The paper that explains the sequence and its parameter regime is can be found here: https://www.ncbi.nlm.nih.gov/pubmed/23963641. The sequence diagram looks something like this:
There are different readouts implemented so far:
- single-shot EPI
- SMS-Multi-echo (with the WIP recon from MHG)
They are all implemented in VB17-UHF.
The Inversion time “TI2” (as visible in the sequence editor) refers to the time between the center of the inversion pulse and the center of the first excitation pulse of the first readout module.
The other inversion time parameters TI1 and TI1s are not used and are remnants from earlier versions of the sequence. Just leave them at the very short default value.
The time difference of the two readout modules is adjusted with the sequence card “TR” parameter. It is adjusted in a way that all images are equally spaced across time. There is no option to independently adjust the second readout pulse from the first one.
The sequence is set up in a way that the delay period between BOLD and VASO is the same as the delay period between VASO and BOLD. This is theoretically not necessary for the contrast generation of VASO. Its just the easiest way how it is implemented in the sequence code.
Calculation of the blood nulling time.
Conventional steady-state VASO approaches use the equation TI = T1 * (ln (1+chi) – ln(1+chi*exp(-TR/T1))). With chi being the inversion efficiency [0:1]. Since SS-SI VASO assumes that the blood magnetization is inverted only once, I use the equation TI = T1 * ln(1+chi).
The inversion efficiency chi is affected by two things: (A) the phase skip of the inversion pulse, (B) T2 relaxation during the adiabatic inversion. The (B) part is negligibly small and is often ignored. If you want to include everything in one equations it reads:
TI = T1_blood * ln(1+cos(phase_skip)*exp(-(0.5*pulse_duration)/(T2_blood))). Usually, I assume a blood T1 = 2100 ms, blood T2 = 45 ms. In M1 I use a phase skip of 30deg. In V1, I use a phase skip of 0deg (with Nova inversion coils).
In V1, I would end up with desired TIs of 2100 * ln(1+cos(0)*exp(-(0.5*10)/(45))) = 1342 ms.
In M1, I would end up with desired TIs of 2100 * ln(1+cos(30)*exp(-(0.5*10)/(45))) = 1200 ms.
Adjustment of the TI parameter in the protocol editor of the sequence.
Note, that the above discussion refer to effective “TI”, the blood nulling time. The desired time of the blood signal acquisition. The “TI” in the sequence editor, however, only refers to the first excitation pulse, not the center time point of the imaging volume. Hence, when the readout block takes about 600ms. The “TI” in the sequence editor must be correspondingly earlier.
In M1, for a readout duration of 600 ms, I would adjust the “TI” in the sequence editor to 900 ms. In V1, it would be 1050 ms.
Inversion pulse parameters
The adiabatic inversion pulse is calculated in real time and can be adjusted within the sequence editor. Optimal parameters are duration 10 ms, BWDTH 300%, amplitude between 90 and 110 %.
Those are also the default values and have been adjusted based on a series of pilot experiments (http://pubman.mpdl.mpg.de/pubman/item/escidoc:1752750:3/component/escidoc:1752749/VASO_Thesis_Huber.pdf Figures 6.2, 6.3, 6.15, 6.18, 6.22).
The performance of the pulse should be stable -independent of the amplitude. Only if the adiabaticity threshold is undershot, the inversion efficiency might suffer. The minimum B1+ needed for good inversion efficiency is 7-10 muT. With the Nova coil, this is achieved with amplitudes of > 90% (Ref ampl 220V). There is almost no disadvantage of going to high amplitudes. It is only limited by SAR. Common inversion efficiency parameters are in the range of 90% (in phantoms).
If you change the bandwidth, it will affect the sharpness of the pulse (Fig. 6.3). Values below 300 % (6.355 kHz) make the pulse less sharp. Higher values reduce the adiabaticity and make the pulse more susceptible to B1+ inhomogeneities.
The thickness parameter is only used, when the Phase-skip is switched off. For a head transmit coil, it is advised to always use a phase skip. E.g. in visual cortex for flickering checkerboard use, Phase-skip = 1. In Motor cortex for Phase skip use 30 deg. When doing breath hold , phase skips of 60deg might ne necessary (with loss in SNR).
While the original VASO paper shows gradients during the inversion pulse, I have come to the conclusion that at 7T it is best to have the slab-thickness as large as possible. Hence, from the sequence code perspective, the inversion is a global inversion. In reality it is an “slab-selective” inversion, where the slab thickness is determined by the size of the transmit coil. Only at 3T with a body transmit coil the gradients become important again.
Since the advent of VASO in 2003, inflow effects and corresponding CBF-contamination have been a serious concern in functional VASO experiments. Manus Donahue investigated inflow effects for conventional VASO variants with steady-state blood in his paper. He found out that in conventional VASO, inflow effects result in a negative VASO signal change. Hence, they amplify the negative CBV-weighted VASO signal change.
This is different for SS-SI VASO. In SS-SI VASO, inflow effects have the opposite effect. There, CBF contaminations are caused by inflow of blood that has not been inverted at all. I.e. blood that was outside the inversion coverage below the circle of willis during the adiabatic inversion and has flown into the imaging slice until the signal is acquired. This ‘fresh’ blood has a large (equilibrium) magnetization. When the vessel is engaged in the functional task and changes its diameter and flow velocity, the arterial arrival time becomes shorter during activation. As a result, the amount of the large arterial equilibrium magnetization increases even more and results in a CBF-dependent signal increase during activation, which counteracts the negative CBV-contrast in VASO.
