PURPOSE: To determine the diagnostic performance of MRI in assessing local tumour extent and evaluate associations between MRI features and survival in patients undergoing MRI before pelvic exenteration for persistent or recurrent gynaecological cancers. METHODS AND MATERIALS: The study included 50 patients with persistent or recurrent gynaecological malignancies who underwent pelvic exenteration between January 1999 and December 2011 and had MRI at most 90 days before surgery. Two radiologists independently assessed invasion of adjacent organs (on a 5-point scale). Diagnostic accuracy, inter-reader agreement, and associations between organ invasion on MRI and patient survival were evaluated. RESULTS: Areas under receiver operating characteristic curves (AUCs) for invasion of the bladder, rectum and pelvic sidewall were 0.96, 0.90 and 0.98 for reader 1 and 0.95, 0.88 and 0.90 for reader 2. Corresponding sensitivities/specificities were 87.0 %/92.6 %, 81.3 %/97.0 % and 87.5 %/97.2 % for reader 1, and 87.0 %/100.0 %, 75.0 %/97.0 % and 75.0 %/94.4 % for reader 2. Inter-reader agreement was excellent for organ invasion (κ = 0.81-0.85). Pelvic sidewall invasion on MRI was associated with overall and recurrence-free survival (P = 0.01-0.04 for the two readers). CONCLUSION: Preoperative MRI is accurate in predicting organ invasion. It may guide surgical planning and serve as a predictive biomarker in patients undergoing pelvic exenteration for gynaecological malignancies. KEY POINTS: • MRI can accurately assess bladder and rectal wall invasion before major surgery. • MRI identifies patients requiring extended pelvic exenteration by detecting sidewall invasion. • Inter-reader agreement for detecting organ invasion and tumor size is excellent. • Pelvic sidewall invasion on MRI is associated with shorter overall and recurrence-free survival.