Dataset Information

Automatic measurement plane placement for 4D Flow MRI of the great vessels using deep learning.

ABSTRACT:

Purpose

Despite 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.

Methods

Our 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.

Results

The 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).

Conclusion

Our 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.

SUBMITTER: Corrado PA

PROVIDER: S-EPMC8851604 | biostudies-literature | 2022 Jan

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Automatic measurement plane placement for 4D Flow MRI of the great vessels using deep learning.

Corrado Philip A PA Seiter Daniel P DP Wieben Oliver O

International journal of computer assisted radiology and surgery 20210817 1

<h4>Purpose</h4>Despite 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.<h4>Methods</h4>Our algorithm works by (1) subsampling the 3D image into 3D patches, (2) predicting the probability of each pat ...[more]

PMID: 34403045

Similar Datasets

Project description:ObjectivesTo evaluate the feasibility of Doppler-ultrasound (DUS)-gated 4D flow MRI of the fetal great thoracic vessels at 3T in a clinical setting.MethodsSixteen consecutive fetuses (range 30+4-38+5 weeks) with (n = 11) and without (n = 5) cardiovascular anomalies underwent 4D flow MRI of the great thoracic vessels at 3T. Direct fetal cardiac gating was obtained using a MR-compatible DUS device. 4D flow MRI-based visualisation and quantification of four target regions (ascending aorta (AAo), descending aorta (DAo), main pulmonary artery (MPA), and ductus arteriosus (DA)) were performed using dedicated software.ResultsFetal 4D flow MRI of the great thoracic vessels was successful in 12/16 fetuses (75%) by adopting clinical 4D flow MR protocols in combination with direct fetal cardiac DUS-gating. Four datasets were excluded due to artefacts by fetal movement or maternal breathing. 4D flow MRI-derived time-velocity curves revealed typical arterial blood flow patterns in the aorta. 4D flow quantification was achieved for the pre-defined target regions. Average velocity and flow volume were 21.1 ± 5.2 cm/s and 6.0 ± 3.1 mL/s in the AAo, 24.3 ± 6.7 cm/s and 8.4 ± 3.7 mL/s in the DAo, 21.9 ± 6.4 cm/s and 7.8 ± 4.2 mL/s in the MPA, and 23.4 ± 4.7 cm/s and 5.9 ± 3.6 mL/s in the DA, respectively.ConclusionsCombination of DUS-gating of the fetal heart and 4D flow MRI allows comprehensive visualisation and quantification of haemodynamics in the fetal great thoracic vessels. DUS-gated fetal 4D flow MRI may provide a new diagnostic approach for prenatal assessment of blood flow haemodynamics.Key points• Fetal cardiac Doppler-ultrasound (DUS) gating and 4D flow MRI can be successfully combined. • DUS-gated fetal 4D flow MRI allowed visualisation and evaluation of streamline directionality, illustration of blood flow variations, and pulsatile arterial waveforms in the target vessels. • 4D flow MRI-based visualisation and quantification of the fetal great thoracic vessels were successful and flow metrics agreed with echocardiographic reference values.

Project description:ObjectivesTo use 4D-flow MRI to describe systemic and non-systemic ventricular flow organisation and energy loss in patients with repaired d-transposition of the great arteries (d-TGA) and normal subjects.MethodsPathline tracking of ventricular volumes was performed using 4D-flow MRI data from a 1.5-T GE Discovery MR450 scanner. D-TGA patients following arterial switch (n = 17, mean age 14 ± 5 years) and atrial switch (n = 15, 35 ± 6 years) procedures were examined and compared with subjects with normal cardiac anatomy and ventricular function (n = 12, 12 ± 3 years). Pathlines were classified by their passage through the ventricles as direct flow, retained inflow, delayed ejection flow, and residual volume and visually and quantitatively assessed. Additionally, viscous energy losses (ELv) were calculated.ResultsIn normal subjects, the ventricular flow paths were well ordered following similar trajectories through the ventricles with very little mixing of flow components. The flow paths in all atrial and some arterial switch patients were more irregular with high mixing. Direct flow and delayed ejection flow were decreased in atrial switch patients' systemic ventricles with a corresponding increase in residual volume compared with normal subjects (p = 0.003 and p < 0.001 respectively) and arterial switch patients (p < 0.0001 and p < 0.001 respectively). In non-systemic ventricles, arterial switch patients had increased direct flow and decreased delayed ejection fractions compared to normal (p = 0.007 and p < 0.001 respectively) and atrial switch patients (p = 0.01 and p < 0.001 respectively). Regions of high levels of mixing of ventricular flow components showed elevated ELv.Conclusions4D-flow MRI pathline tracking reveals disordered ventricular flow patterns and associated ELv in d-TGA patients.Key points• 4D-flow MRI can be used to assess intraventricular flow dynamics in d-TGA patients. • d-TGA arterial switch patients mostly show intraventricular flow dynamics representative of normal subjects, while atrial switch patients show increased flow disorder and different proportions of intraventricular flow volumes. • Flow disruption and disorder increase viscous energy losses.

Project description:BackgroundThe beating heart is the generator of blood flow through the cardiovascular system. Within the heart's own chambers, normal complex blood flow patterns can be disturbed by diseases. Methods for the quantification of intra-cardiac blood flow, with its 4D (3D+time) nature, are lacking. We sought to develop and validate a novel semi-automatic analysis approach that integrates flow and morphological data.MethodIn six healthy subjects and three patients with dilated cardiomyopathy, three-directional, three-dimensional cine phase-contrast cardiovascular magnetic resonance (CMR) velocity data and balanced steady-state free-precession long- and short-axis images were acquired. The LV endocardium was segmented from the short-axis images at the times of isovolumetric contraction (IVC) and isovolumetric relaxation (IVR). At the time of IVC, pathlines were emitted from the IVC LV blood volume and traced forwards and backwards in time until IVR, thus including the entire cardiac cycle. The IVR volume was used to determine if and where the pathlines left the LV. This information was used to automatically separate the pathlines into four different components of flow: Direct Flow, Retained Inflow, Delayed Ejection Flow and Residual Volume. Blood volumes were calculated for every component by multiplying the number of pathlines with the blood volume represented by each pathline. The accuracy and inter- and intra-observer reproducibility of the approach were evaluated by analyzing volumes of LV inflow and outflow, the four flow components, and the end-diastolic volume.ResultsThe volume and distribution of the LV flow components were determined in all subjects. The calculated LV outflow volumes [ml] (67 +/- 13) appeared to fall in between those obtained by through-plane phase-contrast CMR (77 +/- 16) and Doppler ultrasound (58 +/- 10), respectively. Calculated volumes of LV inflow (68 +/- 11) and outflow (67 +/- 13) were well matched (NS). Low inter- and intra-observer variability for the assessment of the volumes of the flow components was obtained.ConclusionsThis semi-automatic analysis approach for the quantification of 4D blood flow resulted in accurate LV inflow and outflow volumes and a high reproducibility for the assessment of LV flow components.

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.

Dataset Information

Automatic measurement plane placement for 4D Flow MRI of the great vessels using deep learning.

Purpose

Methods

Results

Conclusion

Publications

Automatic measurement plane placement for 4D Flow MRI of the great vessels using deep learning.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets