To explore the diagnostic accuracy improved by magnetic resonance imaging (MRI) biomarkers for lymph node metastasis in T1-2 stage rectal cancer before treatment. Medical records of 327 patients with T1-2 rectal cancer who underwent pretreatment MRI and rectal tumor resection between January 2015 and November 2019 were retrospectively analyzed. Fifty-seven cases were divided into the lymph node metastasis group (N+ group) while other 270 cases in the non-lymph node metastasis group (N-group) according to the pathologic diagnosis. Two radiologist evaluated the tumor characteristics of MRI images. The relationship of the clinical and imaging characteristics of lymph node metastasis was assessed by using univariate analysis and multivariable logistic regression analysis. Receiver operating characteristic (ROC) curve was used to evaluate the diagnostic abilities for the differentiation of N- from N+ tumors. Among the 327 patients, MR-N evaluation was positive in 67 cases, which was statistically different from the pathological diagnosis (<0.001). The sensitivity, specificity and accuracy of MRI for lymph node metastasis were 45.6%, 84.8% and 78.0%, respectively. Multivariate regression analysis showed that tumor morphology (=0.002), including mucus or not (<0.001), and MR-N evaluation (<0.001) were independent influencing factors for stage T1-2 rectal cancer with lymph node metastasis. The area under the ROC curve of rectal cancer with lymph node metastasis analyzed by the logistic regression model was 0.786 (95% 0.720~0.852). Tumor morphology, including mucus or not, and MR-N evaluation can serve as independent biomarkers for differentiation of N- and N+ tumors. The model combined with these biomarkers facilitates to improve the diagnostic accuracy of lymph node metastasis in T1-2 rectal cancers by using MRI.