A Multichannel 2D Convolutional Neural Network Model for Task-Evoked fMRI Data Classification

Table 2

Classification accuracy over different sample sizes.

Model	Accuracy (mean ± std) over different sample sizes
	2000 samples (200 subjects) (%)	5000 samples (500 subjects) (%)	9950 samples (995 subjects) (%)
mv2D CNN	53.70 ± 4.20	60.88 ± 2.28	63.36 ± 2.19
3D CNN	72.70 ± 2.54	77.36 ± 1.95	82.34 ± 1.27
3D SepConv	73.60 ± 1.77	77.24 ± 2.79	80.44 ± 1.16
1D CNN	67.40 ± 2.92	76.52 ± 1.09	80.76 ± 1.69
s2D CNN	66.20 ± 3.19	76.64 ± 1.96	81.80 ± 0.89
M2D CNN	71.70 ± 1.81	79.44 ± 1.70	83.20 ± 2.29