A novel algorithm for detecting human circadian rhythms using a thoracic temperature sensor
Abstract
Circadian rhythms undergo high perturbations due to cancer progression and worsening of metabolic diseases. This paper proposes an original method for detecting such perturbations using a novel thoracic temperature sensor. Such an infrared sensor records the skin temperature every five minutes, although some data might be missing. In this pilot study, five control subjects were evaluated over four days of recordings. In order to overcome the problem of missing data, first four different interpolation methods were compared. Using interpolation helps covering the gaps and extending the recordings frequency, subsequently prolonging sensor battery life. Afterwards, a Cosinor model was proposed to characterize circadian rhythms, and extract relevant parameters, with their confidence limits. A divergence study is then performed to detect changes in these parameters. The results are promising, supporting the enlargement of the sample size and warranting further assessment in cancer patients.