A method is proposed for segmentation of organs in CT-image sequences. An important feature of the method is the use of search areas to guide the segmentation process. Time consuming, data-directed operations are restricted to these areas, instead of being applied to the whole image. A search area for a particular anatomical structure is determined by constraints, derived from a model of the imaged scene and by results already obtained in the recognition process. The method has been successfully applied on recognition of the spleen in abdominal X-ray CT scans. Q 1988 Academic Pres. Inc.
I. INTR~DUCTI~N
In medical imaging the process of data acquisition and image processing has been improved enormously over recent years. In contrast with this the process of recognition and interpretation of images has not shown a similar progress in medical practice. This relative slow progress must be attributed to the difficulty of accurate segmentation of images. Recent studies addressing this subject are given by Sokolowska and Newell (I) and Stiehl (2), both with applications to X-ray computer tomography (CT) of the brain, and Shani (3). who uses explicit geometrical organ models to recognize structures in abdominal CT. If reliable routines can be developed for extraction of organ outlines in, for instance, CT-scan images, a wide range of possibilities for fast automated analysis is within reach, such as quantification of tissue characteristics, volumetry, and three-dimensional (3D) morphometry. Also 3D display techniques are expected to be useful in many biomedical problems, although clinical validation of these techniques seems to lag their development (4).
In image processing we can distinguish image segmentation from image understanding or recognition (5). Segmentation decomposes the image into regions, while the recognition process generates a description of the image by assigning labels to these regions, corresponding to a model of the imaged scene. Although it might seem advantageous to keep the segmentation process completely data directed, to remain independent of a special image domain, at