Table 4.

Effect of batch size on FundusNet performance on detecting referable DR.

Batch size	AUC (SD)—test dataset	P value (vs 32 batch size)	P value (vs 2048 batch size)
32	0.77 (0.039)	REF	< 0.001*
64	0.78 (0.058)	0.18	< 0.001*
128	0.78 (0.045)	0.16	< 0.001*
256	0.80 (0.077)	0.06	< 0.001*
512	0.86 (0.082)	0.02*	< 0.001*
1024	0.87 (0.034)	< 0.01*	< 0.01*
2048	0.91 (0.014)	< 0.001*	REF
4096	0.91 (0.011)	< 0.001*	0.08

AUC area under the ROC curve, SD standard deviation, Ref reference.

*Significant difference; P value from measuring statistical significance using DeLong’s test for comparing pairwise AUC values among the batch size (all vs batch size 32 in third column, and all vs batch size 2048 in fourth column). This pairwise comparison was done to show whether the AUCs coming from experiments with all the batch sizes are significantly different that AUCs coming from experiments with batch size 32 (initial batch size) and 2048 (optimal batch size).