A new application of pattern classification technique to classify the recovery patterns of patients after a coronary artery bypass graft (CABG) surgery is presented. Patients are classified into three types by the physicians: rapid recovery, satisfactory recovery, and marginal recovery. Eleven selected clinical measurements are observed before, immediately after, 1 day, and 2 days after surgery. The feature vector used for the trajectory classification is based on the normalized distances to five physiological states defined in terms of these clinical measurements, at each epoch. The existing data based on 86 patients is divided into two groups: 32 for analysis and the remaining 54 for test. For analysis, the time-dependent and multidimensional characterizing parameters such as the mean and covariance matrices for these three types of patients are estimated from the time history of the 32 analysis CABG patients. A two-level nonlinear recognition algorithm is developed for this study. Very good agreement is found between the physician's and the computer classifications: 100% on analysis data and 86.1% on the test data accumulated up to 2 days after surgery. The result also shows that the earlier prediction of the patient recovery patterns is feasible. The percentage of the correct prediction for the 54 test patients is 63.3% before surgery, 70.6% immediately after surgery, 86.3% one day after the surgery. 100% of the highest-risk test patients can be correctly predicted less than 1 day after surgery, while they normally can be definitely classified by physicians only after 1 day, 2 days, or sometimes 3 days after surgery. © 1979.