. 2020 Oct 19;25(20):4792. doi: 10.3390/molecules25204792

Table 7.

Comparison with related works of PD classification.

Author (year) [reference]	Category	Sample Size	Method/# Feature	Classifier	Accuracy
R. Prashanth et al. (2016) [37]	Normal	208	Machine learning/34	SVM	0.97
R. Prashanth et al. (2016) [37]	PD	427	Machine learning/34	SVM	0.97
Abdelbasset Brahim et al. (2017) [38]	Normal	111	Voxels as Features approach and Principal Component Analysis	SVM	0.88
Abdelbasset Brahim et al. (2017) [38]	PD	158		SVM	0.88
Ehsan Adeli et al. (2017) [39]	Normal	169	Kernel-based Feature	SVM	0.95
Ehsan Adeli et al. (2017) [39]	PD	369	Kernel-based Feature	SVM	0.95
Mosarrat Rumman et al. (2018) [40]	Normal	100	ROI detection and area calculation	ANN	0.94
Mosarrat Rumman et al. (2018) [40]	PD	100	ROI detection and area calculation	ANN	0.94
Presented Method: Deep CNN	4 stage	Healthy = 6 early = 245 mid = 265 late = 470	Popular Pre-trained models in CNN	Alex Net (grayscale)	0.83
	4 stage	Healthy = 6 early = 245 mid = 265 late = 470	Popular Pre-trained models in CNN	DenseNet (color)	0.85
	6 stage	Healthy =6 HYS1= 110 HYS2 = 135 HYS3 = 265 HYS4 = 435 HYS5 = 35	Popular Pre-trained models in CNN	VGG19 (grayscale)	0.78
	6 stage		Popular Pre-trained models in CNN	DenseNet (color)	0.78