Automatic differentiation of Glaucoma visual field from non-glaucoma visual filed using deep convolutional neural network

Table 2 Performance of the algorithm and the compared methods

	Methods	Accuracy	Specificity	Sensitivity
Ophthalmologists	resident #1	0.640	0.767	0.513
	resident #2	0.593	0.680	0.507
	resident #3	0.587	0.630	0.540
	attending #1	0.533	0.213	0.853
	attending #2	0.570	0.670	0.473
	attending #3	0.653	0.547	0.760
	glaucoma expert #1	0.663	0.700	0.647
	glaucoma expert #2	0.607	0.527	0.687
	glaucoma expert #3	0.607	0.913	0.300
Rule based methods	AGIS	0.459	0.560	0.343
Rule based methods	GSS2	0.523	0.500	0.550
Traditional machine learning methods	SVM	0.670	0.618	0.733
	RF	0.644	0.453	0.863
	k-NN	0.591	0.347	0.870
	CNN	0.876	0.826	0.932

ISSN: 1471-2342