To establish a decision tree using preoperative PET/computed tomography (CT) parameters for detecting contralateral lymph node metastasis (CLNM) in oral cancer patients. In total, 140 patients with a confirmed histopathological diagnosis of oral carcinoma showed fluorodeoxyglucose accumulation of contralateral lymph nodes in PET images. Of 260 lymph nodes, eight were metastatic. We compared metastatic and non-metastatic nodes using Fisher-Freeman-Halton exact and Fisher's exact and Mann-Whitney U tests. The diagnostic performance was analyzed using receiver operating characteristic curves. We established a decision tree using exhaustive chi-squared automatic interaction detection algorithm. Five PET/CT parameters were significantly different between metastatic and non-metastatic nodes: the longest ( P = 0.015) and shortest ( P = 0.023) diameter, and the maximum standardized uptake values (SUVmax) of the contralateral node ( P = 0.030), primary tumor ( P < 0.001), and ipsilateral node ( P < 0.001). The area under the curves of SUVmax of the primary tumor (0.887), ipsilateral node (0.886), and longest diameter (0.752) were the largest and these three parameters were used as predictive criteria in the decision tree. The accuracy of the decision tree was 97.7% with 100% sensitivity and 97.6% specificity. Preoperative PET/CT imaging with a decision tree has the potential to diagnose CLNM in patients with oral cancer.