Deep-Learning Based Automatic Determination of Cardiac Planes in Survey MRI Data

Inference of the radiological planes of the heart in MRI is a crucial step for valid data acquisition to examine the structure and function of the human heart in detail. In this paper, we present a deep learning model for automatic inference of the radiological plane of the heart from 3D survey sequences. The proposed neural network is based on the V-Net [6] architecture that has been developed to perform inference on the radiological positions of the hearts. The network is designed to take a 3D image as input and generate a regressed heatmap of probable plane positions as output. The results show that the proposed method is feasible for automatic geometry planning. It has the potential to increase the efficiency of medical imaging. The presented networks show that they can locate cardiac landmarks even from data with anisotropic voxels. It can improve the accuracy and speed of diagnosis, allowing for faster and more effective treatment.