Table 2.

Summary of performance metrics of all configurations

Input images	Id	Binarizer	Accuracy	Sensitivity	Specificity	AUC
Aligned T1w	A1	None	71.12 ± 5.01% [61.34%, 82.52%]	67.47 ± 9.90% [51.66%, 85.94%]	74.76 ± 7.07% [62.03%, 85.45%]	0.71 ± 0.05 [0.62, 0.83]
	B1	13.75%	62.51 ± 5.45% [53.39%, 72.09%]	62.26 ± 9.48% [47.70%, 84.93%]	62.79 ± 8.35% [46.56%, 74.42%]	0.63 ± 0.054 [0.53, 0.72]
	C1	27.50%	72.74 ± 5.49% [61.41%, 82.11%]	71.37 ± 9.10% [56.86%, 88.87%]	74.15 ± 9.94% [51.67%, 88.37%]	0.73 ± 0.055 [0.61, 0.82]
	D1	41.25%	77.95 ± 4.57% [70.90%, 86.34%]	76.74 ± 9.41% [60.10%, 94.56%]	79.15 ± 6.80% [64.54%, 89.56%]	0.78 ± 0.045 [0.71, 0.86]
Skull-stripped T1w	A2	None	81.63 ± 3.77% [74.36%, 88.01%]	81.22 ± 6.94% [69.59%, 93.15%]	82.11 ± 7.92% [65.50%, 93.60%]	0.82 ± 0.037 [0.74, 0.88]
	B2	13.75%	78.12 ± 4.63% [70.79%, 85.79%]	76.83 ± 7.03% [62.65%, 87.05%]	79.40 ± 6.76% [65.53%, 89.91%]	0.78 ± 0.046 [0.71, 0.86]
	C2	27.50%	79.57 ± 3.92% [73.46%, 86.45%]	78.32 ± 7.74% [66.79%, 93.87%]	80.92 ± 7.71% [67.53%, 91.95%]	0.80 ± 0.039 [0.74, 0.86]
	D2	41.25%	81.56 ± 4.63% [72.31%, 88.67%]	79.69 ± 9.42% [62.59%, 96.48%]	83.50 ± 6.77% [72.48%, 96.29%]	0.82 ± 0.046 [0.72, 0.89]

Column Id refers to the preprocessing defined in Fig. 1

Values between brackets show the 95% confidence interval

AUC area under receiver operating characteristic curve

Values in bold highlight best result for metric