Rapid and automatic hemodynamic assessment: integration of deep learning-based image segmentation, vessel...

Currently, vascular hemodynamic analyses are typically conducted using commercial software. This process usually involves reconstructing the three-dimensional (3D) geometry of blood vessels, generating a computational mesh, and performing a computational fluid dynamics (CFD) analysis. It requires skilled medical personnel to manually process medical images, which is time consuming and prone to errors. This study aimed to develop a deep learning-based method to quickly and accurately extract vascular hemodynamic feature data to address these issues. This was accomplished by automating the processes of computed tomography (CT) image segmentation, vessel reconstruction, and CFD prediction.

An improved convolutional neural network (CNN) was developed to automatically segment preprocessed vascular CT images. Additionally, a marching cubes (MC) algorithm was used to reconstruct the segmented images into a 3D model. The geometrical model was then meshed for hemodynamic simulation using OpenFOAM.

The proposed Res2Net-ConvFormer-Dilation-UNet (Res2-CD-UNet) model achieved the best results in both the lower-limb and aortic-artery datasets. In the aortic-artery dataset, it achieved an accuracy of 92.76%, which was 1.32% higher than that of the second-best model. In the lower-limb artery dataset, it achieved an accuracy of 94.57%, surpassing the second-best model by 1.12%. The maximum relative geometric error for the lower-limb arteries was only about 2.05%. The overall computational time for the process significantly decreased from several hours to a few minutes, substantially enhancing diagnostic efficiency.

The method developed in this study facilitates the automated segmentation, 3D reconstruction, and CFD simulation of arterial regions in CT images. Our proposed method exhibits high accuracy, and enables the rapid and intuitive visualization of hemodynamic changes in the arteries.