We evaluated the accuracy of PET/CT with 68 Ga-PSMA-HBED-CC-a 68 Ga-conjugated ligand of human prostate-specific membrane antigen (PSMA)-to localize cancer in the prostate and surrounding tissue at initial diagnosis. Methods: Twenty-one patients with biopsy-proven prostate cancer underwent 68 Ga-PSMA-HBED-CC ( 68 Ga-PSMA) PET/CT at a median of 4 d (range, 0-47 d) before radical prostatectomy. Based on a 6-segment model, the Gleason score and proportion of tumor tissue within each segment (segmental tumor burden, or STB) as determined by histopathology (STB HP ) were correlated with SUV max and STB as determined by different SUV cutoffs for 68 Ga-PSMA PET (STB ). Furthermore, the involvement of seminal vesicles and other extracapsular extension were assessed by histopathology and PET/CT. Results: Histopathology-positive segments (n 5 100 of 126; 79%) demonstrated a significantly higher mean ± SD SUV max (11.8 ± 7.6) than histopathology-negative segments (4.9 ± 2.9; P , 0.001). Receiver-operating-characteristic analysis revealed an optimal SUV max cutoff of 6.5 for discrimination of histopathology-positive segments from histopathology-negative segments (area under the curve, 0.84; P , 0.001), which gave 67% sensitivity, 92% specificity, a 97% positive predictive value, a 42% negative predictive value, and 72% accuracy. STB PET3 as determined by (2 · blood SUV) 1 (2 · SD) correlated best with STB HP (Pearson ρ 5 0.68; P , 0.001; mean difference ± SD, 19% ± 15%). PET/CT correctly detected invasion of seminal vesicles (n 5 11 of 21 patients; 52%) with 86% accuracy and tumor spread through the capsule (n 5 12; 57%) with 71% accuracy. Conclusion: 68 Ga-PSMA PET/CT accurately detected the location and extent of primary prostate cancer. Our preliminary findings warrant further investigation of 68 Ga-PSMA PET/CT in conjunction with needle biopsy.