2 3 Shadowed 2D image-based machine learning and automatic 3D landmark detection method and apparatus using thereof

摘要

The present invention relates to an automatic 3D landmark detection method and device using 2D shadow image-based machine learning. The method according to the present invention comprises: a step of segmenting a landmark detection target body part from 3D medical image data acquired by an image acquisition device; a step of generating a plurality of 2D shade images in which the lighting position and the view direction are differently applied to the segmented body parts; a step of detecting a 2D landmark of the segmented body part in the plurality of 2D shade images generated using at least one neural network model; and a step of projecting the detected 2D landmark on the surface of the segmented body part to detect a 3D landmark of the body part. According to the present invention, by generating a multi-shade 2D image capturing a 3D geometric clue to learn the neural network model, it is possible to learn with a relatively small number of learning data, and to solve the memory shortage problem, while accurately detecting a landmark.