Table 2.

Areas under the receiver-operating characteristic curves (AUCs) and cutoff values for assessing pain in the spectrogram-convolutional neural network index and surgical pleth index (SPI) in surgical patients.

Parameter	Spectrogram–CNNa index			SPI
	Training set	Validation set	Test set
AUC (95% CI)	0.992 (0.991-0.993)	0.932 (0.921-0.942)	0.757 (0.746-0.768)	0.659 (0.646-0.671)
P value	<.01	<.01	<.01	<.01
Cutoff valueb	50	54	48	44
Sensitivity, %, specificity, %	94.6, 96.4	83.1, 88.2	68.1, 73.8	65.2, 59.5
Balanced accuracy, %c	95.5	85.7	71.0	62.4

^aCNN: convolutional neural network.

^bCutoff values used for the calculation of sensitivity and specificity were calculated as “best fit” (highest combined sensitivity and specificity).

^cBalanced accuracy is the corrected accuracy of the imbalance of a class set, calculated as (sensitivity + specificity)/2.