. Author manuscript; available in PMC: 2021 Mar 10.

Published in final edited form as: Proc SPIE Int Soc Opt Eng. 2021 Feb 15;11597:115971F. doi: 10.1117/12.2580358

Table 1:

Performance of each individual neural network (AP-AUC, AP-PH, lateral-AUC, and lateral-PH) and the ensembled networks (averaging, grid search, and differential evolution directed optimization) in classifying the respective case based on the level of reperfusion as defined by the mTICI scale. Performance is displayed in the form of average accuracies, AUROCs, MCCs, sensitivities, and specificities along with their standard deviations and 95% confidence intervals. The best results are in bold.

Metric	Reperfusion level assessment: mTICI 0,1,2a versus mTICI 2b,2c,3
Metric	AP AUC	AP PH	Lateral AUC	Lateral PH	Averaged	Grid Weights	Differential Evolution Directed Optimization
Accuracy	74.4 ± 5.4 (72.0, 76.7)	74.2 ± 3.3 (72.8, 75.7)	74.9 ± 6.3 (72.2, 77.7)	76.9 ± 5.9 (74.4, 79.5)	78.3 ± 5.1 (76.1, 80.5)	83.0 ± 4.2 (81.2, 84.8)	82.7 ± 4.5 (80.7, 84.7)
AUROC	0.81 ± 0.05 (0.79, 0.83)	0.83 ± 0.04 (0.81, 0.84)	0.82 ± 0.05 (0.8, 0.84)	0.84 ± 0.05 (0.82, 0.87)	0.86 ± 0.04 (0.84, 0.88)	0.86 ± 0.05 (0.84, 0.88)	0.86 ± 0.05 (0.84, 0.88)
MCC	0.48 ± 0.11 (0.43, 0.53)	0.49 ± 0.07 (0.46, 0.52)	0.51 ± 0.12 (0.45, 0.56)	0.54 ± 0.11 (0.49, 0.59)	0.56 ± 0.10 (0.5, 0.61)	0.66 ± 0.08 (0.63, 0.70)	0.66 ± 0.09 (0.62, 0.70)
Sensitivity	0.78 ± 0.08 (0.75, 0.81)	0.78 ± 0.1 (0.74, 0.83)	0.78 ± 0.13 (0.72, 0.84)	0.84 ± 0.1 (0.79, 0.88)	0.83 ± 0.10 (0.80, 0.87)	0.90 ± 0.09 (0.87, 0.93)	0.89 ± 0.08 (0.86, 0.93)
Specificity	0.70 ± 0.10 (0.65, 0.74)	0.69 ± 0.12 (0.64, 0.75)	0.71 ± 0.13 (0.65, 0.77)	0.68 ± 0.14 (0.62, 0.75)	0.72 ± 0.10 (0.67, 0.76)	0.74 ± 0.09 (0.70, 0.78)	0.74 ± 0.09 (0.70, 0.78)