A robust Parkinson’s disease detection model based on time-varying synaptic efficacy function in spiking neural network

Table 5 Performance comparison with different percentage split for dataset #1

Technology	Accuracy (in %)	Sensitivity (in %)	Specificity (in %)	MCC	Precision	F1 Score	Gmean
90% training set − 10% testing set
MLP-NN	95	83.33	100	0.8819	1	0.9091	0.9129
RBF-NN	85	66.67	92.86	0.6299	0.8	0.7273	0.7868
RNN	90	87.5	100	0.7638	1.0	0.9333	0.9354
LSTM	85	81.25	100	0.6813	1.0	0.8965	0.9014
SEFRON [Dataset#1]	100	100	100	1	1	1	1
85% training set − 15% testing set
MLP-NN	90	83.33	91.67	0.7092	0.7143	0.7692	0.8740
RBF-NN	90	66.67	95.83	0.6708	0.8	0.7273	0.7993
RNN	88.7	66	100	0.7512	1.0	0.7911	0.812
LSTM	89.72	100	84.3	0.808	0.7767	0.8743	0.92
SEFRON [Dataset#1]	93.33	100	91.67	0.8292	0.75	0.8571	0.9574
80% training set − 20% testing set
MLP-NN	92.31	87.5	93.55	0.7767	0.7778	0.8235	0.9047
RBF-NN	89.74	62.5	96.77	0.6633	0.8333	0.7143	0.7777
RNN	92.86	100	90.11	0.85	0.80	0.89	0.95
LSTM	90	74.5	94.24	0.9274	1.0	0.9836	0.9837
SEFRON [Dataset#1]	92.31	87.5	93.55	0.7767	0.7778	0.8235	0.9047
70% training set − 30% testing set
MLP-NN	93.22	84.62	95.65	0.8027	0.8462	0.8462	0.8996
RBF-NN	84.74	53.85	93.47	0.5223	0.7	0.6087	0.71
RNN	93.22	93.61	91.67	0.8068	0.9778	0.9565	0.9263
LSTM	93.22	93.61	91.67	0.8068	0.9778	0.9565	0.9263
SEFRON [Dataset#1]	89.83	100	86.96	0.7713	0.6842	0.8125	0.9325

ISSN: 1471-2377