Glaucoma Diagnosis with Machine Learning Based on Optical Coherence Tomography and Color Fundus Images

Table 2

The AUC of the machine learning models alone and in combination.

Number	Cases	AUC (mean ± SD)
#1	Disc fundus image (green channel)	0.940 ± 0.039
#2	Disc RNFL thickness map	0.942 ± 0.037
#3	Macular GCC thickness map	0.944 ± 0.032
#4	Disc deviation map	0.949 ± 0.030
#5	Macular deviation map	0.952 ± 0.029
#6	Combination of #2 and #4 (images from disc OCT data)	0.953 ± 0.032
#7	Combination of #3 and #5 (images from macular OCT data)	0.954 ± 0.031
#8	Combination of #1, #2, and #4 (images from disc OCT data with fundus image)	0.959 ± 0.031
#9	Combination of #1, #2, and #3 (automatically detected disc and macular center were not used in creating images)	0.961 ± 0.029
#10	Combination of #2, #3, #4, and #5 (images from OCT data)	0.963 ± 0.030
#11	Combination of all images	0.963 ± 0.029