3 METHOD AND APPARATUS FOR DETECTING LANDMARKS FROM MEDICAL 3D VOLUME IMAGE BASED ON DEEP LEARNING

摘要

A method for automatically detecting feature points of 3D medical image data using deep learning includes receiving a 3D volumetric medical image, generating a 2D brightness value projection image based on the 3D volumetric medical image, and the 2D brightness value Based on the projection image, automatically detecting an initial anatomical feature point using a first convolutional neural network, generating a three-dimensional volume of interest region based on the initial anatomical feature point, and generating the three-dimensional volume of interest region and automatically detecting detailed anatomical feature points based on a second convolutional neural network different from the first convolutional neural network.