Chemical shift imaging with in-phase and opposed-phase sequences at 3 T: what is the optimal threshold, measurement method, and diagnostic accuracy for characterizing marrow signal abnormalities?

Objective: To determine the threshold signal drop on 3-T chemical shift imaging (CSI), with in-phase (IP) and opposed-phase (OP) sequences, for accurately identifying bone marrow replacement with 100% sensitivity, and determine a clinically useful measurement method for deriving such a threshold. Materials and methods: From a convenience series of 157 MRIs, 36 cases with histologically proven marrow-replacing lesions and 22 sites of red marrow (histologically proven (2) or with minimum 6-month stability) with 3-Tesla CSI were included. Two musculoskeletal radiologists performed two measurement methods (first: multiple algorithmic ROIs at the top, middle, and bottom of lesions (M-ROI); second: an ROI was drawn where there appeared to be the least opposed-phase signal reduction qualitatively/visually (Q-ROI)). Lesional and red marrow signal change (%,[(IP-OP)signal/IP signal]*100) was determined. Statistical analyses included Student’s t test, Cohen’s kappa, and receiver operator characteristic curve generation. Results: By M-ROI, lesion signal change was − 0.508% (confidence interval (CI) = − 5.537:4.521) and 1.348% (CI = − 3.541:6.311) for readers 1 and 2. By Q-ROI, lesion signal change was − 11.03% (CI = − 17.01:- 5.046) and − 5.657% (CI = − 12.36:1.048) for readers 1 and 2. For all M-ROI and Q-ROI measurement strategies, signal change between lesional tissue and red marrow was significantly different (p < 0.0001). QROI produced the best composite sensitivities and specificities with a maximized Youden index of 0.955–1. A threshold signal drop of 25% with Q-ROI produced at least 100%/86% sensitivity/specificity for both readers for identifying marrow replacement. Conclusions: For 3-T CSI, a single visually targeted measurement using a 25% threshold is accurate for identifying marrow-replacing lesions.