Background: The new gold standard for myocardial viability assessment is late gadolinium enhancement-cardiovascular magnetic resonance (LGE-CMR); this technique has demonstrated that the transmural extent of scar predicts segmental functional recovery. We now asked how the number of viable and number of viable+normal, segments predicted recovery of global left ventricular (LV) function in patients undergoing CABG. Finally, we examined which segmental transmural threshold of scarring best predicted global LV recovery. Methods and Results. Fifty patients with reduced LV ejection fraction (EF) referred for CABG were recruited, and 33 included in this analysis. Patients underwent CMR to assess LV function and viability pre-operatively at 6 days and 6 months. Mean LVEF 38% 11, which improved to 43% 12 after surgery. 21/33 patients improved EF by 3% (EF before 38% 13, after 47% 13), 12/33 did not (EF before 39% 6, after 37% 8). The only independent predictor for global functional recovery after revascularisation was the number of viable+normal segments: Based on a segmental transmural viability cutoff of <50%, ROC analysis demonstrated 10 viable+normal segments predicted 3% improvement in LVEF with a sensitivity of 95% and specificity of 75% (AUC = 0.9, p < 0.001). Transmural viability cutoffs of <25 and <75% and a cutoff of 4 viable segments were less useful predictors of global LV recovery. Conclusions: Based on a 50% transmural viability cutoff, patients with 10 viable+normal segments improve global LV function post revascularisation, while patients with fewer such segments do not. LGE-CMR is a simple and powerful tool for identifying which patients with impaired LV function will benefit from CABG. Trial registration. Research Ethics Committee Unique Identifier: NRES:05/Q1603/42. The study is listed on the Current Controlled Trials Registry: ISRCTN41388968. URL: http://www.controlled- trials.com.