Research is proposed for the development of a computer-based expert radiology assistant to aid in the diagnosis of breast cancer using clinical findings, x-ray mammograms, and ultrasound images as input data. Based upon encouraging feasibility results which demonstrated that the Phase I prototype system produced interpretations comparable to expert radiologists, it is proposed to extend the system to more complex diagnostic situations, to incorporate a data base for statistical analysis of mammographic data, to optimize the diagnostic rules, to refine the methods of combining evidence, to improve the user interface, and to investigate the role of objective image analysis.
Specific aims of Phase II are designed to provide experimental evidence to determine the role of expert system technology in breast cancer diagnosis, training of residents, management and reporting of examination data, and assistance in detecting and quantifying of abnormalities. The resulting expert system will be extensively tested, both retrospectively and prospectively, using clinical data to evaluate the practically attainable diagnostic accuracy, overall clinical utility, and potential for commercialization.
