Table 5.

Comparison results between B-DDLN and other classifiers for classifying the same extracted image features.

Classifier	Testing accuracy(%)	F1-score	Precision	Specificity	Sensitivity	Kappa value	AUC
B-DDLN1${}^{\text{a}}$	98.8889	0.9900	0.9900	0.9875	0.9900	0.9775	0.9888
B-DDLN2${}^{\text{b}}$	98.8889	0.9900	0.9900	0.9875	0.9900	0.9775	0.9888
${\text{BPNN}}^{\text{c}}$	95.5556	0.9588	0.9894	0.9875	0.9300	0.9107	0.9587
Naive Bayesian	92.7778	0.9353	0.9307	0.9125	0.9400	0.8536	0.9012
${\text{LDA}}^{\text{d}}$	93.3333	0.9406	0.9314	0.9125	0.9500	0.8647	0.9313
${\text{QDA}}^{\text{e}}$	94.4444	0.9485	0.9787	0.9750	0.9200	0.8883	0.9475
Linear ${\text{SVM}}^{\text{f}}$	96.6667	0.9697	0.9796	0.9750	0.9600	0.9327	0.9675
Gaussian SVM	95.5556	0.9600	0.9600	0.9500	0.9600	0.9100	0.9550
k-${\text{NN}}^{\text{g}}$	95.0000	0.9569	0.9174	0.8875	1.0000	0.8976	0.9437
Logistic regression	90.5556	0.9171	0.8952	0.8625	0.9400	0.8075	0.9012
Decision tree	93.3333	0.9394	0.9490	0.9375	0.9300	0.8653	0.9337
Bagging tree	92.7778	0.9366	0.9143	0.8875	0.9600	0.8528	0.9237
Boosting tree	96.1111	0.9648	0.9697	0.9625	0.9600	0.9213	0.9612

The superior performance of the proposed B-DDLN diagnosis model are highlighted in bold.

^aB-DDLN in case 1.

^bB-DDLN in case 2.

^cBack propagation neural network.

^dLinear discriminant analysis.

^eQuadratic discriminant analysis.

^fSupport vector machine.

^g$k$-nearest neighbor.