A generic labeling scheme for segmented cardiac MR images
Michel Bister, Jan P. Cornelis, Yves Taeymans, Nils Langloh · 2002
M. Bister et al. (1989) developed an algorithm for the automatic segmentation of medical images: the cavity detector. A labeling scheme, the performance of which has been tested on cardiac magnetic resonance (MR) images, was designed based on this algorithm. The labeling scheme consists of three steps: the first step standardizes the image position, the second step checks the standardized image against spatial knowledge about the position of anatomical objects, and the third step generates a proposal for the labeling. The system is self-learning, and although it does not use typical AI tools, it makes use of some AI concepts such as trainability and separation between model and 'reasoning engine'. The performance in terms of labeling quality and speed is discussed.>