Objective Studies have demonstrated that the circulating cell-free DNA (cfDNA) may be a potential biomarker for diagnosis of thyroid cancer. Therefore, this study was performed to investigate the value of a scoring model established by plasma circulating cfDNA and changes in ultrasound characteristics of the thyroid nodules in differential diagnosis between benign and malignant thyroid nodules.Methods Two hundred and forty patients with thyroid nodules (132 cases of thyroid cancer and 108 cases of benign thyroid nodules) admitted from June 2018 to October 2020 were enrolled for this study. They were randomly divided into the modeling group and validation group using a 1∶1 ratio. The plasma cfDNAs were extracted from the 240 patients by blood DNA extraction kit, and the DNA concentrations were further detected by qRT-PCR. Thyroid ultrasound was performed in all patients. The scoring model was constructed based on the cfDNA concentration and thyroid ultrasound characteristics, and then, its sensitivity, specificity, positive predictive value and negative predictive value for diagnosis of thyroid cancer were analyzed. The effectiveness of the model was evaluated in the validation group.Results The cfDNA concentration in patients with thyroid cancer was significantly higher than that in patients with benign nodules (P<0.001). The results of univariate analysis showed that there were significant differences in cfDNA concentration and ultrasound imaging features that included the aspect ratio, internal echo, integrity of the capsule, calcification and cystic lesions between patients with malignant and benign nodules (all P<0.05). The results of multivariate Logistic regression analysis showed that cfDNA≥56.84 ng/mL, and thyroid ultrasound presenting the aspect ratio ≥1, incomplete capsule, hypoechoic, calcification, and non-cystic lesions were independent risk factors for diagnosis of thyroid cancer (all P<0.05). According to the standard regression coefficients of above variables, a scoring model was established. The area under the ROC curve (AUC) of the model for diagnosis of thyroid cancer was 0.958 (95% CI=0.926-0.989), and its optimal cut-off value was 5.5, with a diagnostic sensitivity, specificity, positive predictive value, and negative predictive value of 85.5%, 89.7%, 89.8%, and 85.2%, respectively, which were all superior to those of the predictive power of each single variable. The validation results showed that the AUC was 0.902 (95% CI=0.848-0.957) in validation group.Conclusion The scoring model based on plasm cfDNA and ultrasound features of the thyroid nodules has high predictive value for diagnosis of thyroid cancer, and it provides a reference for the differential diagnosis of thyroid benign and malignant nodules. Clinical intervention should be aggressively performed when the score of a thyroid nodule ≥5.5.