Adaptive circadian rhythm estimator and its application to locomotor activity

Disruption of the circadian rhythm is detrimental to human well being, with consequences ranging from lower productivity, sleep disorder, to more serious health problems. Accurate estimation of circadian rhythm is critical to the assessment and treatment of circadian disruption. Circadian estimate is also essential for light-based circadian entrainment. Direct measurements of circadian rhythm markers such as dim light melatonin onset are inconvenient and acquired at best at low rate. Wearable continuous measurements such as actigraph and body temperature are convenient but masked by many other factors. In this paper, we present a new circadian rhythm estimation scheme based on adaptive notch filter (ANF) which is commonly used in signal processing. The ANF is designed to track the gain and phase of a single sinusoid from noisy data. We extend the classic ANF to multiple harmonics with non-zero bias needed in circadian rhythm tracking. The modified ANF is tested on rat and Drosophila locomotor activity data to extract circadian argument. The estimation results are compared with linear regression and gliding cosinor. The ability to generate circadian estimate opens up the tantalizing possibility of personalized circadian rhythm estimator and light therapy.