Integration of multiple knowledge sources in a system for brain CT-scan interpretation based on the blackboard model

Medical image interpretation is a complex task that requires the integration of knowledge acquired from different domains, such as medicine, computer vision and image processing. This paper describes a knowledge based brain CT scan interpretation system that uses the blackboard model to integrate various sources of knowledge. The frame-based representation technique is employed to represent the geometric model of the human brain. The knowledge on low level image processing algorithms and high level interpretation is partitioned into knowledge sources (KSs) that operate on and communicate through the domain blackboard. Several numeric image processing algorithms are coded into KSs that segment the images or extract features from the image primitives. For the mapping of image primitives to brain objects, there are two groups of mapping KSs, namely model-directed and data-directed. The system achieves the successful labeling and delineation of about 25 brain objects