Feasibility research is proposed for the development of a computer-based expert radiology assistant (XRA) to aid in the diagnosis of breast cancer. The proposed expert system has the potential to assist radiologists in the interpretation of diagnostic images and clinical data using the artificial intelligence of the system formed by encoding the knowledge of human experts in the computer software. The system will assist in the diagnosis of difficult cases and provide ready access to the knowledge of experts, leading to increased distribution and application of mammographic expertise.
Specific aims of Phase I are designed to provide experimental evidence to test the capability of an expert system to make diagnoses based upon clinical and image descriptors provided by radiologists. A prototype system will be implemented and tested using verified clinical cases in order to aid in the discovery of reasoning methods and data descriptors that constitute sufficient knowledge to make diagnoses at expert levels of performance. Having defined a sufficient set of image descriptors, a Phase II effort will concentrate on full implementation of the interactive image understanding portion of the system, integration of the diagnostic and image analysis subsystems, expansion of the knowledge base, and definitive clinical testing.
