A characterization of cardiac-induced noise in R₂* maps of the brain

Abstract

Purpose Cardiac pulsation increases the noise level in brain maps of the MRI parameter R2*. Cardiac-induced noise is challenging to mitigate during the acquisition of R2* mapping data because the characteristics of this noise remain largely unknown. In this work, we characterize cardiac-induced noise in brain maps of the MRI parameter R2*. Methods We introduce a sampling strategy that enables the acquisition of multi-echo data in several intervals of the cardiac cycle. From this data, we estimate the variability of brain R2* maps due to cardiac-induced fluctuations and identify the most sensitive area of k-space. From these characteristics, we derive a novel sampling strategy that successfully mitigates cardiac-induced noise in R2* maps of the brain. Results In inferior brain regions, cardiac pulsation accounts for R2* variations of up to 3s-1 across the cardiac cycle, i.e. ~35% of the overall variability. Cardiac-induced fluctuations occur throughout the cardiac cycle and exhibit a reduced intensity during the first quarter of the cycle after the detection of the systole at the finger. 50-60% of the overall cardiac-induced noise is contained near the k-space centre (k < 0.074 mm-1), corresponding to 22% of voxels at 4mm resolution. The proposed cardiac noise mitigation strategy reduces the variability of R2* estimates across repetitions by 11% in the brainstem and 6% across the whole brain. Conclusion We provide characteristics of cardiac-induced noise in brain maps of the MRI parameter R2* that constitute a basis for the design of mitigation strategies during data acquisition.

A characterization of cardiac‐induced noise in R2* maps of the brain

“…Initial results indicate that these strategies provide a much improved mitigation of cardiac-induced noise. 64 The spatial resolution in the assessment of the proposed mitigation strategy was higher than the 5D datasets used for the characterization of cardiac-induced noise, leading to a higher contribution of thermal noise to the overall variability of R 2 * maps. Furthermore, cardiac gating was set to act on a restricted area of k-space that contains ∼40% of the total cardiac-induced noise in brain voxels and ∼25% in non-brain voxels, leaving a large part of cardiac-induced noise in other k-space regions intact.…”

“…We are currently investigating alternative strategies that involve the synchronization of data acquisition with cardiac pulsation for Cartesian trajectories, as well as alternative sampling trajectories. Initial results indicate that these strategies provide a much improved mitigation of cardiac‐induced noise 64 . The spatial resolution in the assessment of the proposed mitigation strategy was higher than the 5D datasets used for the characterization of cardiac‐induced noise, leading to a higher contribution of thermal noise to the overall variability of R 2 * maps.…”

Tensor denoising of quantitative multi‐parameter mapping

“…As a surrogate for the noise level on the R 2 * estimates, we evaluated the RMS error (RMSE) between data points (PDw images) and the R 2 * relaxation model 31 . Fit residuals were normalized with the extrapolated value at TE = 0 and averaged across white matter (WM) and gray matter (GM).…”

In-vivo characterization of magnetic inclusions in the subcortex from non-exponential transverse relaxation decay

“…The cardiac pulsation of the participants was recorded using a finger pulse oximeter. To minimize the effect of cardiac-induced noise on the images, data acquisition was suspended during the systolic period of the cardiac cycle, taken to last for a duration of 300 ms (Raynaud et al, 2023). For a heart rate of 80 beats per minute, this resulted in an increase in scan time by approximately 40%.…”

“…The lack of evidence of non-exponential signal decay in subcortical regions has been attributed to the short timescale of the transition between the Gaussian and exponential behaviours, below the range of achievable echo times (Yablonskiy et al, 2021). Here, we combined a dense sampling of the decay curve with acquisition strategies that reduce the noise level in the data (Castella et al, 2018;Raynaud et al, 2023) to collect reliable evidence of non-exponential signal decay. Transverse relaxation decay exhibits a Gaussian behaviour at short echo times (TE≲5ms) with a transition towards exponential decay (Fig.…”