TMS timing test: with vs without TMS pulses

[BenInglis]

Test of data stability for TMS pulses timed to coincide with the 4 ms crusher gradient preceding the fat saturation pulse in ep2d_neuro.

Doby's test data from 9th April, 2015. 200 total EPI volumes, the first 100 of which were acquired as follows (from Doby's email):

"the 200-volume time series was designed to test to allow me to run different analyses. Each of the first 100 volumes received 5 TMS pulses targeted at the first crusher gradient of slices 2, 4, 6, 8, and 10. The last 100 volumes were TMS free."

In the summary statistical image, on the left is the std dev for the first 100 volumes (with TMS pulses) on the right the std dev for the second 100 volumes:

The only visible artifact arises because of TMS coil heating. The TMS coil has created a magnetic susceptibility gradient in the phantom signal and the additional local field drift caused by the TMS coil warming up is changing the distortion/dropout. But there is no other visible consequence of the TMS pulses.

Here is a time series of an ROI in slice 4 for the whole 200 volumes:

It's arguable if there is any difference between the first and second 100 volumes in the series, I think we need stats to differentiate anything. But we see a clear scanner drift in the entire series. Might the presence of the warming (then cooling) TMS coil be generating additional variance? I don't see anything big enough to be attributable to direct effects of TMS pulses in this slice.

[BenInglis]

In the above data, slice 1 is at bottom right in the mosaic, slice number increases right to left and from bottom row to top. So slice 4 is 4th slice from the bottom right corner.

[danshel]

So we should redo the test keeping the TMS timing relative to MRI the same but this time move the coil to a location at the opposite end. I think we should do two such tests: One with the coil right up against the phantom as done for the above data and one with the coil at distance approximately equal to the scalp to pia mater distance.

[BenInglis]

New timing tests with the first TMS pulse applied 38 ms after the TTL. A total of 220 TMS pulses were applied in increments of 0.1 ms from 38 ms, i.e. 38-60 ms from the TTL. This takes the TMS pulses through the tail end of the k-space of the target slice, through the fatsat then the slice select RF of the next slice in the stack. These effects are seen clearly in MeanCurve analysis of the full EPI time series.

First, this is the effect of the early delays on the high frequency (late portion) k-space of the target slice (blue trace) compared to a control slice (yellow trace). Note that there was minimal effect in the signal region by eye, so a noise ROI was selected:

Next, for the slice >40 ms after the TTL there is significant interaction of the TMS with both the fatsat and slice select RF pulses. The interaction actually perturbs the signal in an ROI in a manner consistent with the amplitude modulation of the RF pulses themselves:

A voxelwise map of the standard deviation of the time series revealed minor effects in two or three adjacent slices due to interaction of the TMS pulses with the fatsat RF pulse. (Compare to the yellow trace in the previous figure which reveals minimal interaction of the TMS pulse with the slice select RF for those slices.) Also visible in the stdev image are the effects in high frequency k-space in the target slice, and the large effect of TMS-slice select interaction for the subsequent slice: