Abstract
Several methods of testing statistically significant differences were used to evaluate the usefulness of one diagnostic imaging system compared with any other imaging system using ROC analysis. Comparison of usefulness between conventional methods such as the two-tailed paired t-test and 95% confidence intervals, and Jackknife method for ROC analysis, which was reported a few years ago, were investigated with actual two experimental results from ROC studies. From the testing results, it was considered that conventional methods reguire large numbers of readers in order to improve the reliability of statistically significant test differenxes between the two imaging systems. When determining differences between two kinds of imaging modalities, the use of the Jackknife method was better than conventional methods in terms of the reliability of test results, because conventional methods do not consider variances between readers and case samples.
