Subjects
Six healthy infants (three females and three males), aged 12 months (range 11–13 months), participated in this study. All parents of the subjects were carefully instructed about the study, and all gave their written informed consent. The study was approved by the Tokyo-Kasei University Institutional Review Board.
Heart rate and 3-axis acceleration were continuously and simultaneously monitored using the wearable heart rate sensor “myBeat” (UNION TOOL CO., Tokyo, Japan). The device was placed on each subject’s left chest with disposable electrode (Vitrode. T-50, Nihon Kohden Co., Tokyo, Japan). Measurement was conducted at subjects’ homes under free-living conditions and started before bedtime and finished when the infants woke up in the morning. The parents were instructed to write down bedtime and waking times.
Spectral analysis of HRV
The RR intervals and acceleration signals were recorded in the device memory and transferred to a computer for analysis. The instantaneous heart rate was calculated from the RR interval sequences. False-negative or false-positive RR intervals were excluded from the analysis. An automated computer analysis system was developed using Microsoft Visual Basic for Applications (VBA), and the Least Square Cosine Spectrum Method was applied with a window range of 30 s of the heartbeat time series. The formula of the cosine curve is as follows:
$$ Y=M+A\cos \left(2\pi t/\omega -\theta \right) $$
(1)
where Y is the estimated cosine curve, M is the MESOR (Midline Estimating Statistic Of Rhythm), A is amplitude (a measure of half the extent of predictable variation within a cycle), t is time, ω is period (duration of one cycle), and θ is acrophase (a measure of the time of overall high values recurring in each cycle).
By changing the period every 0.1 s sequentially, a high-frequency component of HRV was extracted. The cosine curve that showed the minimum value of the residual sum of squares calculated by subtracting the estimated cosine curve from the original data was referred to as the best-fit cosine curve. The series of the probability in each given cycle was calculated using the direct method by Sasaki [13, 14]. Then, the reciprocal logarithm of the probability was calculated and defined as the RA (RA-COSPEC: Respiratory Area obtained by COSPEC), which is expressed as:
$$ \mathrm{RA}=\log \left(1/\mathrm{Probability}\right) $$
(2)
The value of RA is higher when a cosine curve is fitted to the original data series.
In addition, we analyzed fitting curve for the time sequential changes of RA by using the Least Square Cosine Spectrum Method when the period was changing from 30 to 120 min. We calculated the probability of the fitness of cosine curve directly [13, 14].
Analysis of triaxial acceleration
To determine the overall magnitude of physical activity, the vector magnitude (G) was calculated by taking the square root of the sum of squares from each axis:
$$ G=\sqrt{\left({X}^2+{Y}^2+{Z}^2\right)} $$
(3)
where G = 9.8 m/s2, the X-axis of triaxial acceleration reads right or left, the Y-axis reads up or down, and the Z-axis reads back or front movement. The sleeping position was estimated as in our previous study [15]: when the Y-axis value was more than −0.7, the subject was considered to be in the lying position. In addition, the left lateral position was defined as an X-axis value ≥ 0.5 G, the right lateral position was defined as an X-axis value < − 0.5 G, the supine position was defined as a Z-axis value ≥ 0.7 G, and the prone position was defined as a Z-axis value < − 0.7 G.
Statistical analysis
The t test was used to examine the differences. P value < 0.05 was considered statistically significant.