Background: Accurate preoperative staging optimizes treatment of pancreatic cancer. Predicting unresectability in many cases is very difficult, despite advances in diagnostic imaging technology. We propose a scoring system using computed tomography (CT) scan and endoscopic ultrasound (EUS) that more accurately predicts unresectability in patients with pancreatic cancer than either modality alone. Methods: Reports of preoperative EUS and CT scans from 79 patients who underwent exploration for pancreatic cancer were reviewed, as were operative notes and surgical pathology reports. Chi-squares were used to identify radiologic factors significantly correlated with unresectability. A total of five factors were identified as predictors of unresectability and incorporated into a scoring system. Each patient received a score of 0-5 by counting the number of criteria fulfilled. Results: Five criteria were identified that predicted unresectability: 1) suspicious liver lesions too small to biopsy or characterize (p < 0.001); 2) adenopathy (> 1 cm) identified by EUS (p = 0.05); 3) adenopathy (> 1 cm, long axis) identified by CT (p < 0.001; 4) suspicion of vascular involvement on EUS (p < 0.001); and 5) suspicion of vascular involvement on CT (p 0.53; specificity 91%). Patients with scores of 0, 1, 2, and 3 had resectability rates of 83.0%, 36.7 %, 15.4%, and 0%, respectively. The most accurate results were achieved in the group of 49 patients who were evaluated with CT and EUS. In this group of patients, all unresectable patients had a score ≥ 1. 24/30 resectable patients who underwent both CT and EUS had a score of “0,” for an overall accuracy of 88.2%. Conclusion: We propose a scoring system combining EUS and CT to assess risk for unresectability in pancreatic cancer. This scoring system is an effective method of stratifying pancreatic cancer patients into three groups: 1) patients with a score of zero: high probability (83.0%) of successful resection; 2) patients with a score of one: mediocre probability (36.7%) of successful resection; and 3) patients with a score of two or higher: low probability (11.1%) of successful resection. This stratification of resection probabilities can be used to guide surgical management.