Table 6 Confusion matrices for predicting heart disease and HF severity based on neural networks retrained with the augmented synthetic [PCr], [ATP] and kf data to balance class size. (a) Results from the network as in Table 2, retrained with 388 cases comprised of groups of 97 with HF, and 97 synthetically-augmented cases of HCM, MI and control subjects. (b) Results from the network as in Table 4, retrained with 168 cases of HF comprised of 56 patients with NYHA Class II severity and two synthetically-augmented groups of 56 each, with NYHA Class I and III
