Accuracy of Opposed-phase Magnetic Resonance Imaging for the Evaluation of Treated and Untreated Spinal Metastases.

RATIONALE AND OBJECTIVES: To assess whether the accuracy of opposed-phase magnetic resonance (MR) imaging to differentiate spinal metastases from benign lesions is influenced by treatment.

MATERIALS AND METHODS: We retrospectively evaluated 25 benign lesions, 25 untreated spinal metastases, and 89 treated spinal metastases in 101 patients who underwent opposed-phase MR spine imaging at our institution. The largest possible region of interest was placed over the lesion in question on out-of-phase and in-phase MR sequences, and the signal intensity ratio (SIR) of the lesions was calculated. The SIRs were compared between benign, untreated, and treated lesions. Receiver operator characteristic (ROC) curves were used to identify the optimal threshold to differentiate benign lesions from untreated spinal metastases, and the accuracy of this threshold was assessed for treated spinal metastases, chemotherapy-treated spinal metastases, and radiated spinal metastases.

RESULTS: Benign lesions had lower mean SIR than untreated (P = 2.4 × 10-8 , 95% confidence interval [0.29, 0.51]) and treated spinal metastases (P = .51; 95% confidence interval [-0.13, 0.06]). A cutoff SIR of 0.856 had an accuracy of 88.00% for untreated lesions, 77.48% for previously treated lesions, and 70.45% for previously radiated lesions. The ROC curve to differentiate benign lesions from radiated spinal metastases was significantly different from the ROC curve to differentiate benign lesions from untreated spinal metastases (P = .0180). The ROC curve to differentiate benign lesions from lesions treated with chemotherapy only was significantly different from the ROC curve to differentiate between benign lesions and radiated spinal metastases (P = .041).

CONCLUSIONS: Opposed-phase imaging is less accurate for treated spinal metastases, in particular after radiation.