Proton resonant frequency shift is the most frequently used method for MR-thermometry. This method is sensitive to patient motion which poses issues that affect the MR-thermometry accuracy. Few studies have evaluated MR thermometry in the pelvis, but only in volumes far from the regions with air motion, and without an objective patient data exclusion criteria. In this study, we assessed accuracy of MR thermometry for a selected group of patients with cervical carcinoma. We showed that changing gastrointestinal air volume was an important confounder for MR thermometry accuracy and that this can be exploited for selection criteria prior to treatment.