Fig. 4: User validation test of DeepDx assistance.

A The results of individual cases in UT1 (without AI assistance) and UT2 (with AI assistance). B Kappa and quadratic-weighted kappa values of the grouping against the reference standard in UT1 and UT2. The error bars indicate 95% confidence intervals. C The time spent for UT1 and UT2. Squares and triangles indicate the records for every 30 cases. Stars denote the average time consumed for one case.