In Brief:
A computer-based artificial intelligence neural network was developed to integrate clinical, exercise and imaging data from radionuclide stress testing and thereby obtain an objective, reliable, reproducible method for interpretation of these tests which could be used by anyone with a personal computer. The artificial intelligence method showed a sensitivity of 88% and a specificity of 65% for the diagnosis of ischemic heart disease and was comparable to that of our standard clinical method (sensitivity 80%, specificity 69%), which employs dual expert readers with full access to all relevant clinical information. Incorporation of clinical/exercise data significantly improved predictive accuracy of the network over that of a network constructed using image data alone (Z=2.26, p<05). Finally, the neural network performed as well as the expert readers in the diagnosis of 1, 2 and 3 vessel disease and in identification of the specific coronary vessel(s) involved.