3D EPI Scanning with high temporal resolution

by | Jan 10, 2026 | Neuroimaging Facility

Overview

We recently acquired a 3D EPI sequence developed by researchers at the German Center for Neurodegenerative Diseases (DZNE; Poser et al., 2010; Wang et al., 2022). This sequence allows for rapid whole-brain fMRI with high acceleration without the signal-to-noise hit experienced by high simultaneous multi-slice (AKA multi-band) values (Stirnberg et al., 2017). This post describes a use-case for this sequence in which we try to increase the temporal resolution of a whole-brain fMRI sequence to localize oscillatory activity. This type of sequence has also been shown to be effective for increasing spatial resolution (e.g., Narsude et al., 2016) as well. Please get in touch if you’re interested in exploring this sequence for your research.

The Pilot Study

In pilot testing, we were interested in leveraging the high temporal resolution of 3D-EPI to try to isolate oscillatory activation with the goal of isolating delta oscillations during memory consolidation paradigms. As a proof of concept, we scanned a flickering checkerboard paradigm using the 64-channel head coil and the following EPI parameters, which were optimized from Koch et al. (2025). TR = 410ms

  • TE = 30ms
  • Flip angle = 17°
  • Parallel Acquisition Technique = CAIPIRINHA
  • Parallel Reduction Factor: 6
  • Partial Fourier = 7/8
  • Matrix = 76 x 76
  • FOV = 228mm
  • Slice thickness = 3mm
  • 44 slices
  • 3x3x3mm spatial resolution

Temporal SNR looks good. In particular, we’re seeing much better SNR values in inferior temporal regions than we get with other fast functional sequences such as MREG, which uses a stack of spirals sequence (Assländer et al., 2013). Compare the 3D EPI sequence on the left with the MREG sequence (which has the same 3mm3 spatial resolution but a much faster temporal resolution of TR=200ms).

Task activation was about what you’d expect. The behavioral paradigm used alternating 20-second blocks of flickering checkerboard and fixation cross with music. The activation map below is the contrast of checkerboard > music and the time course is the “hottest” voxel in the visual cortex.

To examine oscillatory activity, I defined three masks. The first was the univariate activation map for significant voxels in either task condition with a voxel-wise p-value of 1×10-20 and spatial extent of k<40 contiguous voxels. Here, I was looking for voxels where the periodic activation varied at the same frequency as the task design, so about 0.025Hz. To look at higher frequency bands, I created a periodogram for each voxel using the AFNI program 3dPeriodogram. I then thresholded for the top ~5% of voxels centered on frequencies corresponding to respiration (0.303Hz) and cardiac cycle (1.108Hz). Below are the masks and corresponding time/frequency or spectrogram plots.

As you can see, task activation voxels cluster about where you’d expect them to in primary sensory areas (keep in mind the mask was defined for any active voxel regardless of sign so it also included primary auditory cortices, not pictured in the midsagittal slice above). The respiration and cardiac masks tended to identify major blood vessels and CSF spaces. I’m encouraged by seeing strong signals in the ~1Hz range that are physiologically sensible. I anticipate that it might be challenging to separate cardiac cycle from delta oscillations, but independently tracking heart rate would allow us to regress cardiac signals from the time course data.

What’s up with the peak at 0.606Hz?

As you might have noticed in the spectrogram plots, there is a band of increased power at the 0.606Hz range, corresponding to a cycle every 1.65 sec. This is a very strong signal in the frequency data. When we consider the frequency power spectrum for all voxels in the brain, we see a distinct peak in power at this frequency across many voxels:

If this were an EEG experiment, I’d say that looks a lot like line noise (which is when you see a peak in power at 60Hz corresponding to the oscillations in AC circuits in North America), so I suspected this might be some artifact in the signal. If it is artifact, we’d expect it to be uniformly present, or at least not follow any structural consistency in the brain. Looking at the periodogram map, however, reveals that this signal is strongest in CSF regions around the periphery of the brain, as well as in some gray matter structures in cortex. The figures below are the thresholded periodogram and power spectrum for the “hottest voxel”, i.e., the voxel at the center of the crosshairs with the highest power at this frequency.

 

To further investigate whether this is an artifact, I scanned a phantom with the same scan parameters as the pilot subject above. There was no evidence of a peak in frequency power around 0.6Hz.

So, I’m a bit stumped. I cannot think of any physiological processes that cycle at this frequency. I don’t think I’m aliasing something that cycles at a faster frequency, but that’s definitely a possibility. A further follow-up might be to independently measure cardiac and respiratory signals to see if there are faster signals that are getting aliased.

References

Assländer, J., Zahneisen, B., Hugger, T., Reisert, M., Lee, H.-L., LeVan, P., & Hennig, J. (2013). Single shot whole brain imaging using spherical stack of spirals trajectories. NeuroImage, 73, 59–70. https://doi.org/10.1016/j.neuroimage.2013.01.065

Koch, A., Stirnberg, R., Estrada, S., Zeng, W., Lohner, V., Shahid, M., Ehses, P., Pracht, E. D., Reuter, M., Stöcker, T., & Breteler, M. M. B. (2025). Versatile MRI acquisition and processing protocol for population-based neuroimaging. Nature Protocols, 20(5), 1223–1245. https://doi.org/10.1038/s41596-024-01085-w

Narsude, M., Gallichan, D., van der Zwaag, W., Gruetter, R., & Marques, J. P. (2016). Three-dimensional echo planar imaging with controlled aliasing: A sequence for high temporal resolution functional MRI. Magnetic Resonance in Medicine, 75(6), 2350–2361. https://doi.org/10.1002/mrm.25835

Poser, B. A., Koopmans, P. J., Witzel, T., Wald, L. L., & Barth, M. (2010). Three dimensional echo-planar imaging at 7 Tesla. NeuroImage, 51(1), 261–266. https://doi.org/10.1016/j.neuroimage.2010.01.108

Stirnberg, R., Huijbers, W., Brenner, D., Poser, B. A., Breteler, M., & Stöcker, T. (2017). Rapid whole-brain resting-state fMRI at 3 T: Efficiency-optimized three-dimensional EPI versus repetition time-matched simultaneous-multi-slice EPI. NeuroImage, 163, 81–92. https://doi.org/10.1016/j.neuroimage.2017.08.031

Wang, D., Ehses, P., Stöcker, T., & Stirnberg, R. (2022). Reproducibility of rapid multi-parameter mapping at 3T and 7T with highly segmented and accelerated 3D-EPI. Magnetic Resonance in Medicine, 88(5), 2217–2232. https://doi.org/10.1002/mrm.29383