STIR (Short TI Inversion Recovery) Sequence

STIR (Short TI 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 Version 1.0 — May 2026

1. Introduction: Historical Evolution and Clinical Purpose

STIR — Short TI Inversion Recovery — is one of the most clinically important sequences in musculoskeletal, oncological, and whole-body MRI, and one of the few that has actually increased rather than decreased in clinical relevance since its introduction. Its dominance in bone marrow oedema detection, off-isocentre fat suppression, and whole-body oncological staging derives from a single physical property that no other fat suppression technique can replicate: complete independence from magnetic field homogeneity.

The STIR sequence was introduced into clinical MRI practice by Graeme Bydder and Ian Young at Hammersmith Hospital, London, in 1985 [6]. The technique itself derives from the inversion recovery (IR) family described by Hahn in 1949 and developed throughout the 1950s–1970s for NMR spectroscopy T1 measurement. Bydder and Young recognised that by choosing the TI to null the fat signal — exploiting fat's uniquely short T1 — the resulting image would show oedematous, infiltrated, and pathological tissues with dramatically enhanced conspicuity against a uniformly suppressed fat background. This was immediately clinically transformative for musculoskeletal imaging, where the contrast between oedematous bone marrow (bright) and normal yellow marrow (suppressed) represented a diagnostic capability with no CT equivalent.

The clinical problem STIR was designed to solve is specific: fat signal dominance in T2-weighted imaging. In regions of the body where adipose tissue is abundant — the bone marrow, periarticular fat pads, orbital fat, subcutaneous fat, mesenteric fat — the T2 signal of fat competes with and partially obscures the signal of pathological tissue (oedema, tumour, inflammation). Spectral fat saturation techniques (CHESS, ChemSat, SPAIR) solve this problem adequately near isocentre where B0 is homogeneous, but they fail in regions of field inhomogeneity — precisely the peripheral extremities, the small bones, and the regions that most commonly require bone marrow oedema assessment. STIR is not a frequency-selective technique — its suppression is T1-based — and therefore it cannot be defeated by B0 inhomogeneity.

The current clinical role of STIR:

• Universal bone marrow oedema screen in the spine, pelvis, and extremities
• Gold standard fat suppression for off-isocentre musculoskeletal MRI (wrist, ankle, fingers, toes, mid-foot)
• Primary oedema sequence in inflammatory arthropathy (rheumatoid, psoriatic, spondylarthropathy)
• Whole-body MRI oncology (DWIBS protocols; lymph node and bone metastasis detection)
• Orbital and optic nerve imaging
• Spinal cord and paraspinal soft tissue assessment

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 specifically on the STIR-specific physics.

STIR applies an IR preparation 180° inversion pulse, waits for TI = TI_null(fat), then applies a 90°–180° readout (SE) or a 90° followed by a TSE echo train. The TI is chosen such that fat longitudinal magnetisation Mz = 0 at the time of the readout excitation:

TI_STIR = T1_fat × ln(2) ≈ 0.693 × T1_fat

At 1.5T: T1_fat ≈ 250–270 ms → TI_STIR ≈ 150–175 ms At 3T: T1_fat ≈ 370–390 ms → TI_STIR ≈ 200–230 ms

The critical distinction from spectral fat suppression: STIR suppresses fat based on its T1 value, not its resonance frequency. Any tissue with T1 ≈ T1_fat will also be suppressed (or partially suppressed) — this is both the mechanism of the sequence and its primary limitation (see Section 7 and 9).

After the TI interval, the readout sequence excites all tissues simultaneously. The signal of each tissue at the echo time is:

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

(magnitude reconstruction)

The two terms represent: (1) the longitudinal magnetisation at TI after the IR preparation — which is zero for fat and positive for all other tissues; (2) the T2 decay during the echo time TE.

The combination of these two terms makes STIR simultaneously a fat-suppressed T2-weighted image: tissues with long T2 (oedema, free fluid, tumour) appear bright; tissues with short T2 (muscle, cortical bone, fibrosis) appear dark; fat appears nulled. This simultaneous T2-weighting and fat suppression is the diagnostic foundation of STIR's utility in bone marrow pathology.

