OBJECTIVES: To build radiomics based OS prediction tools for local advanced cervical cancer (LACC) patients treated by concurrent chemoradiotherapy (CCRT) alone or followed by adjuvant chemotherapy (ACT). And, to construct adjuvant chemotherapy decision aid. METHODS: 83 patients treated by ACT following CCRT and 47 patients treated by CCRT were included in the ACT cohort and non-ACT cohort. Radiomics features extracted from primary tumor area of T2-weighted MRI. Two radiomics models were built for ACT and non-ACT cohort in prediction of 3 years overall survival (OS). Elastic Net Regression was applied to the the ACT cohort, meanwhile least absolute shrinkage and selection operator plus support vector machine was applied to the non-ACT cohort. Cox regression models was used in clinical features selection and OS predicting nomograms building. RESULT: The two radiomics models predicted the 3 years OS of two cohorts. The receiver operator characteristics analysis was used to evaluate the 3 years OS prediction performance of the two radiomics models. The area under the curve of ACT and non-ACT cohort model were 0.832 and 0.879, respectively. Patients were stratified into low-risk group and high-risk group determined by radiomics models and nomograms, respectively. And, the low-risk group patients present significantly increased OS, progression-free survival, local regional control, and metastasis free survival compare with high-risk group (P < 0.05). Meanwhile the prognosis prediction performance of radiomics model and nomogram is superior to the prognosis prediction performance of Figo stage. CONCLUSION: The two radiomics model and the two nomograms is a prognosis predictor of LACC patients treated by CCRT alone or followed by ACT.