Development of a Computer-aided Diagnosis System Using Fuzzy Inference in ^ TlCl Exercise Myocardial Scintigraphy

We have been working on the development of a computer-aided diagnosis system for images in nuclear medicine by using artificial neural networks. A physician's diagnosis of coronary artery disease in nuclear cardiology is performed not only on the basis of imaging data but also includes other factors such as measurement data from exercise scintigraphy. Therefore, we propose an expert system that uses fuzzy inference to estimate the number of abnormal vessels in cases of single-or multi-vessel disease(including normal vessels)of the coronary arteries. The main characteristic of this system is that it integrates information from various sources, including the physician's impressions. In this study, we investigated the system's clinical effectiveness. Results indicated a rate of agreement between the system's confidence level of inference and the physician's diagnosis of 62.2% and a rate of sensitivity of 83% for coronary artery disease. The computer made it possible to utilize vague factors such as a physician's assessment based on experience and intuition. These results demonstrate the usefulness of our proposed technique.