STIR pulse sequence timing diagram

2.2 Key Property: Additive T1 and T2 Contrast

A consequence of the STIR signal equation: STIR signal is proportional to |1 − 2·e^(−TI/T1)| × e^(−TE/T2). This means that at short TI (near the fat null):

• Tissues with long T1 are less recovered → have smaller |Mz(TI)| → contribute less signal
• Tissues with long T2 retain more signal at TE → contribute more signal

The net result: tissues with simultaneously long T1 AND long T2 (oedema, tumour, free fluid) appear most conspicuously bright on STIR. Tissues with short T1 (fat) are nulled; tissues with intermediate T1 and T2 (muscle, normal marrow-replaced-by-red-marrow) have intermediate signal. This additive T1+T2 contrast is unique to STIR and is the reason STIR provides higher lesion-to-background CNR for bone marrow pathology than either T2-FS alone (which suppresses fat but has no T1 weighting effect) or T1 alone.

3. Key Parameters and Their Clinical Meaning

3.1 Parameter Table

3.2 Parameter Interdependence

TR and TI are interdependent: if TR is too short, the longitudinal magnetisation of tissues with long T1 (CSF, free fluid) does not fully recover before the next inversion pulse. The initial Mz before each inversion is then not +M₀ but a partially recovered value, which shifts the effective null point. For tissues with very long T1 (CSF: T1 ≈ 3600 ms at 1.5T), a TR of 3000 ms provides only partial recovery. With TR = 4000–5000 ms, CSF recovery is substantially more complete. The practical rule: TR must be at least 3× T1_longest_tissue in the image, which for muscle/liver (T1 ≈ 800–1000 ms) requires TR ≥ 3000 ms.

3.3 Temporal Magnetisation Diagrams

4. Tissue Contrast Profiles

Critical Interpretation Pitfalls

Incomplete fat nulling mimicking pathology: if TI is not calibrated for field strength, fat appears as intermediate signal that can mimic oedema or tumour in periarticular fat pads or bone marrow. Verify by reviewing the T1 sequence: if the structure is bright on T1 and intermediate on STIR, TI miscalibration (or post-contrast imaging) is more likely than true pathology.

STIR signal in red marrow: normal red marrow (haematopoietic marrow, predominant in axial skeleton and proximal long bones in adults) has a T1 intermediate between fat and fluid. Red marrow appears as mild-to-moderate signal on STIR — not fully dark (unlike yellow marrow) but not as bright as oedema. This is the most common source of false-positive interpretation in bone marrow STIR screening.

Short-T1 substances partially suppressed: any substance with T1 near T1_fat will be partially or fully suppressed on STIR. This includes protein-rich lesions (mucoid, protein-containing cysts), melanin, manganese-containing tissue, and haemorrhage at certain ages. These may appear falsely dark on STIR and should always be compared with T1 sequences.

5. Vendor Implementations

Key Implementation Differences

TI calibration: all vendors provide default TI values appropriate for each field strength. However, scanner-specific fat T1 measurements differ slightly between platforms. Siemens and Philips default TI values may differ by 10–20 ms from GE defaults at identical nominal field strength. This is clinically insignificant for most applications (fat null is broad enough to tolerate ±20 ms variation) but may become relevant in research quantitative STIR applications.

SAR at 3T: the extra 180° inversion pulse adds SAR to the standard TSE acquisition. At 3T, this may trigger automatic TR extension by the scanner's SAR monitoring. Using adiabatic inversion pulses (available on all major platforms at 3T) adds more SAR than standard rectangular pulses — this is the expected trade-off for improved inversion uniformity. For whole-body STIR sequences at 3T, SAR monitoring should be verified before extending coverage.

STIR-Dixon hybrid: some vendors (Philips mDixon-STIR; GE IDEAL-STIR) offer a combination of IR preparation with Dixon fat-water separation, providing B0-independent AND B1-improved fat suppression. This is particularly useful at 3T for body STIR sequences. Not widely standardised across platforms.

6. Clinical Applications Overview

When STIR is NOT the Right Choice

Post-gadolinium imaging: absolute contraindication — gadolinium T1 shortening partially or fully nulls enhancing tissue at the standard STIR TI, producing false-negative or unpredictable signal changes. Dixon or SPAIR are the fat suppression methods for all post-contrast sequences.

