Table 1.

Summary of published machine learning systems for prediction of apnoeic events in preterm infants.

Author	Sample size, gestational age (weeks), Birthweight (g)	Study duration (mins), No. of apnoeic event	Respiratory detection	Input features	Pre-processing technique	Machine learning technique	Pre-apnoea window (mins)	Evaluation
Williamson et al. (15)	6, 28.6, 1,240	2015,34	Abdominal respiratory inductance plethysmography	Inter-breath intervalR-R interval	Linear interpolation Log transformation Conversion to standard units for each patient	Equal prior quadratic classifier	7.5	*AUC = 0.73, p = 0.13
Williamson et al. (14)	6, 28.6, 1,240	2015,34	Abdominal respiratory inductance plethysmography	Inter-breath intervalR-R intervalMovement features	Linear interpolation Log transformation Conversion to standard units for each patient	Bayesian adaptation of Gaussian mixture models	7.5	*AUC = 0.80, p = 0.00
Shirwaikar et al. (16)	299, ND, ND	ND,ND	Not stated	23 features recorded including demographic, maternal co-variates and physiological input	Transformation Normalization	Decision tree (C5.0)	ND	Accuracy = 0.75, Sensitivity = 0.20, Specificity = 0.88
						Support vector machine using radial kernel		Accuracy = 0.75, Sensitivity = 0.28, Specificity = 0.72
						Ensemble approach: ∙ Bagged decision tree ∙ Auto tuned boosted C5.0 ∙ Random forest		Accuracy = 0.88, Kappa = 0.72

Gestational age and birthweight are presented as mean only.

^*Area under the receiver operating characteristic curve; AUC, Area under curve.

ND, Not described.