutes or after 30+ minutes post-injection) provides an additional lesion detection mechanism through: (1) gadolinium entering the CSF from leaky meninges → CSF T1 shortens → bright sulcal signal; (2) enhancing meningeal lesions visible on both T1 and FLAIR. Post-contrast 3D FLAIR has been shown to be more sensitive than standard T1 for leptomeningeal carcinomatosis and brain metastases.

10. Evidence Gaps and Ongoing Debate

Optimal TI at 3T — theoretical vs empirical: the theoretically correct TI for CSF null at 3T (based on the steady-state correction formula) is approximately 1800–2000 ms, but published vendor defaults and clinical validation studies use 1700–1950 ms with slight institutional variation. No large multicentre study has prospectively validated the optimal 3T FLAIR TI against diagnostic outcome endpoints across all lesion types.

3D FLAIR acquisition time and diagnostic equivalence with compressed sensing: publications show that compressed sensing (CS)-accelerated 3D FLAIR achieves equivalent lesion detection to fully sampled 3D FLAIR at 50–60% of the acquisition time. However, CS reconstruction can introduce texture differences and smoothing artefacts that may alter radiologist performance for subtle cortical lesions. The clinical equivalence threshold for CS acceleration in 3D FLAIR has not been formally established.

Post-contrast FLAIR timing standardisation: the optimal timing window for post-contrast FLAIR to detect leptomeningeal enhancement (before gadolinium artefact overwhelms genuine enhancement signal) varies between studies: some suggest 5 minutes, others 10 minutes. No randomised protocol has defined the optimal post-injection timing for specific pathologies.

FLAIR-DWI mismatch for wake-up stroke: the WAKE-UP trial established DWI-FLAIR mismatch (DWI positive, FLAIR negative) as an imaging biomarker for thrombolysis eligibility in wake-up stroke. Ongoing debate concerns the specificity of FLAIR negativity as a proxy for onset-to-image time, given that FLAIR can be negative for up to 4–6 hours after infarct onset in some cases.

11. Miscellaneous and Related Topics

Post-Contrast FLAIR for Brain Metastasis

Post-contrast 3D FLAIR has been proposed as a supplement to standard post-contrast T1 for brain metastasis detection. The rationale: gadolinium in the CSF surrounding small leptomeningeal deposits brightens the sulcal background, making small surface metastases conspicuous. Published series show improved small metastasis detection with post-contrast FLAIR but standardisation of injection-to-acquisition timing is required.

T2-FLAIR Mismatch Sign in Glioma

The T2-FLAIR mismatch sign — T2-bright tumour with FLAIR signal lower than the T2 signal — has been described as a relatively specific imaging marker for IDH-mutant 1p/19q non-codeleted astrocytoma (WHO grade 2–3). The pathophysiological basis is the homogeneous myxoid matrix of these tumours, which has a T1 near the FLAIR null point. This is a child-page topic (glioma imaging) but the FLAIR-specific mechanism is documented here for completeness.

FLAIR in Posterior Fossa: A Persistent Limitation

The posterior fossa limitation of FLAIR — incomplete CSF nulling from pulsation and B1 inhomogeneity, combined with susceptibility from petrous bone and mastoid air cells — means that FLAIR cannot replace T2 TSE for brainstem and cerebellar lesion detection. The standard brain protocol therefore always includes both FLAIR (for supratentorial lesion detection) and T2 TSE (for posterior fossa). This is a protocol design principle, not a limitation to be solved by optimisation.

12. Bibliography

A. Guidelines / Consensus / Society Recommendations

B. Systematic Reviews / Meta-analyses

C. Important Prospective / Original Studies

D. Technical MRI Papers

E. Landmark Historical References

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

Pulse diagram (2.1) and magnetisation diagrams (3.3) are rendered as interactive educational widgets on the MRIninja page.