What you’ll be able to do after this lesson
- Explain how a 180° inversion pulse plus a chosen TI nulls a tissue
- Derive TI_null from T1 and say why the clinical value differs from the textbook one
- State when STIR beats spectral fat suppression, and the one situation where STIR is the wrong tool
- Explain what FLAIR removes and why that makes periventricular lesions visible
Inverting first, then waiting
Inversion recovery adds one step before the ordinary sequence: a 180° pulse applied to fully recovered magnetisation, which drives it from +z to −z. Nothing is transverse, so nothing is detectable. Then you wait.
During that wait, each tissue climbs its own T1 curve from −M0 back towards +M0. Every curve must pass through zero on the way, and each does so at a time set by its own T1. Fire your 90° excitation pulse at the exact moment a chosen tissue is crossing zero, and that tissue has no longitudinal magnetisation to tip. It contributes nothing to the image. It is nulled.
The delay between the 180° and the 90° is the inversion time, TI, and it is the only new parameter. Everything after the 90° is a normal spin echo or fast spin echo, with its own TR and TE doing their usual jobs.
Choosing TI
For magnetisation starting from full inversion with a long TR, the null occurs at:
TI_null = T1 × ln 2 ≈ 0.69 × T1
Fat’s T1 at 1.5 T is conventionally taken as about 260 ms, which puts its null at about 180 ms — the textbook figure, and the one Elster quotes for a very long TR.
Clinical STIR sequences usually sit a little below that, commonly between about 150 and 180 ms at 1.5 T. This is not an error. Elster gives two reasons, and neither is the one people usually assume: adipose tissue contains variable amounts of water, so its effective T1 is not a single number; and the readout is nearly always fast spin echo, whose train of extra 180° pulses and gradients perturbs the recovery the simple formula describes. Published fat T1 measurements themselves range from roughly 230 to 350 ms depending on site and method, which is enough on its own to move the calculated null by tens of milliseconds.
There is a third effect that matters much more elsewhere in this lesson than it does here. TI_null depends on TR as well as T1, because magnetisation that has not fully recovered before the inversion starts from less than −M0 and therefore reaches zero sooner. For fat, with a T1 of ~260 ms and a STIR TR of 3000–4000 ms, that correction is negligible. For CSF on FLAIR, where T1 and TR are the same order of magnitude, it is large — and it is why FLAIR inversion times differ so much between sequences, as the next section shows.
What travels to every console is simpler: TI_null depends on T1, on TR, and therefore on field strength. A TI copied from a 1.5 T protocol will not null fat correctly at 3 T, where T1 is longer and the required TI is longer — roughly 190–230 ms is the range usually quoted. Do not port inversion times between field strengths, and use your scanner’s own null calculator rather than a remembered number.
STIR: nulling fat
Set TI to the fat null and you have STIR — short-tau inversion recovery. Fat goes black. What remains is a T2-weighted image with the fat signal removed, and because oedema and most pathology have long T1 and long T2, they end up conspicuously bright against a dark background. STIR is the most sensitive routine sequence for marrow oedema for exactly this reason: the two contrast effects add rather than compete.
Why STIR sometimes beats spectral fat saturation. Spectral fat-sat works by frequency: it transmits a pulse tuned to fat’s resonance, about 220 Hz from water at 1.5 T, and spoils that signal. It therefore depends on the field being uniform enough that fat everywhere in the field of view really is at that frequency. Off-isocentre, near metal, in a large or awkwardly shaped body part, or in a poorly shimmed shoulder, it fails — patchily, on one side of the image.
STIR does not care about frequency. It works on T1, which is a tissue property that field imperfection barely disturbs. So STIR is the robust choice where shim is difficult: shoulders, off-centre extremities, whole-body imaging, near metal.
The costs: STIR has lower SNR than a comparable T2 fat-sat sequence, because you sample magnetisation that has recovered only partway from inversion, and it takes longer, because the inversion and the wait sit inside every TR.
The one situation to be careful about. STIR nulls by T1, not by chemical composition — it nulls anything whose T1 sits near fat’s. Gadolinium shortens the T1 of enhancing tissue, potentially towards the fat-like range, so STIR after contrast can suppress the enhancement you gave the contrast to see. The conventional teaching is to use spectral fat-sat, or Dixon, for post-contrast fat-suppressed T1 imaging rather than STIR. Protocol choices around contrast are your department’s to set — the physics point is simply that STIR’s selectivity is T1-based, and gadolinium changes T1.
FLAIR: nulling CSF
Same mechanism, different target. CSF has a very long T1 — around 3000–4000 ms at 1.5 T, longer again at 3 T — so its null falls at a long TI, conventionally quoted around 2000–2500 ms on clinical sequences.
