Background and Purpose: Infarct volumes predict malignant infarcts in patients undergoing decompressive hemicraniectomy (DH) for large middle cerebral artery territory infarcts. The aim of the study was to determine the optimal magnetic resonance imaging infarct volume threshold that predicts a catastrophic outcome at 1 year (modified Rankin Scale score of 5 or death). Methods: We included consecutive patients who underwent DH for large middle cerebral artery infarcts. We analyzed infarct volumes before DH with semi-automated methods on b1000 diffusion-weighted imaging sequences and apparent diffusion coefficient maps. We studied infarct volume thresholds for prediction of catastrophic outcomes, and analyzed sensitivity, specificity, and the area under the curve, a value ≥0.70 indicating an acceptable prediction. Results: Of 173 patients (109 men, 63%; median age 53 years), 42 (24.3%) had catastrophic outcomes. Magnetic resonance imaging b1000 diffusion-weighted imaging and apparent diffusion coefficient infarct volumes were associated to the occurrence of 1-year catastrophic outcome (adjusted odds ratio, 9.17 [95% CI, 2.00–42.04] and odds ratio, 4.18 [95% CI, 1.33–13.19], respectively, per 1 log increase). The optimal volume cutoff of were 211 mL on b1000 diffusion-weighted imaging and 181 mL on apparent diffusion coefficient maps. The 2 methods showed similar sensitivities and specificities and overlapping area under the curve of 0.64 (95% CI, 0.54–0.74). Conclusions: In patients with large middle cerebral artery infarcts, optimal magnetic resonance imaging infarct volume thresholds showed poor accuracy and low specificity to predict 1-year catastrophic outcome, with different b1000 diffusion-weighted imaging and apparent diffusion coefficient thresholds. In the setting of DH, optimal infarct volumes alone should not be used to deny DH, irrespectively of the method used.