Accelerated 4D‐flow MRI with 3‐point encoding enabled by machine learning

Abstract

Purpose To investigate the acceleration of 4D‐flow MRI using a convolutional neural network (CNN) that produces three directional velocities from three flow encodings, without requiring a fourth reference scan measuring background phase. Methods A fully 3D CNN using a U‐net architecture was trained in a block‐wise fashion to take complex images from three flow encodings and to produce three real‐valued images for each velocity component. Using neurovascular 4D‐flow scans (n = 144), the CNN was trained to predict velocities computed from four flow encodings by standard reconstruction including correction for residual background phase offsets. Methods to optimize loss functions were investigated, including magnitude, complex difference, and uniform velocity weightings. Subsequently, 3‐point encoding was evaluated using cross validation of pixelwise correlation, flow measurements in major arteries, and in experiments with data at differing acceleration rates than the training data. Results The CNN‐produced 3‐point velocities showed excellent agreements with the 4‐point velocities, both qualitatively in velocity images, in flow rate measures, and quantitatively in regression analysis (slope = 0.96, R2 = 0.992). Optimizing the training to focus on vessel velocities rather than the global velocity error and improved the correlation of velocity within vessels themselves. The lowest error was observed when the loss function used uniform velocity weighting, in which the magnitude‐weighted inverse of the velocity frequency uniformly distributed weighting across all velocity ranges. When applied to highly accelerated data, the 3‐point network maintained a high correlation with ground truth data and demonstrated a denoising effect. Conclusion The 4D‐flow MRI can be accelerated using machine learning requiring only three flow encodings to produce three‐directional velocity maps with small errors.

Deep Learning-Based Reconstruction for Cardiac MRI: A Review

“…Kim et al demonstrated a DL-based reconstruction for the 4D-Flow that produced three directional velocities from only three acquired flow encodings, without requiring the conventional fourth reference scan to measure the background phase [ 102 ]. They used the complex images from four velocity encodings (flow-compensated reference plus x, y, and z velocities) of 144 neurovascular patient datasets to train a 3D U-Net to predict the x, y, and z velocities.…”

Four‐dimensional flow MRI for quantitative assessment of cerebrospinal fluid dynamics: Status and opportunities

“…39,115 Because CSF net flow rates are typically small, uncorrected phase offset errors may lead to false interpretation of the direction of CSF flow in compartments that are central to understanding normal and pathological CSF physiology (e.g., the CA 39 ). More recently, machine learning implementations have demonstrated the capability to remove background field offsets in neurovascular 4D flow MRI, 81 something that could be extended to 4D CSF flow.…”

“…78 Other studies have demonstrated the feasibility of accelerating neurovascular 4D flow acquisitions by 25% using machine learning and a convolutional neural network (CNN) that produces three directional velocities from three flow encodings, without requiring a fourth reference scan measuring background phase. 81 The authors used neurovascular 4D flow scans to train a CNN using supervised learning. The produced three-point velocities showed excellent agreement with the velocities derived from standard reconstruction using all four flow encodings with $1.5% errors in the velocity values, and correlations as high as 0.992.…”

“…Neurovascular 4D flow requires high spatiotemporal resolution to detect blood flow in smaller structures 76,77 . Advances in neurovascular 4D flow acquisition, 76–79 reconstruction, 80,81 and the development of postprocessing 82–85 tools have turned 4D flow MRI into a versatile technique for the assessment of CBF dynamics. This includes derivation of a large number of hemodynamic biomarkers (reviewed elsewhere) 67,68,80,84,86–92 .…”

“…More recently, ~35% scan time reductions in neurovascular dual‐V enc 4D flow scans were demonstrated by undersampling of the high V enc acquisition while preserving the dynamic range and VNRs of dual‐V enc 78 . Other studies have demonstrated the feasibility of accelerating neurovascular 4D flow acquisitions by 25% using machine learning and a convolutional neural network (CNN) that produces three directional velocities from three flow encodings, without requiring a fourth reference scan measuring background phase 81 . The authors used neurovascular 4D flow scans to train a CNN using supervised learning.…”