Overcoming Data Loss in Wearable Disease Detection with GAN-Based Imputation
Read the study design and limitations alongside its findings.
Read the limitations ↓Summary
Summary
Studies generative reconstruction of missing wearable heart-rate data and its potential to support disease-signal detection.
The summary describes the published study and the sources listed below.
Scientific notes: 2026-11-11 · