Fig 1.
The model architecture of Image2Flow.
The hybrid image and graph convolutional neural network architecture of Image2Flow. It takes as input a 3D cardiac MRI and a template volume mesh of a pulmonary artery. It outputs the patient-specific pulmonary artery mesh with associated pressure and flow at each node.
Fig 2.
A schematic of the steps involved in point-point correspondent volume mesh generation.
(A) raw patient-specific volume mesh created from manual segmentation, (B) initial template volume mesh creation, (C) point-point correspondent volume mesh generation by transforming the initial template, (D) final template volume mesh creation by averaging the point-point correspondent meshes of the training data. Red indicates non-corresponding meshes and blue represents corresponding meshes. The wireframe rendering denotes surface meshes, while the solid rendering denotes volume meshes.
Fig 3.
The best, median and worst Image2Flow segmentations compared to ‘MeshDeformNet’ and a 3D UNet.
Table 1.
Segmentation metrics evaluating Image2Flow, ‘MeshDeformNet’ and the 3D UNet compared to the ground truth.
Fig 4.
(A) MNAES values of the Image2Flow predictions compared to the ground truth on the test set (n = 15) for pressure and velocity. (B) Difference in MNAES values between CFDI2F and CFDDL-seg.
Fig 5.
The best, median and worst blood pressure, and velocity predictions of Image2Flow by MNAEs.
The size and positioning between the true and predicted meshes are to scale.
Fig 6.
The distribution of node-wise error (MNAEN) of the test set (n = 15) projected onto the template pulmonary artery volume-mesh.
(A) Distribution of error across the cross-section of the pulmonary artery, (B) Bland-Altman analysis of the pressure and velocity magnitude errors.
Fig 7.
MNAES values of the Image2Flow pressure and velocity magnitude predictions for different inlet velocities (0.1, 0.2, and 0.3m/s).
Fig 8.
The best, median and worst blood pressure predictions of Image2Flow with varying input inlet velocity (0.1–0.3m/s).
The size and positioning between the true and predicted meshes are to scale.
Fig 9.
The best, median and worst blood velocity predictions of Image2Flow with varying input inlet velocity (0.1–0.3m/s).
The size and positioning between the true and predicted meshes are to scale.
Fig 10.
Comparison between Image2Flow CFD prediction and conventional CFD on the same geometry.