Table 1.

EMA performance with feature selection.

	Mean PMA	SVR	SVR with FS
MSE (days2)	438.0	97.3*	108.2*
r	0	0.936	0.940
SD (days)	20.94	9.86	10.30
SE (%)	9.72	4.51	4.75
Bias (days)	0	−0.45	−1.33
% EMA (1 week)	20.8	50.5	43.6
% EMA (2 weeks)	39.6	82.2	85.2

Mean PMA is the worst case/best guess EMA estimate. ^*Is a significant reduction in the MSE compared to the mean PMA (p < 0.001), SVR is support vector regression, FS is feature selection, PMA is post-menstrual age, EMA is EEG maturational age, SD is the standard deviation and SE is the percentage error and %EMA denotes the percentage of EEG recordings where the EMA was within 1 week and 2 weeks of the PMA and MSE is the mean square error estimated on a per recording basis (n = 101). The adjusted correlation coefficient, r, is estimated from a linear mixed model: the slope of EMA over time was a fixed effect and the intercept was a random effect per infant (n = 39).