T1-dependent tissue characterisation: STIR intrinsically suppresses short-T1 tissues. This makes it unsuitable for fat detection (lipoma, fatty tumour, dermoid), for T1-bright haematoma characterisation, or for gadolinium enhancement detection.

Brain lesion detection: FLAIR, not STIR, is the standard brain T2 fat-suppressed equivalent. STIR suppresses fat but does not null CSF, and brain STIR sequences have lower SNR than FLAIR at equivalent acquisition time.

High-SNR demanding applications at 3T: STIR has approximately 25–30% lower SNR than equivalent PD-FS TSE at the same TR and TE, because the inversion preparation further reduces the effective available magnetisation. At 3T where SNR is the primary advantage, using PD-FS (or Dixon-FS) instead of STIR for sequences near isocentre preserves the SNR advantage.

7. Artefacts

8. Advanced Technical Parameters

ETL and T2 Blurring in STIR-TSE

STIR in clinical practice almost universally uses a TSE (turbo/fast spin echo) readout rather than a conventional SE readout (which is prohibitively slow). The ETL considerations are identical to TSE/FSE (see MRIninja TSE sequence page), with one STIR-specific note: the effective starting magnetisation for the TSE readout is the reduced value |Mz(TI)| rather than the full M₀. This means that the steady-state signal per echo is lower than in a conventional TSE at equivalent TR/TE, and T2 blurring manifests at shorter ETL than for equivalent TSE without IR preparation. For fine-structure applications (tendon assessment, cartilage, small ligaments), ETL should be limited to 8–12 for STIR, compared with 12–16 for standard PD-FS TSE.

Bandwidth Optimisation

Wider bandwidth reduces echo spacing (ES), which reduces the T2 modulation across the echo train, reducing blurring. For STIR at 3T, a bandwidth of 250–400 Hz/px is recommended to maintain echo spacing ≤ 10 ms.

Whole-Body STIR (DWIBS context)

Whole-body STIR at 1.5T or 3T is acquired as a series of overlapping multi-station acquisitions, each covering approximately 40–50 cm. The key technical parameters: TR ≥ 5000 ms (to allow adequate recovery between stations even with partial k-space sharing); TI calibrated for field strength; parallel imaging R=2; partial Fourier (5/8 or 6/8) for time efficiency. At 3T, SAR monitoring is mandatory for whole-body STIR to avoid automatic TR extension that would invalidate the nominal fat null TI.

3D STIR (SPACE STIR / CUBE STIR / VISTA STIR)

3D STIR uses a volumetric IR-TSE acquisition with VFA readout (see TSE sequence page for VFA principles). The 3D acquisition enables isotropic voxels (1–2 mm) and full MPR capability. Clinical applications: perineural oedema mapping (brachial plexus, sciatic nerve), bone marrow lesion volumetry for treatment response, complex joint anatomy with 3D multiplanar reconstruction. 3D STIR at 3T requires careful SAR management — the combination of inversion pulse + long VFA readout frequently triggers automatic TR extension on all vendor platforms.

SNR Considerations

STIR has approximately 25–30% lower SNR than equivalent PD-FS at identical TR/TE/matrix due to the reduced starting magnetisation after IR preparation. This SNR penalty must be factored into protocol design: for equivalent diagnostic quality as PD-FS, STIR requires:

• Slightly higher NSA (2 vs 1 for equivalent SNR) or
• Slightly longer TR (increasing the fraction of magnetisation recovered before readout) or
• Reduced matrix (trading spatial resolution for SNR)

This SNR penalty is the primary reason why STIR is used as a screening/oedema sequence rather than as the primary fine-structure sequence: for ligament, cartilage, and plantar plate assessment requiring sub-0.4 mm in-plane resolution, PD-FS or Dixon-FS is preferred over STIR.

9. Comparison with Alternative Fat Suppression Techniques

This section extends the comparison table in the parent IR sequence page with STIR-specific clinical detail.

STIR vs PD-FS (SPAIR/Dixon): PD-FS provides higher SNR and better spatial resolution for equivalent acquisition time. It is the preferred choice at isocentre where B0 homogeneity is adequate. STIR is the preferred choice when B0 cannot be guaranteed — peripheral joints (wrist, ankle, toes, fingers), patients with metallic hardware, obese patients with poor shim quality, and any examination > 15 cm from isocentre. In clinical practice, STIR and PD-FS should be considered complementary rather than competing: STIR for bone marrow oedema screening (STIR is more sensitive for bone marrow oedema than PD-FS due to the additive T1+T2 contrast effect), PD-FS for soft tissue structural assessment.

