Background: Experimentally and clinically collected time series data are often contaminated with significant confounding noise, creating short, non-stationary time series. This noise, due to natural variability and measurement error, poses a challenge to conventional changepoint detection methods. Results: We proposed a novel, real-time changepoint detection method for effectively extracting important time points in non-stationary, noisy time series. We validated our method with three simulated time series, a widely benchmark data set, two geological time series, and two physiological data sets, and compared it to three existing methods. Our method demonstrated significantly improved performance over existing methods in several cases. Conclusion: Our method is able to effectively discern meaningful changes from system noise and natural fluctuations, accurately detecting change points. The ability of the method to extract meaningful change points with minimal user interaction opens new possibilities in clinical monitoring dealing with Big Data.