Adaptation of the Hierarchical Factor Segmentation method to noisy activity data

ABSTRACT

The Hierarchical Factor Segmentation (HFS) method is a non-parametric statistical method for detection of the phase of a biological rhythm shown in an actogram. The detection accuracy of this method was measured on actograms showing only circadian rhythms with a constant ratio of signal to noise (S/N). In the present study, we generated 84 types of artificial actograms including circadian or circatidal rhythms by using three parameters: α/ρ, S/N and period length τ, and evaluated the effectiveness of our devised adaptation of the HFS method, the cycle-by-cycle adaptation. The results showed the effectiveness of the cycle-by-cycle adaptation was high even though S/N or τ was fluctuating through a whole actogram. These suggested that the cycle-by-cycle adaptation could be effectively applied to various kinds of rhythmic activity data. The C++ source code of the cycle-by-cycle adaptation is available on the website at https://github.com/KazukiSakura/cHFS.git.

KEYWORDS:

• Hierarchical Factor Segmentation method
• statistical method
• phase
• actogram
• circadian rhythm
• circatidal rhythm

Acknowledgments

We thank Hideharu Numata for critical reading of the manuscript and Elizabeth Nakajima for linguistic corrections.

Supplementary Material

Supplemental data for this article can be accessed here.