Table 2.

Perfusion estimates for the eight testing data by the different methods

	F0	F1	N1	N2	N3	N4
Mean	103.8	102.3	102	90	100	109
SD	81.8	74.7	80	62	78	77
Median	72.4	78.6	70	80	70	80
		F1‐F0	N1‐F0	N2‐F0	N3‐F0	N4‐F0
Mean	‐	−1.5	−1.5	−13.8	−3.6	3.7
RMSE	‐	37.5	26.0	43.8	27.0	33.5
MAE	‐	22.3	15.9 &	24.9 #	15.8 &	23.5 #
Correlation	‐	0.889	0.949	0.850	0.944	0.912

Abbreviations: F₀, values from multigrid model fitting; F₁, regular model fitting; MAE, mean absolute error; N₁, 20 diverse data; N_1‐4, estimates from NN trained by different strategies; N₂, 20 homogeneous data; N₃, 40 combined data; N₄, 20 diverse data with regular fit as reference; RMSE, root means square error; SD, standard deviation. Superscript “&” indicates that MAE was significantly lower than that of F₁. Superscript “#” indicates that MAE was significantly higher than that of N₁. Unite for all values except for correlation: ml min⁻¹ 100 g^−1.