Exploring missing data prediction in medical monitoring: A performance analysis approach

Medical monitoring represents one of the most critical components in existing healthcare system. The accurate and reliable acquisition of various physiological data can help physicians and patients to properly detect and identify potential health risks. However, this process suffers from severe limitations in terms of missing or degraded data, which may lead to a rather high false alarm rate and potentially compromised diagnostic results. In this paper, we investigated three different approaches for missing data prediction in clinical settings: mean imputation, Gaussian Process Regression (GPR), and Kalman Filter (KF). Experimental results show that, the heart rate (HR) signals largely rely on most recent data and missing data prediction will be less effective for further prediction.