You are here: Home Equine Articles STIR-ing the Pot

STIR-ing the Pot

How STIR sequences can aid diagnosis by standing MRI


One of the great advantages of MRI is that the images it creates are based on signals from fat and water, giving physiological as well as anatomical information about the patient.

The signals read by MRI are displayed as shades of grey. On most MR sequences both fat and water appear the same shade, rendering them indistinguishable.  In the standing MRI system the STIR sequence is most commonly used to see just water, suppressing the signal from fat.

The STIR sequence comprises three parts: an initial inversion pulse, a delay time, and a readout image formation block.   The MRI world loves its acronyms, and STIR stands for "Short τ Inversion Recovery" where τ, the Greek letter tau, itself stands for the delay time, more commonly referred to as the inversion time T.I. .  Choosing the right τ, or T.I., value is critical.

The STIR pulse sequence, consisting of an inversion pulse A, a delay time TI or τ, and a readout block (in this case a spin-echo) including further pulses and the acquisition of the signal echo

The inversion pulse flips the magnetisation from both fat and water by 180°.  Starting from the fully relaxed, equilibrium condition this aligns the magnetisation in the opposite direction to the applied magnetic field. 

From there it will gradually return to the equilibrium alignment, with a time constant T1.  The magnetisation due to nuclei in fat recovers slightly faster than that in water, due to the increased mobility of the smaller water molecules.  After a certain time the magnetisation in the fat has recovered exactly half-way between fully inverted and fully at equilibrium, becoming zero.  The magnetisation in water, recovering more slowly, is not zero but instead remains at a lower, negative value. 

Collecting an image at this point will give no signal from fat, while still giving some, if weak, signal from water.

Immediately following the first pulse of the STIR sequence the magnetisation becomes inverted.  It recovers broadly on two exponential curves, with different time constants T1, toward the equilibrium state.  Protons in fat have shorter T1 than those in water, so recover faster. 

After the time TI, or τ, magnetisation from fat has reached a null position while that from water remains. 

The image formation process is insensitive to the sign of the magnetisation, so the negative magnetisation from water still gives a positive signal in the image (dotted line)

STIR Bone bruise

STIR image (top left) showing a region in the middle phalanx containing signal from water, while the remainder of the bone is black as signal from fat has been suppressed. The other three images show different sequences/orientations of the same bone bruise.

To get the best images from the STIR sequence you need to pay attention to each phase

Inversion pulse

In a perfect world the RF coil surrounding the sample (in our case a horse limb) will be perfectly uniform, and an exactly measured duration and power of RF pulse would flip the magnetisation exactly 180° throughout the sample.  But in reality the field generated by the coil is not uniform - to make it so would also make it weaker, and so give a poorer quality image.  So a clever type of pulse is used (called a hyperbolic secant adiabatic inversion pulse) where the magnetisation over a defined frequency bandwidth (and thus the target slice in a multi-slice imaging sequence) spirals down to the inverted state.  A bit too much power just has the magnetisation staying put fully inverted.  Magnetisation outside the defined slice also tips down, but if it doesn't get as far as half-way it then spirals back up again during the second half of the pulse, ending up back in the equilibrium position and giving no signal.

However insufficient power leaves magnetisation spread all over the place, so it is vital before starting a STIR sequence to make sure the RF power calibration (coil calibration) is reasonably recent, and repeat it if not.

τ delay

The recovery toward equilibrium is determined by spin-lattice relaxation, which is very dependent on molecular mobility.  This in turn depends on the temperature, on the size of the molecules containing the hydrogen atoms being observed, and on other restrictions such as diffusion limitation.

Neither fat nor water in biological tissue are pure and homogeneous, so the time constants T1(fat) and T1(water) only approximately represent the true recovery of magnetisation.  Nevertheless there is a time TI at which the signal fat is well suppressed, and in general this is determined by measurement in a sample representing the biological tissue of interest. The optimum TI can be different for different patients. Older horses and certain breeds have a slightly shorter optimum TI, as do limbs at a lower temperature. 

