Dataset Information

Deep learning-based velocity antialiasing of 4D-flow MRI.

ABSTRACT:

Purpose

To develop a convolutional neural network (CNN) for the robust and fast correction of velocity aliasing in 4D-flow MRI.

Methods

This study included 667 adult subjects with aortic 4D-flow MRI data with existing velocity aliasing (n = 362) and no velocity aliasing (n = 305). Additionally, 10 controls received back-to-back 4D-flow scans with systemically varied velocity-encoding sensitivity (vencs) at 60, 100, and 175 cm/s. The no-aliasing data sets were used to simulate velocity aliasing by reducing the venc to 40%-70% of the original, alongside a ground truth locating all aliased voxels (153 training, 152 testing). The 152 simulated and 362 existing aliasing data sets were used for testing and compared with a conventional velocity antialiasing algorithm. Dice scores were calculated to quantify CNN performance. For controls, the venc 175-cm/s scans were used as the ground truth and compared with the CNN-corrected venc 60 and 100 cm/s data sets RESULTS: The CNN required 176 ± 30 s to perform compared with 162 ± 14 s for the conventional algorithm. The CNN showed excellent performance for the simulated data compared with the conventional algorithm (median range of Dice scores CNN: [0.89-0.99], conventional algorithm: [0.84-0.94], p < 0.001, across all simulated vencs) and detected more aliased voxels in existing velocity aliasing data sets (median detected CNN: 159 voxels [31-605], conventional algorithm: 65 [7-417], p < 0.001). For controls, the CNN showed Dice scores of 0.98 [0.95-0.99] and 0.96 [0.87-0.99] for venc = 60 cm/s and 100 cm/s, respectively, while flow comparisons showed moderate-excellent agreement.

Conclusion

Deep learning enabled fast and robust velocity anti-aliasing in 4D-flow MRI.

SUBMITTER: Berhane H

PROVIDER: S-EPMC9050855 | biostudies-literature | 2022 Jul

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Deep learning-based velocity antialiasing of 4D-flow MRI.

Berhane Haben H Scott Michael B MB Barker Alex J AJ McCarthy Patrick P Avery Ryan R Allen Brad B Malaisrie Chris C Robinson Joshua D JD Rigsby Cynthia K CK Markl Michael M

Magnetic resonance in medicine 20220405 1

<h4>Purpose</h4>To develop a convolutional neural network (CNN) for the robust and fast correction of velocity aliasing in 4D-flow MRI.<h4>Methods</h4>This study included 667 adult subjects with aortic 4D-flow MRI data with existing velocity aliasing (n = 362) and no velocity aliasing (n = 305). Additionally, 10 controls received back-to-back 4D-flow scans with systemically varied velocity-encoding sensitivity (vencs) at 60, 100, and 175 cm/s. The no-aliasing data sets were used to simulate velo ...[more]

PMID: 35381116

Similar Datasets

Project description:PurposeTo generate fully automated and fast 4D-flow MRI-based 3D segmentations of the aorta using deep learning for reproducible quantification of aortic flow, peak velocity, and dimensions.MethodsA total of 1018 subjects with aortic 4D-flow MRI (528 with bicuspid aortic valve, 376 with tricuspid aortic valve and aortic dilation, 114 healthy controls) comprised the data set. A convolutional neural network was trained to generate 3D aortic segmentations from 4D-flow data. Manual segmentations served as the ground truth (N = 499 training, N = 101 validation, N = 418 testing). Dice scores, Hausdorff distance, and average symmetrical surface distance were calculated to assess performance. Aortic flow, peak velocity, and lumen dimensions were quantified at the ascending, arch, and descending aorta and compared using Bland-Altman analysis. Interobserver variability of manual analysis was assessed on a subset of 40.ResultsConvolutional neural network segmentation required 0.438 ± 0.355 seconds versus 630 ± 254 seconds for manual analysis and demonstrated excellent performance with a median Dice score of 0.951 (0.930-0.966), Hausdorff distance of 2.80 (2.13-4.35), and average symmetrical surface distance of 0.176 (0.119-0.290). Excellent agreement was found for flow, peak velocity, and dimensions with low bias and limits of agreement less than 10% difference versus manual analysis. For aortic volume, limits of agreement were moderate within 16.3%. Interobserver variability (median Dice score: 0.950; Hausdorff distance: 2.45; and average symmetrical surface distance: 0.145) and convolutional neural network-based analysis (median Dice score: 0.953-0.959; Hausdorff distance: 2.24-2.91; and average symmetrical surface distance: 0.145-1.98 to observers) demonstrated similar reproducibility.ConclusionsDeep learning enabled fast and automated 3D aortic segmentation from 4D-flow MRI, demonstrating its potential for efficient clinical workflows. Future studies should investigate its utility for other vasculature and multivendor applications.

