Observer reliability and accuracy of nuclear magnetic resonance

The purpose of this research was to: (1) evaluate two sources of error in the performance of magnetic resonance imaging (MRI) in three anatomic areas and (2) compare the error estimates for MRI to computerized tomography (CT). The sources of error assessed were inter- and intra-observer reliability. The traditional accuracy measurements of sensitivity and specificity and ROC were used along with a newly created accuracy measured called the RA index. For the MRI neuroscans, the superconductor MRI had significantly greater inter-observer reliability over the resistive MRI although the accuracy measures were equivalent. Among the extremity scans, the resistive MRI had greater inter-observer reliability than the superconductor MRI. However, when considering reliability and accuracy together, the superconductor MRI had a significantly higher RA index. The comparisons of MRI to CT yielded statistically different performance measures among the abdomen/pelvis scans only. MRI had greater inter-observer reliability, intra-observer reliability, and RA index scores. The statistical difference observed in the RA scores, however, was due to the higher RA scores of the superconductor MRI. In fact, the overall RA score for the resistive MRI was equivalent to the overall RA score for CT