A muti-organ localization method in CT volumes
Yuxuan Zhang, Jiwei Liu, Jianfei Liu · 2017
This paper proposes a robust method for multi-organ localization of 3D CT images. Improved Haar-like feature is implemented to narrow down the beginning and ending positions of organ slices. Relying on the fact that voxel intensity of one organ may approximate to others', characteristics metrics are generated based on intensity and spatial position of every single voxel in a target area and trained by random forests afterward, such that a more robust output is presented as a probability density. A polished user interface is also designed deliberately for display of patient's organ bounding boxes.