Project description:PurposeDespite the great potential and flexibility of 4D flow MRI for hemodynamic analysis, a major limitation is the need for time-consuming and user-dependent post-processing. We propose a fast four-step algorithm for rapid, robust, and repeatable flow measurements in the great vessels based on automatic placement of measurement planes and vessel segmentation.MethodsOur algorithm works by (1) subsampling the 3D image into 3D patches, (2) predicting the probability of each patch containing individual vessels and location/orientation of the vessel within the patch via a convolutional neural network, (3) selecting the predicted planes with highest probabilities for each vessel, and (4) shifting the plane centers to the maximum velocity within each plane. The method was trained on 283 scans and evaluated on 40 unseen scans by comparing algorithm-derived processing times, plane locations, and flow measurements to those of two manual observers (graduate students) using t-tests, Pearson correlation, and Bland-Altman analysis.ResultsThe average processing time for the algorithm (18 s) was shorter than observer 1 (362 s; P < 0.001) and observer 2 (317 s; P < 0.001). The distance between planes placed by the algorithm and those placed by manual observers was slightly greater (O1 vs. algorithm: 9.0 mm, O2 vs. algorithm: 10.3 mm) than the distance between planes placed by the two manual observers (8.3 mm). The correlation between flow values for planes placed by the algorithm and those placed by manual observers was slightly lower (O1 vs. algorithm: R = 0.68, O2 vs. algorithm: R = 0.72) than the flow correlation between the two manual observers (R = 0.81).ConclusionOur method is a feasible and accurate approach for fast, reproducible, and automated flow measurement and visualization in 4D flow MRI of the great vessels, with similar variability compared to a manual annotator as the variability between two manual observers. This approach could be applied in other anatomical regions.

Project description:BACKGROUND:Complex and reverse flow in the aorta has been implicated in aneurysm development and stroke via retrograde embolization. PURPOSE:To evaluate global and regional differences between standard 2D plane-based and volumetric voxel-based quantification of regional forward/reverse flow, and reverse flow fraction (RFF) in the aorta. STUDY TYPE:Retrospective. SUBJECTS:In all, 35 subjects: 10 healthy controls (age: 57?±?7 years, nine male), nine patients without aortic valve regurgitation (AR) (age: 63?±?10 years, seven male), six patients with mild AR (age: 66?±?6 years, five male), and 10 with moderate or severe AR (age: 60?±?16 years, eight male). FIELD STRENGTH/SEQUENCE:4D flow MRI (3T and 1.5T) was employed to acquire 3D blood flow velocities with entire thoracic aorta in all subjects. ASSESSMENT:Data analysis included standard 2D plane-based quantification of forward/reverse flow, and RFF-plane. In addition, a new semiautomatic workflow based on 3D segmentation and extraction of an aorta centerline was developed for voxel-by-voxel visualization (forward/reverse flow and RFF-voxel maps) and quantification of regional voxel-by-voxel forward/reverse flow in the entire thoracic aorta. STATISTICAL TESTS:Kruskal-Wallis tests were performed to test for differences between groups. A two-sample t-test or Wilcoxon rank sum test was used to compare voxel-based and plane-based results. RESULTS:Semiautomatic plane-based analysis showed excellent agreement with standard manual plane-based analysis for net flow and RFF-plane (RFF-plane: y?=?0.99x-0.0, net flow: y?=?1.00x-0.21, R > 0.99, P < 0.0001). Voxel-by-voxel maps demonstrated marked regional flow reversal in the ascending aorta in all patients and RFF-voxel was significantly increased (P < 0.001) compared to RFF-plane for all four groups, with the most pronounced differences for mild AR (18.0?±?15.2% vs. 4.7?±?5.4%). Voxel-based flow and RFF-voxel along the aorta showed areas with marked regional flow reversal (eg, vortex flow) compared to plane-based analysis. DATA CONCLUSION:Voxel-based analysis demonstrated regional flow reversal that was not detected by plane-based analysis. LEVEL OF EVIDENCE:3 Technical Efficacy: Stage 1 J. Magn. Reson. Imaging 2018;47:1276-1286.

Dataset Information

Deep learning-based velocity antialiasing of 4D-flow MRI.

Purpose

Methods

Conclusion

Publications

Deep learning-based velocity antialiasing of 4D-flow MRI.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets