Systematic protocol for assessment of the validity of BOLD MRI in a rabbit model of inflammatory arthritis at 1.5 tesla

Abstract

Blood-oxygen-level-dependent (BOLD) MRI has the potential to identify regions of early hypoxic and vascular joint changes in inflammatory arthritis. There is no standard protocol for analysis of BOLD MRI measurements in musculoskeletal disorders. To optimize the following BOLD MRI reading parameters: (1) statistical threshold values (low, r > 0.01 versus high, r > 0.2); (2) summary measures of BOLD contrast (percentage of activated voxels [PT%] versus percentage signal difference between on-and-off signal intensities [diff_on_off]); and (3) direction of BOLD response (positive, negative and positive + negative). Using BOLD MRI protocols at 1.5T, arthritic (n = 21) and contralateral (n = 21) knees of 21 juvenile rabbits were imaged at baseline and on days 1, 14 and 28 after a unilateral intra-articular injection of carrageenan. Nine non-injected rabbits served as external control knees (n = 18). By comparing arthritic to contralateral knees, receiver operating characteristic curves were used to determine diagnostic accuracy. Using diff_on_off and positive + negative responses, a threshold of r > 0.01 was more accurate than r > 0.2 (P = 0.03 at day 28). Comparison of summary measures yielded no statistically significant difference (P > 0.05). Although positive + negative (AUC = 0.86 at day 28) and negative responses (AUC = 0.90 at day 28) for PT% were the most diagnostically accurate, positive + negative responses for diff_on_off (AUC = 0.78 at day 28) also had acceptable accuracy. The most clinically relevant reading parameters included a lower threshold of r > 0.01 and a positive + negative BOLD response. We propose that diff_on_off is a more clinically relevant summary measure of BOLD MRI, while PT% can be used as an ancillary measure.