Nyquist ghost correction of breast diffusion weighted imaging using referenceless methods

Abstract

Purpose Correction of Nyquist ghosts for single‐shot spin‐echo EPI using the standard 3‐line navigator often fails in breast DWI because of incomplete fat suppression, respiration, and greater B0 inhomogeneity. The purpose of this work is to compare the performance of the 3‐line navigator with 4 data‐driven methods termed “referenceless methods,” including 2 previously proposed in literature, 1 introduced in this work, and finally a combination of all 3, in breast DWI. Methods Breast DWI was acquired for 41 patients with SS SE‐EPI. Raw data was corrected offline with the standard 3‐line navigator and 4 referenceless methods, which modeled the ghost as a linear phase error and minimized 3 unique cost functions as well as the median solution of all 3. Ghost levels were evaluated based on the signal intensity in the background region, defined by a mask auto‐generated from a T1‐weighted anatomical image. Ghost intensity measurements were fit to a linear mixed model including ghost correction method and b‐value as covariates. Results All 4 referenceless methods outperformed the standard 3‐line navigator with statistical significance at all 4 b‐values tested (b = 0, 100, 600, and 800 s/mm2). Conclusions Referenceless methods provide a robust way to reduce Nyquist ghosts in breast DWI without the need for any additional calibration scan.

Comparison of simultaneous multi-slice single-shot DWI to readout-segmented DWI for evaluation of breast lesions at 3T MRI

“…Like in conventional ss-EPI sequences, these artifacts were more commonly found in the SMS-ss-EPI acquisitions than in the rs-EPI acquisitions, which would currently be a major drawback for the use of SMS-ss-EPI in practice. However, recent studies proposed alternative navigator approaches to reduce these artefacts in breast imaging [28],…”

Reconstruction for 7T high‐resolution whole‐brain diffusion MRI using two‐stage N/2 ghost correction and L1‐SPIRiT without single‐band reference

“…In stage 2, residual ghosts were further removed using the reference‐free entropy method 30 . The phase error $$$ \theta $$$ in stage 2 was modeled as a first‐order polynomial 31 : $$$$ \theta =\pi \kappa x/{N}_x+{\theta}_0, $$$$ where $$$ {\theta}_0 $$$ = spatially constant phase shift, $$$ \kappa $$$ = phase shift slope (with $$$ \kappa =1 $$$ referring to a one‐pixel shift between positive and negative lines), and $$$ {N}_x $$$ = number of read‐out points. The cost function $$$ {f}_{entropy}\left(\kappa, {\theta}_0\right) $$$ of the image entropy was formulated and minimized to estimate the $$$ \kappa $$$ and $$$ {\theta}_0 $$$ solutions.…”

A simultaneous multi‐slice T2 mapping framework based on overlapping‐echo detachment planar imaging and deep learning reconstruction

“…For the acquired SMS‐OLED data, we used a phase search algorithm 45 to reduce the Nyquist ghost artifacts. The acquired data were processed the same as the training samples with MATLAB (Mathworks), and then sent to the trained network for T 2 map reconstruction.…”