Fig. 3

The receiver operating characteristic (ROC) curve illustrates the comparison of the true and false positive rates as the minimum “positive” vote necessary to label a lymph node metastatic. The three criteria are a single vote (1:3), majority vote (2:3) and unanimous vote (3:3). Majority voting, initially presumed, proved to be the most effective method for trained reader post-operative OCT analysis, since, of the three data points, it is the furthest from the “random guess” (50:50) line and closest to the “perfect classification” limit. Each of the individual OCT reader data are shown for comparison