Rest-redistribution 201Tl imaging is currently being used for myocardial viability detection, but the ideal parameters for territory classification have not yet been defined. The aim of this study was to define the optimal criteria for detecting viable myocardium and predicting postrevascularization recovery with rest-redistribution 201Tl SPECT. METHODS In 29 patients with left ventricular dysfunction, tracer activity within asynergic segments was quantified on rest and redistribution 201Tl SPECT. Viability was defined by the presence of functional recovery, which was detected by comparing wall motion in baseline and follow-up echocardiography. Discriminant function analysis and receiver operating characteristic (ROC) curve analysis were used to evaluate the relationship between 201Tl data and viability. RESULTS Of 214 dysfunctioning segments (135 a-/dyskinetic), viability was demonstrated in 115 (75 a-/dyskinetic). Both rest and redistribution 201Tl activity in these segments were significantly higher than they were in the nonviable segments (p < 0.0001). Significant (> 10%) reversibility was observed in 39% of the viable and in 36% of the nonviable segments (p = 0.81). Discriminant analysis identified redistribution activity, followed by rest activity, as the most effective predictors of functional recovery. Similar areas were found under the ROC curve for rest (0.68 +/- 0.037) and for redistribution activity (0.70 +/- 0.036) (p = 0.13). ROC curve analysis identified the optimal cutoff for redistribution activity at < 60%, with 147 of 214 (69%) segments correctly classified (sensitivity = 78% and specificity = 58%). In the subset of a-/dyskinetic segments, redistribution activity presented a significantly larger ROC curve area (0.81 +/- 0.038 compared to 0.77 +/- 0.042, p < 0.05), and 103 of 135 (76%) segments were correctly classified (sensitivity = 81% and specificity = 70%). CONCLUSIONS Redistribution activity is the most important parameter to be considered in rest-redistribution 201Tl to differentiate viable from nonviable segments; rest activity is also valuable, whereas the meaning of reversibility appears limited. Cutoff values about 60% appear to give the most reasonable balance between sensitivity and specificity.