This is where the finite-TR correction earns its keep. The full expression is TI_null = T1 × ln[2 / (1 + e^(−TR/T1))], and with TR only two or three times T1 it makes a difference of hundreds of milliseconds. For CSF with T1 ≈ 4300 ms at 3 T, a TR of 9000 ms puts the null near 2500 ms, while a TR of 5000 ms — typical of a 3D FLAIR — puts it near 1800 ms. Both are correct; they are answers to different sequences. A FLAIR TI is therefore only meaningful alongside its TR, which is why 3D FLAIR protocols legitimately quote inversion times far shorter than 2D ones at the same field strength.
Set TI there, then run a long TE readout, and you get a T2-weighted image with the CSF removed. That combination is the point. On plain T2, bright CSF sits directly against periventricular white matter, and a bright lesion abutting a bright ventricle is hard to see. Null the CSF and the lesion stands out against black. FLAIR is consequently the workhorse for periventricular and cortical/juxtacortical disease, and for subarachnoid signal that should not be there.
Two practical caveats belong in the same breath. CSF flow — particularly pulsatile flow through the aqueduct, fourth ventricle and basal cisterns — can spoil the null and produce bright CSF where you expect black, a classic mimic. And on a fast spin echo readout, the effective TI depends on the sequence timing, so vendor defaults differ; incomplete nulling is a sequence-parameter problem, not necessarily pathology.
At the console
An ankle for suspected stress fracture. The site protocol offers sagittal T2 with spectral fat-sat and sagittal STIR, and the list is running behind, so you consider dropping the STIR.
Two questions decide it. Is this anatomy well shimmed? An ankle is small and awkwardly shaped and often not perfectly at isocentre, which is exactly where spectral fat-sat becomes patchy — and patchy fat-sat over the calcaneus is precisely the failure that hides a stress fracture. And what is the finding made of? Marrow oedema: long T1 and long T2, the combination STIR is built to maximise. On this indication the STIR is the sequence you keep and the other you consider dropping.
Second case. A post-contrast fat-suppressed T1 of the spine comes back and the enhancing lesion the previous study showed is far less obvious. Check what produced the fat suppression. If someone has substituted STIR for spectral fat-sat after gadolinium — perhaps for the perfectly good reason that the spectral fat-sat was failing near instrumentation — the contrast-shortened T1 of the enhancing tissue may have moved it towards the null. The image is behaving exactly as the physics says it should. Dixon-based fat suppression is the usual answer to that combination of problems, and what your department uses is a protocol decision to raise, not to improvise mid-list.
Check yourself
Without a calculator: a tissue has T1 = 900 ms at 1.5 T and you want it nulled with a very long TR. Roughly what TI? Now, that same tissue at 3 T has T1 around 1300 ms — does the TI go up or down, and by roughly how much? Finally, explain in one sentence why STIR keeps working over a shoulder when spectral fat-sat does not.
This is a private prompt. Nothing you write is stored or assessed.
Key takeaways
- An inversion pulse drives magnetisation to −z; each tissue crosses zero at a time set by its own T1. Exciting at that instant nulls it.
- TI_null ≈ 0.69 × T1 when TR is long, and TI_null = T1 × ln[2/(1 + e^(−TR/T1))] when it is not. Clinical STIR TIs sit slightly below the textbook 180 ms at 1.5 T because fat is not a single T1 and because the FSE readout perturbs recovery; FLAIR TIs vary widely because they track TR. All of it shifts with field strength.
- STIR nulls fat by T1, so it is robust where shim is poor: off-isocentre, near metal, whole-body. It costs SNR and time.
- STIR’s T1 selectivity is also its limitation: gadolinium shortens T1 and can push enhancing tissue towards the null.
- FLAIR nulls CSF with a long TI and reads out with a long TE, making periventricular and juxtacortical lesions conspicuous. Watch for flow-related incomplete nulling.
References
- Bydder GM, Young IR. MR imaging: clinical use of the inversion recovery sequence. Journal of Computer Assisted Tomography. 1985;9(4):659–675 — PMID 2991345 https://pubmed.ncbi.nlm.nih.gov/2991345/
- Hajnal JV, Bryant DJ, Kasuboski L, et al. Use of fluid attenuated inversion recovery (FLAIR) pulse sequences in MRI of the brain. Journal of Computer Assisted Tomography. 1992;16(6):841–844 — PMID 1430427 https://pubmed.ncbi.nlm.nih.gov/1430427/
- Elster AD. Questions and Answers in MRI — ‘TI to null a tissue’ and ‘STIR’ https://mriquestions.com/ti-to-null-a-tissue.html
- Radiopaedia. Short tau inversion recovery (STIR); Fluid attenuated inversion recovery (FLAIR) https://radiopaedia.org/articles/short-tau-inversion-recovery
- Bernstein MA, King KF, Zhou XJ. Handbook of MRI Pulse Sequences. Elsevier Academic Press, 2004 — §14.2, inversion recovery