Measuring Arterial Pulsatility With Dynamic Inflow Magnitude Contrast

Figures

• 

  FIGURE 1. (A) An example of the slice placement at the M1 segment of the MCA for the “Hypercapnia Challenge-Experiment”. (B) An example dynamic inflow magnitude contrast (DIMAC) image, including MCA mask. (C) Slice placement for the subject in the “Thigh Cuff Release-Experiment” placed to include best perpendicular placement of bilateral ICAs and VAs. (D) Subject’s DIMAC EPI image and artery masks.

• 

  FIGURE 2. (A) The average amount of variance explained in the beat-to-beat fit as a function of the number of Fourier terms used. (B) An example showing detected diastolic troughs and the beat-to-beat pulsatile fits taken from a representative subject.

• 

  FIGURE 3. (A) Relative sensitivity of spoiled gradient-recalled-echo (GRE) signal to pulsatile cerebral blood flow velocity (CBFV) compared with pulsatile cerebral blood volume (CBV), where each line represents a different flip angle. The annotated arrow represents how the GRE signal moves into the DIMAC regime of high sensitivity to pulsatile CBFV over pulsatile CBV as TR decreases and flip angle increases. (B) Scatterplot of simulated spoiled GRE signal in the DIMAC regime of low TR (15 ms) and high flip angle (90°) as a function of CBFV. Note that the spread of data points reflects signal variance due to pulsatile CBV, which is small compared with variance due to CBFV within the normal physiological range. In every case the signal plateaus at the critical velocity 66 cm s–1, i.e., when flow velocity increases to the point when spins only experience a single RF pulse and the signal becomes sensitive to CBV alone.

• 

  FIGURE 4. (A) Simulated DIMAC signal over four beats. Each panel represents a different partial volume (row) and SNR condition (column). Note the different Y-axis scale for each partial volume. Thus, for a given SNR, increased partial volume of arterial blood increases signal magnitude and yields a stronger pulsatile signal. (B) Each panel shows the amount of variance explained in the true CBFV waveform by the simulated DIMAC signal as a function of number of beats included in the average. Naturally, as the number of beats included in the average increases more noise is averaged out, and thus the R2 increases. However, this also shows that when both partial volume and SNR are sufficiently high, individual beats can be resolved with high fidelity (i.e., a high R2 with respect to the true CBFV waveform).

• 

  FIGURE 5. (A) Example (subject 1) time series and their power spectra for each TR condition. Both the time series and power spectrum make it clear that there is a strong degree of pulsatility for the TR15 conditions, which falls away with higher TR values as predicted by the theory. (B) The pulsatility index for each TR and HC condition, which quantitatively confirms that which is observed in panel (A).

• 

  FIGURE 6. (A) First 15 s of time series for each subject for HC-challenge data (TR15 and HC0 condition), with beat-to-beat fit overlaid. (B) The mean pulsatile waveform (±SD shaded area) for each subject (TR15 and HC0 condition). (C) The across subject mean pulsatile waveforms (±within subject error shaded area) for each HC condition of the HC-challenge. Qualitatively one can see the pulsatile waveform shape is modulated by hypercapnia.

• 

  FIGURE 7. (A) First 15 s of the TCR-challenge time series taken from the left ICA ROI. (B) The systolic (blue) and diastolic (red) time series (i.e., signal envelope) averaged across the five repeats, for both ICA and VA bilaterally. The onset of the thigh-cuff release is highlighted with a dotted grey line. Both left and right internal carotid arteries (ICA) show clear time locked responses to the TCR-challenge in both systolic and diastolic time series. The same responses are far less evident in the VA time series, particularly in the right side.