Circadian Rhythm Estimator Using Adaptive Notch Filter and Its Application to Locomotor Activity of Drosophila Melanogaster

The traditional method of circadian rhythm estimation is based on a double actogram that provides phase estimation on a daily basis, but not continuously and also fails to effectively filter noise. Our goal is to develop a circadian rhythm estimator that accurately and continuously estimates circadian phase while filtering noise. The estimator is a modified Adaptive Notch Filter (ANF), which is commonly used in signal processing. Drosophila Activity Monitor boxes house Drosophila fruit flies previously entrained to the same phase and contain sensors that measure the activity of flies. The actogram and ANF convert the raw data to a phase response curve (PRC), which describes the relationship between a stimuli (light) and circadian rhythm phase shift. White noise-corrupted data was also introduced to determine filtering effects. The PRC obtained from the actogram and ANF methods were nearly identical and also matched with literature values. With noise-corrupted data, the ANF PRC had an average error of 0.57 hours while the actogram PRC’s average error was 1.84 hours. The modified ANF is shown to be a valid and accurate method of modeling circadian rhythm continuously. The ANF was also able to successfully reject noise and the error against noise was very small compared to that of the actogram. Continuous circadian phase estimation allows more complex study of human circadian rhythm and further opens the possibility of personalized circadian rhythm estimation and light therapy.