Lauri Raitamaa1, Vesa Korhonen1, Niko Huotari1, and Vesa Kiviniemi1
1Research Unit of Medical Imaging Physics and Technology (MIPT), University of Oulu, Oulu, Finland
Synopsis
Ultra-fast 10 Hz whole brain fMRI with magnetic resonance encephalography (MREG) enables separation of cardiovascular waves from brain pulsation without temporal aliasing. This enables analysis of spectral characteristics of respiration and cardiovascular pulsations by extending ALFF and fALFF methods to cardiovascular and respiration frequencies . Modulation of the cardiovascular pulsation has been detected in ECG and SpO2 and in this study show the modulation of the cardiovascular brain pulsation in 3D brain in healthy controls and how age and blood pressure affects the modulation.
Introduction
In
this study we investigated the power distributions of each brain
pulsation mechanism. The spatial distribution of the cardiovascular and
respiratory were mapped using amplitude spectral map using established
ALFF and fALFF techniques extended to cardiovascular and respiratory frequencies.1
Due to the fast scanning method, we were able to differentiate of
cardiorespiratory brain pulsations known to exist in CSF pressure and
flow. Also,
we investigate how respiration return oscillations modulate
cardiovascular pulses of the brain. This modulation has been shown to
occur on three different forms: baseline, amplitude and frequency
modulation, the last one known better as the heart rate variability
[ref]. Mostly these modulations have been detected in electrocardiogram
(ECG) and peripheral O2 saturation (SpO2) curves, but also in fast MRI
techniques.3,4 Finally, we did linear regression analysis to show effects of blood pressure and age on the brain pulsations.Methods
A total of 48/52 healthy subjects (age: 41.1 ± 17.2, 29 females) were scanned using 3T Skyra MREG sequence (TR=100 ms, TE=36 ms, flip angle=25°, 3D matrix=643, FOV=192 mm) for 5 to 10 min. First 6 s was discarded, and all data sets were set to 4.8 min (2880 volumes). Data was despiked using AFNI 3dDepsiked with –NEW option and processed with typical FSL pre-processing steps (Fig 1 a).5 During image preprocessing four subject were excluded because of defective images. The remaining 48 subjects were used in this study.
Preprocessed MREG data frequency bands were inspected by using AFNI 3dPeriodogram to detect cardiac and respiratory peaks (Fig 1 d).
Location of sidebands were calculated by adding/subtracting respiration
peak frequency from cardiac peak frequency. LF was from 0.01 Hz to 1.0
Hz. Respiration and cardiac bands were set to 0.1 Hz band centered
around the peaks. Locations of cardiac side peaks were calculated using
heterodyne principle: multiplied sinusoidal waveforms maybe written as
the sum and the difference of the multiplied frequencies.6 In our case frequencies are f(card) +/- f(respiration). Frequency bands of side peaks were also set to 0.1 Hz centered around peaks.
ALFF
was computed from 3dperiodogram calculating square root and taking sum
over frequency bands of interest: VLF, respiration, cardiac and side
peaks (Fig 2). fALFF was calculated taking sum of frequency band of interest and dividing it by sum of all frequency bands of interest (Fig 3).
Regression
analysis was performed for 43 subjects as 5 subjects didn’t have blood
pressure information. Analysis were done for both ALFF (Fig 4) and fALFF (Fig
5). Voxel-wise comparisons were performed one-sample t-test using
non-parametric permutation tests (10000 permutation) implemented vlisa_onesample in LISA.7Results
Classical
ALFF in very low frequencies (VLF < 0.1 Hz) occurs exactly in the
previously described cortical areas dominantly in default mode and
visual areas and frontal cortex.1
The cardiac ALFF extends
significantly in the periarterial areas along anterior, posterior and
medial cerebral arteries and immediately adjacent perivascular spaces.
The strong cardiac pulsation extends in the midline and lateral, 3rd and 4th
CSF ventricles. Basal CSF spaces around brain stem, pons and basal
cisternae are also strongly pulsating with cardiac frequency.
Importantly cardiovascular pulsation also clearly present in posterior
parts of sagittal sinus.
The
respiratory ALFF pulsation is relatively strong over the whole brain
tissue. Only periarterial structures and ventricles have relatively
reduced respiratory power. Also, the cortical structures that have high
VLF amplitude have relatively strong respiratory pulsations in addition
to VLF. The respiratory power also extends to white matter quite
uniformly.
The cardiac pulsation is modulated strongly by respiration in centrobasal CSF and periartial spaces. Modulation of cardiac brain pulsation by respiration occurs mostly in lateral and 3rd ventricle and periarterial spaces in areas where the cardiac pulsation frequency is also strong. In frontolateral
areas the modulation also enters in ventromedial prefrontal parts of
default mode network, frontal and sensorimotor cortices bilaterally, in
posterior fossa cerebellar vermis, medulla and especially spinal cord.
Pulse
pressure correlates more strongly with side peaks than with
fundamental cardiac frequency. When comparing proportion of side peaks
power to fundamental frequency with pulse pressure areas that show
positive correlation are in arterial spaces.Discussion
The
critically sampled data allows aliasing free spectral analytics and the
results indicate that the VLF distribution is not significantly
affected by cardiorespiratory aliasing.8 The
intracranial space and they induce CSF and brain tissue pulsations. The
relative power of venous counter pulsations and CSF flow pulses into
the brain tissue is to be determined. The strength of the modulation is
governed by the elastic properties of the pulsating substance: the CSF
pulses and modulations are stronger due to smaller resistance than in
the brain tissue. Furthermore, the pulse pressure (I.e. systolic –
diastolic BP) of the arteries governs the cardiovascular pulsation
amplitude that then becomes modulated by the respiration. Conclusion
Respiratory
modulate cardiac brain pulsations in CSF spaces, frontal brain cortex
and in cerebellar/medulla areas. Physiologically the respiration
controls venous return and CSF pulses, which then modulate arterial
pulses in the brain vault. As the cardiac brain pulsation has been shown
to be a main driver of the glymphatic brain clearance, the results add
further proof that also respiration affects brain clearance.Acknowledgements
No acknowledgement found.References
1.
Zou, Qi-Hong, et al. "An improved approach to detection of amplitude of
low-frequency fluctuation (ALFF) for resting-state fMRI: fractional
ALFF." Journal of neuroscience methods 172.1 (2008): 137-141.
2. Saykrs, B. McA. "Analysis of heart rate variability." Ergonomics 16.1 (1973): 17-32.
3. Charlton, Peter H., et al. "Extraction of respiratory signals from the electrocardiogram and photoplethysmogram: technical and physiological determinants." Physiological measurement 38.5 (2017): 669.
4.
Abi-Abdallah, Dima, et al. "Cardiac and respiratory MRI gating using
combined wavelet sub-band decomposition and adaptive filtering." Annals of biomedical engineering 35.5 (2007): 733-743.
5. Jenkinson, Mark, et al. "Fsl." Neuroimage 62.2 (2012): 782-790.
6. Graf, R. (1999), Modern dictionary of electronics, 7th Ed. USA: Newnes. p. 344. ISBN 0-7506-9866-7
7.
Lohmann, Gabriele, et al. "LIPSIA—a new software system for the
evaluation of functional magnetic resonance images of the human brain." Computerized medical imaging and graphics 25.6 (2001): 449-457.
8. Huotari, Niko, et al. "Sampling rate effects on resting state fMRI metrics." Frontiers in neuroscience 13 (2019): 279.