Methodological study demonstrates an iterative smoothing algorithm for hormone time series, suggesting reliable identification of ultradian pulses and overlapping secretory episodes.
Key Points
To develop a standardized algorithmic procedure for identifying ultradian pulsations and resolving overlapping secretory episodes in hormonal time series.
Applied a robust smoothing technique to remove long-term baseline trends (time constants <6–12 h) and generate a residual time series.
Rescaled residuals by assay standard deviation and identified peaks exceeding duration-dependent cutoff thresholds [G(n)].
Downweighted detected peaks iteratively during subsequent smoothing passes and evaluated resolved peaks for overlapping sub-episodes before calculating peak frequency and amplitude.
Established an iterative algorithm that successfully isolates both narrow, high-amplitude peaks and broader, lower-amplitude secretory episodes.
Integrated deconvolution steps to separate closely spaced, overlapping secretory bursts without distortion from baseline diurnal drifts.
Implemented the mathematical framework into functional computer programs to automate pulse detection and extract frequency and amplitude statistics.