Seed Grant Profile
Artificial Intelligence Decision Support Tools for Nuclear Cardiac Medicine
The long-term objective of the project is to improve the care of patients by developing imaging expert systems that are automatically updated with the latest scientific/clinical knowledge shown to benefit our population. This will be done by using natural language understanding (NLU) techniques from artificial intelligence (AI) to extract new knowledge from published clinical literature, inserting rules derived from this knowledge into the knowledge base of an expert system, and testing the derived rules against a large database of patientsí imaging, clinical, diagnostic and prognostic results.
We propose to develop a proof-of-concept in the area of Nuclear Cardiac Medicine, the most expensive procedure widely performed under Medicare today and a critical need for our nation. Specifically, we will apply this idea to myocardial perfusion SPECT imaging because: (a) this field is well documented in the published scientific literature, (b) it is the domain expertise of one of the PIs (Dr. Garcia), (c) we have at Emory a large database of patient studies and results, and (d) we have previously developed an expert system in this area which is in use at 10,000 hospitals worldwide (a collaborative effort of Dr. Garcia and a Georgia Tech faculty member, Dr. Norberto Ezquerra).
A major limitation of this (and similar) medical decision support systems is that their hand-engineered knowledge bases rapidly get obsolete as new medical knowledge is discovered. The proposed project will address this limitation by developing NLU tools to extract new knowledge and update the knowledge bases of such systems. The expected outcomes of our project are: (i) an NIH proposal submission by 2007, (ii) NIH funding by 2008, and (iii) commercial distribution of a functioning expert system by 2009, resulting long-term in: (iv) improved patient care through increased accuracy of detecting coronary artery disease (CAD) and (v) further application of this methodology for other imaging modalities. We are investigating whether this overall approach is patentable and thus the reason for the request for confidentiality of the proposal. Both PIs have previously founded startup companies based on university research, and we believe that commercialization is an important objective in developing technologies that have a true impact beyond the academic world.
Investigators: Ashwin Ram (GT, Computing) and Dr. Ernest V. Garcia (Emory, Radiology)