STIR vs Dixon: Dixon provides B0-independent fat suppression equivalent to STIR without the SNR penalty or the post-gadolinium contraindication. For post-contrast sequences, Dixon is the only acceptable alternative to SPAIR — STIR cannot be used. For pre-contrast sequences, the choice between STIR and Dixon is driven by scanner availability (Dixon is not available on all systems as a TSE option), the specific clinical question (STIR provides slightly more bone marrow sensitivity due to the T1 contribution), and field strength (Dixon is strongly preferred at 3T for T1 sequences; STIR remains the most reliable bone marrow screen).

STIR vs whole-body DWI (DWIBS): In oncological whole-body imaging, STIR and DWI are complementary. STIR provides anatomical localisation and bone marrow/soft tissue morphology; DWI provides cellularity and restriction information. The combination of STIR + DWI in the DWIBS protocol provides the most complete lesion characterisation for whole-body oncological staging [3].

10. Evidence Gaps and Ongoing Debate

Optimal TI calibration strategy: the standard TI values (150–175 ms at 1.5T; 200–230 ms at 3T) are based on consensus from technical guidelines and departmental practice. Individual variation in fat T1 due to patient age, body composition, and fat distribution has been documented but not systematically incorporated into adaptive TI selection algorithms for clinical STIR.

STIR vs DWI for bone marrow screening: multiple studies have compared STIR and DWI for bone marrow metastasis detection, with conflicting results depending on primary tumour histology, field strength, and protocol details. No definitive head-to-head randomised trial has established superiority for a specific clinical scenario across all presentations.

3D STIR clinical equivalence to 2D STIR: while 3D STIR (SPACE/CUBE/VISTA) provides isotropic voxels and MPR capability, its clinical diagnostic performance relative to optimised 2D STIR for the primary bone marrow oedema detection task has not been formally validated in prospective comparative studies with histopathological or follow-up reference standard.

STIR-Dixon hybrid availability and standardisation: the STIR-Dixon combination (available on Philips and GE with different implementations) theoretically provides the best of both worlds — B0-independent AND B1-improved fat suppression. Comparative prospective validation against standard STIR for clinical endpoints is lacking.

11. Miscellaneous and Related Topics

STIR for MR Neurography

The combination of STIR suppression (nulling background fat) with TSE T2 contrast (making nerve fluid-equivalent signal bright) produces nerve-selective contrast that is the basis of MR neurography. In peripheral nerve imaging, the nerve signal is primarily water-based and thus not nulled by STIR, while the surrounding fat is nulled. With optimised ETL and TE, nerve fascicles, ganglia, and intraneural pathology appear conspicuously bright against the dark fat background. This is a primary indication for STIR in extremity nerve imaging (brachial plexus, sciatic nerve, femoral nerve).

STIR-EPI DWI

A combined STIR-EPI sequence uses the STIR inversion preparation before EPI-DWI readout to provide T1-based fat suppression rather than spectral fat saturation. This is particularly useful at interfaces (e.g., near metallic hardware) where B0 inhomogeneity causes spectral fat saturation to fail on EPI sequences. STIR-DWI provides more uniform fat suppression but with reduced SNR compared with SPAIR-DWI. Available on Siemens as STIR-DWI (used in DWIBS protocols) and on GE.

Driven Equilibrium STIR

A driven equilibrium RF pulse applied after the TSE readout returns residual transverse magnetisation to the longitudinal axis, shortening the effective T1 recovery time and allowing shorter TR for equivalent fat null quality. This enables faster multi-slice STIR acquisitions with equivalent fat suppression. Available on selected platforms as "DE-STIR" or integrated into the standard STIR acquisition.

12. Bibliography

A. Guidelines / Consensus / Society Recommendations

(STIR is referenced as a component of disease-specific guidelines; no dedicated society recommendation exists solely for STIR sequence design.)

B. Systematic Reviews / Meta-analyses

C. Important Prospective / Original Studies

D. Technical MRI Papers

E. Landmark Historical References

End of document — STIR 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.