Smooth Total variation Regularization for Interference Detection and Elimination (STRIDE) for MRI

TL;DR

STRIDE addresses dynamic electromagnetic interference (EMI) in MRI by integrating total-variation (TV) regularization into an image-domain EMI removal framework. It shifts EMI removal from a purely transfer-function approach to a compressed-sensing-inspired formulation that enforces columnwise smoothness, solved in closed form for the EMI-subspace coefficients. In phantom and in-vivo tests on a 0.5T scanner, STRIDE delivers higher temporal SNR, greater EMI removal, and lower RMSE than the standard EDITER implementations, especially for time-varying, narrow-band noise, while remaining compatible with various sampling patterns. The method is deterministic and transparent, emphasizing the role of EMI-sensor SNR and avoiding data-driven training, which enhances robustness in practical, non-ideal MRI environments.

Abstract

MRI is increasingly desired to function near electronic devices that emit potentially dynamic electromagnetic interference (EMI). To accommodate for this, we propose the STRIDE method, which improves on previous external-sensor-based EMI removal methods by exploiting inherent MR image smoothness in its total variation. STRIDE measures data from both EMI detectors and primary MR imaging coils, transforms this data into the image domain, and for each column of the resulting image array, combines and subtracts data from the EMI detectors in a way that optimizes for total-variation smoothness. Performance was tested on phantom and in-vivo datasets with a 0.5T scanner. STRIDE resulted in visually better EMI removal, higher temporal SNR, larger EMI removal percentage, and lower RMSE than standard implementations. STRIDE is a robust technique that leverages inherent MR image properties to provide improved EMI removal performance over standard algorithms, particularly for time-varying noise sources.

Figures (10)

• Figure 1: Visual comparison of voxel-wise SNR for each EMI removal method for (A) resolution dots and (B) contrast discs in a standard ACR phantom. A higher SNR is better. STRIDE is the proposed method, whereas EDITER A and B are two implementations of the standard EMI removal method EDITER. EDITER A uses one phase encode line per temporal group, while EDITER B dynamically adjusts the number of phase encode lines per temporal group as outlined in the EDITER publication EDITER. The resolution dots in (A) show overall similar SNR between methods, with a small dark spot in the upper-middle third section for STRIDE. In areas corrupted by narrow-band EMI, STRIDE shows consistently higher SNR, indicating superior EMI removal in those regions. The contrast discs in (B) show overall superior SNR in STRIDE compared to EDITER A and EDITER B, again with particular superior performance in areas of narrow-band EMI corruption.
• Figure 2: Graphical comparison of the mean SNR across non-background voxels in Figure \ref{['fig:figure1']} for each EMI removal method for(left) the resolution dots and (right) contrast discs in a standard ACR phantom, as well as the statistical significance of the difference in means as determined by a welch's t-test. $*$ indicates $p < 0.05$, $**$ indicates $p < 0.01$, and $***$ indicates $p < 0.001$, and ns indicates no significance. A higher mean SNR is better. Overall, there were more cases for which STRIDE significantly outperformed both EDITER A and EDITER B, indicating a statistically significant improvement over the two standard implementations.
• Figure 3: Graphical comparison of the mean EMI removal percentage across non-background voxels for each EMI removal method for(left) the resolution dots and (right) contrast discs in a standard ACR phantom, as well as the statistical significance of the difference in means as determined by a Welch's t-test. $*$ indicates $p$ <$0.05$, $**$ indicates $p < 0.01$, and $***$ indicates $p < 0.001$, and ns indicates no significance. A higher mean EMI removal percentage is better. Overall, there were more cases for which STRIDE significantly outperformed both EDITER A and EDITER B, indicating a statistically significant improvement over the two standard implementations.
• Figure 4: Visual comparison of voxel-wise RMSE for each EMI removal method for (A) resolution dots and (B) contrast discs in a standard ACR phantom. A lower RMSE is better. Overall, STRIDE and the two EDITER implementations have visually similar RMSE. In areas of narrow-band EMI contaimination, however, STRIDE shows visually lower RMSE than either EDITER A or EDITER B. The improvement in RMSE is reflected in Table \ref{['tab:RMSEcomparison']}, which shows RMSE over the entire imaging volume.
• Figure 5: Visual comparison of (first row) baseline images with no EMI corruption, (second row) EMI corrupted images, (third row) EMI corrected images using STRIDE, (fourth row) EMI corrected images using EDITER A, and (fifth row) EMI corrected images using EDITER B for the contrast discs in the small ACR phantom. In every case, EDITER A results in horizontal streaking artifacts across the phantom which are not seen in the STRIDE reconstructions. The performance of STRIDE and EDITER B are similar in the cases of Sweep and Vital Sign Monitor noise. In the cases of Square wave and White noise, STRIDE better removes the EMI than EDITER B.