Usage of Probabilistic and General Regression Neural Network for Early Detection and Prevention of Oral Cancer

Table 2

Performance of PNN/GRNN model for classification of various types of oral cancer.

Estimation parameter	Performance (in %)
	Acantholytic	Adenocarcinoma	Basaloid	Lymphoepithelioma-like	Plaque-like	Sarcomatoid	SCC	Verrucous	Benign
Accuracy	99.61	99.41	99.71	99.90	99.90	99.61	99.95	95.61	100
True positive (TP)	0.10	0.10	0.10	0.00	0.00	0.10	56.20	1.37	35.90
True negative (TN)	99.51	99.32	99.61	99.90	99.90	99.51	37.76	94.24	64.10
False positive (FP)	0.00	0.00	0.00	0.00	0.00	0.00	6.05	0.10	0.00
False negative (FN)	0.39	0.59	0.29	0.10	0.10	0.39	0.00	4.29	0.00
Sensitivity	20.00	14.29	25	0.00	0.00	20.00	100	24.14	100
Specificity	100	100	100	100	100	100	86.19	99.90	100
Geometric mean of sensitivity and specificity	44.72	37.80	50	0.0	0.00	44.72	92.84	49	100
Positive predictive value (PPV)	100	100	100	0.00	99.90	100	90.28	93.33	100
Negative predictive value (NPV)	99.61	99.41	99.71	99.90	0.00	99.61	100	95.64	100
Geometric mean of PPV and NPV	99.80	99.71	99.85	0.00	0.00	99.80	95.02	94.48	100
Precision	100	100	100	0.00	0.00	100	90.28	93.33	100
Recall	20.00	14.29	25	0.00	0.00	20.00	100	24.14	100
-measure	0.33	0.25	0.40	0.00	0.00	0.33	0.94	0.38	1.00