The occurrence of such inflow effects is highly dependent on multiple experimental aspects, including: transmit coil coverage, inversion pulse efficiency, and functional task. Since the arterial arrival time is highly variable across brain areas, inflow effects are differently pronounced across ROIs.
While local activation tasks (e.g. flickering checkerboards and/or finger tapping) do not change the arterial arrival times of the large vessels and result in minimal inflow effects, global tasks (e.g. breathhold, CO2-breating) are expected to be much more susceptible to inflow effects.
Identification of inflow effects
- Looking out for bright voxels. In an IR sequence like the VASO sequence, large vessels are refilled for short TIs and are filled with equilibrium magnetization after 200-1000 ms. Hence, they look like very short T1 compartments. Thus, also in VASO, inflow of fresh blood results in easily visible isolated bright voxels (Van De Moortele et al., 2009). Here, we can utilize this effect and quantify the occurrence of inflow-driven voxels as a measure of how much VASO is contaminated by inflow. The very large arteries, that are detectable with this trick do not necessarily pose a problem in functional VASO, because they are not functionally engaged in the functional task. That is because these vessels are similarly refilled with fresh blood during rest and during tapping, so they do not mask the tapping-induced signal changes.
- Looking out for inverse VASO contrast: Inflowing blood has a large magnetization. If the vessel is engaged in the functional task, the arterial arrival time becomes shorter and the amount of this large arterial magnetization increases even more during activation. This results in a CBF-dependent raw signal increase. So, Inflow effects are clearly visible as an inverse VASO contrast. In functional VASO data, this effect is clearly visible above the cortical surface.
Dealing with inflow effects
- Excluding voxels from further analysis. I found that in the majority of cases, inflow effects are very local and easily detectible. Since inflow-voxels exhibit an inverse VASO contrast, they are already inherently excluded from statistical activation maps.
- Reducing the blood-nulling time (TI). This can be achieved by reducing the inversion efficiency of the adiabatic inversion pulse. This approach has been described in section 4.3.1 (Page 123ff) of the PhD thesis here.
Tipps for maximal SNR
- FLASH grappa is better than LIN-PAR (FLEET) is better than PAR-LIN is better than single-shot is better than segmented
- make the TR as short as possible. in conventional VASO is reduced the SNR. in SS-SI-VASO in maximizes the SNR.
- Since, its an inversion-recovery sequence the Ernst-angle can be bigger than 90deg! A flip angle in the range of 120deg has the best SNR. A flip angle in the range of 60deg has the nicest T1-weighting.
- Making a longer TI, increases SNR
- Keep TE as short as possible.
- Do the BOLD correction with the interleaved division instead of the T2* fitting to hypothetical TE=0ms.
- With the high-res 3D-EPI protocol after version 121, Flip angle schemes of 4 or 26 are pretty good. The PSF in slice direction is not perfect, however.
Confusing UI artifacts in protocol editor
- All three labeling schemes are doing the same thing: SS-SI VASO. Previous versions of the sequence had also options of Q2TIPS in the UI. These options were not functional and used SS-SI VASO either way.
- The saturation band in the GUI can be ignored. It is also an artifact from ASL. The pulses are not played out and the saturation band in the patient localizer have absolutely no meaning. I just don’t know how to program the GUI, so they still appear in the viewer, even though the corresponding pulses are excluded. The inversion pulse is automatically adjusted to the imaging volume (without showing it in the GUI)
Tipps for adjusting the protocol
Order of adjusting the parameters
- Number of slices
- Acceleration in slice direction
- Acceleration in phase direction
Issues in adjusting the parameters
- NEVER EVER TRUST THE SOVLVE HANDLER.
- In the worst cases is freezes Syngo,
- If you trust the PNS calculation in one direction it can shut down the gradient amplifier -> Reboot
- Do not go into the read area of parameters. It will open the solve handler saying e.g. that the TR will be adapted. In both cases (clicking “ok” or clicking “undo”) will gray out the entire protocol editor. -> insert sequence from scratch.
- In latest 3D-EPI versions, the FA is not given in degrees, it rather refers to the #th column in a matrix.
Tested Sequence Protocols
I learned that different sites can achieve different TEs ?!? Not clear why. Maybe the gradients are calibrated differently, allowing different effective slew-rates?
TI timing in SMS-VASO
Of course, the TI calculation is dependent on the slice order (Ascending/Descending/Interleaved) in the field “Series:
I think the default is “Descending”, so I will try to explain this scheme in more detail here.
The necessary parameters are:
- #Slices = the total number of slices in the volume
- #SMS = SMS factor, the number of simultaneously acquires slices.
- #SlicesInBlock = number of slices in one block = #slices/#SMS
- TI = inversion time
- TRmin = shortest possible TR allowed by sequence protocol.
- AcquisitionDuration = time that is used to acquire all slices = TRmin – TI
- DeltaTI = time between two consecutive RF-excitation pulses = AcquisitionDuration/#SlicesInBlock
For an descending acquisition order, the top slice is acquired first. And in SMS it is simultaneously acquired with slice 1 x #SlicesInBlock, 2 x #SlicesInBlock, 3 x #SlicesInBlock ….
So, based on the number of slices, the SMS-factor and the Series, the acquisition order can be inferred.
The more complicated part is to find out DeltaTI. Option 1 is that it is given out from the sequence code with mySeqLoop.getlKernelScanTime() (in micro seconds). If you don’t have the sequence code it can be estimated by calculating the difference TRmin – TI and dividing it by the SMS factor. TRmin can be seen in the protocol editor at the scanner: