Investigations of CNN for Medical Image Analysis for Illness Prediction

Table 8

Using digital fundus images of diabetic retinopathy, the DCNN can determine whether or not there are any benign qualities in the soft and hard exudates.

Benignity	Observed		Cumulative		FPR	TPR	AUC
	True	False	True	False
			0	0	1.000000	1.000000	0.064516
1	34	3	34	3	0.935484	0.989247	0.118259
2	63	7	97	10	0.815939	0.964158	0.160998
3	88	11	185	21	0.648956	0.924731	0.184244
4	105	14	290	35	0.449715	0.874552	0.204117
5	123	23	413	58	0.216319	0.792115	0.142791
6	95	60	508	118	0.036053	0.577061	0.009855
7	9	75	517	193	0.018975	0.308244	0.003509
8	6	41	523	234	0.007590	0.161290	0.001224
9	4	30	527	264	0.000000	0.053763	0.000000
10	0	15	527	279	0.000000	0.000000	0.000000
	527	279					0.889515