Classification of Pulmonary Nodules by Using Hybrid Features

Table 2

Overall performance results of the proposed classification approach.

	Sensitivity (%)	TCA (%)	Specificity (%)	AUROC	Kappa	RMSE
Method 1
ANN	87.5	84.9	83.8	0.928	0.693	0.356
RF	87.5	80.2	81.8	0.889	0.594	0.375
Bagging (ANN)	100	61.6	100	0.599	0.145	0.557
Bagging (RF)	87.5	86.0	84.2	0.911	0.717	0.339
Adaboost (ANN)	87.5	81.4	82.4	0.920	0.619	0.416
Adaboost (RF)	83.3	75.6	75.8	0.880	0.498	0.384
Method 2
ANN	85.4	83.7	81.6	0.913	0.670	0.327
RF	83.3	83.7	80.0	0.890	0.672	0.356
Bagging (ANN)	100	60.5	100	0.616	0.116	0.522
Bagging (RF)	83.3	81.4	78.9	0.912	0.623	0.346
Adaboost (ANN)	81.3	80.2	76.9	0.869	0.600	0.385
Adaboost (RF)	83.3	80.2	78.4	0.901	0.598	0.351
Method 3
ANN	83.3	83.7	80.0	0.893	0.672	0.358
RF	81.3	77.9	75.7	0.869	0.551	0.383
Bagging (ANN)	39.6	59.3	52.5	0.738	0.224	0.482
Bagging (RF)	83.3	84.9	80.5	0.912	0.696	0.344
Adaboost (ANN)	83.3	83.7	80.0	0.867	0.672	0.378
Adaboost (RF)	85.4	83.7	81.6	0.906	0.670	0.348
Method 4
ANN	89.6	90.7	87.5	0.940	0.812	0.307
RF	85.4	83.7	81.6	0.908	0.670	0.335
Bagging (ANN)	60.4	66.3	59.6	0.770	0.333	0.463
Bagging (RF)	87.5	86.0	84.2	0.922	0.717	0.325
Adaboost (ANN)	87.5	86.0	84.2	0.921	0.717	0.334
Adaboost (RF)	85.4	80.2	80.0	0.911	0.596	0.343