Fig. 2: Real-world wearable datasets show distinct data completeness profiles across infectious disease studies.
From: Overcoming Data Loss in Wearable Disease Detection with GAN-Based Imputation
a In the COVID-19 cohort, subjects with diagnostic results (black = negative, dark red = positive) generally have longer and more continuous heart rate (HR) recordings, with relatively few short gaps. b In contrast, the malaria cohort exhibits shorter overall recordings, more frequent and prolonged signal dropouts, and denser clusters of positive tests. These patterns highlight variability in data quality across study contexts and the importance of imputation strategies that can adapt to different sparsity profiles.
