This paper presents a self-adaptive expert system for brain tumor diagnosis we have designed. The Brain Tumor Diagnostic System (BTDS) we propose consists of three main components: knowledge building, inference, and knowledge refinement. In the knowledge building component, an inductive learning algorithm, RASSISTANT, constructs an initial knowledge base from noisy examples, eliminating a major difficulty in developing diagnostic expert systems. In the inference component, an inference engine exploits rules in the knowledge base to help diagnosticians determine brain tumor causes according to computer tomography pictures. A simple rule refinement scheme, PCC, is also proposed to modify the existing knowledge base during inference, which dramatically improves the accuracy of the derived rules. BTDS performance has been evaluated on 270 actual brain tumor cases. Results show that BTDS can achieve an accuracy of over 98%.
Relation:
Journal of Information Science and Engineering 11(2):274-294
