BACKGROUND CONTEXT Pedicle screw loosening is common after spinal fusion and can be associated with pseudoarthrosis and pain. With suspicion of screw loosening on standard radiographs, CT is currently considered the advanced imaging modality of choice. MRI with new metal artifact reduction techniques holds potential to be sensitive in detection of screw loosening. The sensitivity and specificity of either of the imaging modalities are yet clear. PURPOSE To evaluate the sensitivity and specificity of three different image modalities (standard radiographs, CT, and MRI) for detection of pedicle screw loosening. STUDY DESIGN/SETTING Cross-sectional diagnostic study. PATIENT SAMPLE Forty-one patients (159 pedicle screws) undergoing revision surgeries after lumbar spinal fusion between August 2014 and April 2017 with preoperative radiographs, CT, and MRI with spinal metal artifact reduction (STIR WARP and TSE high bandwidth sequences). OUTCOME MEASURES Sensitivity and specificity in detection of screw loosening for each imaging modality. METHODS Screw torque force was measured intraoperatively and compared with preoperative screw loosening signs such as peri-screw edema in MRI and peri-screw osteolysis in CT and radiographs. A torque force of less than 60 Ncm was used to define a screw as loosened. RESULTS Sensitivity and specificity in detection of screw loosening was 43.9% and 92.1% for MRI, 64.8% and 96.7% for CT, and 54.2% and 83.5% for standard radiographs, respectively. CONCLUSIONS Despite improvement of MRI with metal artifact reduction MRI technique, CT remains the modality of choice. Even so, CT fails to detect all loosened pedicle screws.