Automated classifiers for early detection and diagnosis of retinopathy in diabetic eyes

Table 2 Test classification performance results obtained in Test 1

TH vs. FD	RNFL (eye/scans)	GCL + IPL (eye/scans)	INL (eye/scans)	OPL (eye/scans)	ONL + IS (eye/scans)	OS (eye/scans)	RPE (eye/scans)
TP	48/288	49/294	48/288	48/288	48/288	50/300	51/306
FN	6/36	5/30	6/36	6/36	6/36	4/24	3/18
TN	10/60	35/210	23/138	36/216	10/60	9/54	11/66
FP	33/198	8/48	20/120	7/42	33/198	34/204	32/192
PPV	0.59	0.86 *	0.71	0.87 *	0.59	0.60	0.61
Sensitivity	0.89	0.91*	0.89	0.89*	0.89	0.93	0.94
Specificity	0.23	0.81*	0.53	0.84*	0.23	0.21	0.26

*denotes the intraretinal layer for which the sensitivity, specificity and PPV are greater than 80%.
Sensitivity, specificity, predictive values (TP, FN, TN, FP) and positive predictive values (PPV) obtained when training the Bayesian radial basis function network using the thickness (TH) and fractal dimension (FD) as the input and target features of the given retinal layers, respectively.