On October 15 at 1:00 p.m. in Room 320, Amon Ties Uerckwitz will defend his master’s thesis on “Design and Implementation of a Camera-Based Method for Noninvasive Dehydration Detection in Office Workers.” The examiners are Prof. Michael Fellmann and Erik Endlicher from the Chair of Business Informatics.
Abstract
A mild fluid deficit of one to two percent of body weight is common among office workers by the afternoon and, even at this level, impairs sustained attention, working memory, and psychomotor speed. Thirst follows the deficit with a delay, and every objective measure in common use today requires a urine sample, a skin electrode, or a wearable device. This study investigates whether the webcam, which is already present in the workplace, can instead serve as a passive indicator. The physiological target variable is cardiovascular compensation: as plasma volume decreases, the resting heart rate increases. The hypothesis tested is therefore that a person’s resting heart rate, as measured by camera, is higher after several hours without fluid intake than after normal intake. Remote photoplethysmography (rPPG) measures heart rate noninvasively based on pulse-synchronous color changes in the facial skin. A systematic literature search revealed no previous study applying this method to hydration. A complete processing pipeline was implemented and tested in four stages. On a public reference dataset with synchronously recorded contact PPG, the mean absolute error is 4.31 BPM across n = 1,290 paired observations and 1.8 BPM in the seated resting condition, which corresponds to the intended application. The hydration issue was then addressed using a crossover protocol with repeated measurements within each subject: 10 subjects were alternately hydrated and dehydrated, with recordings taken in the morning and afternoon, for a total of 40 sessions in which the time of day and day of recording varied within each subject. Among the 7 participants with a regular office workday, the resting heart rate during the dehydrated session was consistently higher than during the hydrated session—by an average of 4.16 BPM—in the direction and magnitude predicted by physiology. Among the 3 individuals without such a workday, the opposite was observed: the resting heart rate was lower in the dehydrated session in every case, a difference of −5.45 BPM. The pooled value of 1.28 BPM thus represents a balance between two opposing groups and not an absence of the effect in either group. Three limitations constrain this result. The pre-registered dose-response relationship with the time elapsed since the last fluid intake does not hold across the entire cohort (r = 0.09); the subgroup showing the effect was formed only after the data were reviewed; and the remaining dose-response relationship survives adjustment for eyelid opening—a measure of fatigue—only within the target population, where it is halved rather than eliminated. The finding is therefore a directional hypothesis, limited to the population specified in the title, and not a confirmed effect. Two metrological findings further limit any such claim. Heart rate variability cannot be measured at the achievable signal-to-noise ratio; the corresponding pre-registered feature was withdrawn before the replication cohort was enrolled. Day-to-day physiological variation—not camera error—sets the resolution limit, which is why adding more participants provides greater statistical power than longer recording times. The study thus provides a validated measurement tool and a quantified statement regarding its resolution, but no validated measurement of hydration. Any camera-based statement regarding hydration status must be based on repeated recordings of an individual under consistent optical conditions, relative to that individual’s own history.