A new knowledge-based lung nodule detection system

We describe a knowledge-based system for segmenting and labeling lung nodules on CT images. The system was developed in a blackboard environment that incorporates a lung knowledge model, image processing model and inference engine. The lung model, which contains anatomical knowledge about the lung in the form of semantic networks, is used to guide the interpretation process. The system works in a hierarchical structure, from large structures to the final nodule candidates, by focusing on the region of interest step by step. The symbolic variables introduced to accomplish high-level inference, are defined by fuzzy confidence functions in the lung model. Composite fuzzy functions are used to map between image and lung model objects. Anatomical lung segment knowledge is embedded in the system to direct 3D validation of suspicious objects. Structures are identified and abnormal objects are reported. Preliminary experiment results are included.