While the precise optimum could be measured for every individual case this would take time, and is not necessary as an approximate value will do.  Hallmarq provides the STIR TEST sequence which quickly collects STIR images with three different durations of TI: short  (-), standard, and long (+).  This test should be run on every case, and the correct STIR chosen based on the image with the best fat suppression.  In case of equally good suppression, choose the most negative as this has the shorter TI and will give a better overall image quality.


STIR TEST image: The position of images with different TI durations when displayed on a 2x2 layout

To be useful the STIR TEST slices must pass through tissue expected to show normal signal from fat. Sagittal or transverse slices through the middle phalanx are a good choice for the foot.

STIR TEST comes with two provisos.

  1. Because the choice is made on the basis of optimum suppression, it is possible to artificially suppress the signal from abnormal tissue.  This is unlikely as the TI giving a null for abnormal tissue is generally outside the range of the (-) and (+) test options.  However when choosing the optimum STIR you should consider whether the choice is reasonable given the age, breed, and temperature of the horse.
  2. The optimum suppression is not necessarily the darkest image.  Even normal bone medulla will contain some water as well as fat, so even after complete suppression of the fat there will still be a slight grey tone to the image.  Selecting for a perfect black will tend to select too long a value for τ, at which point the magnetisation from fat will have recovered to a slightly positive value while the magnetisation from water is still negative, leading to zero net magnetisation.  Such over-suppression reduces sensitivity as what magnetisation there is from water is cancelled by the over-recovered magnetisation from fat.

The STIR images with longer τ values also have lower SNR, giving a more noisy image, again increasing risk of missing pathology.  Hallmarq recommends that the darkest image should normally be selected, but when the levels of darkness are similar the scan with the shortest τ value (eg STIR when choosing between STIR and STIR(+) ) is the best choice.

In very young (< 1 year) horses the bone marrow does not contain the same dominance of fat-containing tissue, so none of the STIR sequences will give good fat suppression.

Readout image

Conventionally the inversion pulse and delay is followed by a spin-echo imaging block.  The rationale for this is that a long TR will be required between successive pulses anyway to allow full recovery after the inversion, so there is no speed benefit in collecting a gradient echo.

While this argument is broadly true there is some speed advantage to the gradient echo.  TR can be reduced to give sufficient rather than full relaxation, and the image collection block is itself shorter. Gradient echo images are also less sensitive to motion artifact, and this can be an advantage e.g. when looking for lesions in TRA scans of the flexor tendons.

Hallmarq therefore provides both a spin echo and a gradient echo version of each STIR sequence.   In common with their non-STIR versions, spin echo sequences generally have more contrast but lower SNR, and conversely gradient echoes have a higher SNR but lower contrast.  In both cases therefore there is the potential to miss some signal from water - either lost in noise, or with insufficient contrast from the surrounding tissue. 

Hallmarq strongly recommends gaining familiarity with one or other sequence before attempting to interpret images.  In particular is it unwise to interpret a gradient echo STIR based on experience of spin echo images, as the higher SNR but lower contrast could be mistaken for a high quality image showing no fluid.

Confirmation of suspected fluid

STIR images are inherently of lower SNR than other sequences.  To compensate the slice thickness of the standard sequences is increased, but the SNR of a STIR image is still low.  Care is needed when interpreting so as not to either miss lesions, or over-interpret noise as lesions.  

Confirmation of any suspected fluid should be obtained either from a STIR image collected in a different orientation, from at T2W image, or from the out-of-phase fat-water cancellation effect of a T2*W image.

As partial compensation for the lower SNR, the slices of a STIR image are thicker than that for sequences of other contrasts.  This can mean that pathology is only visible in one STIR slice, while appearing in two or more slices of T1, T2* or T2 contrast.  Again confirmation is needed from other contrasts and/or other orientations to decide whether any suspected signal intensity is due to real signal from water, or artifacts due to increased noise.

Filed under: