[Retracted] Identification of Diabetic Retinopathy through Machine Learning

Table 5

Average performance of the 15 series of experiments with RBM 500, considering four classes (no DR, light DR, mod/severe, and proliferative).

RBM 500		Machine learning
						Hit %
Specialist		Without DR	With DR
		Absence	Light	Mod/record	Proliferative
Without DR	Absence	9	0	5	0	9/14 (64.2%)
With DR	Light	0	2	2	0	2/4
						50%
	Mod/record	0	2	11	0	11/13
						85%
	Proliferative	0	0	0	7	7/7
						100%

RD: diabetic retinopathy, mod: moderate, and RBM: restricted Boltzmann machine.