Table 4.

Differences in performance between the classification protocol for each view selection protocol.

	Classification Protocol Comparison
View Protocol	B vs. A	A vs. C	B vs. C
All images	+0.01 [0.00; 1]†	+0.04 [0.01; 0.09]*	+0.05 [0.03; 0.10]*
Random images	+0.00 [0.00; 1]†	+0.03 [0.00; 0.09]*	+0.03 [0.00; 0.09]*

Note – Data are indicated as ΔAUC [confidence interval].

Protocol A: Feature averaging

Protocol B: Classifier output averaging

Protocol C: No averaging

^*Two-sided 95% confidence interval

^†Non-inferiority interval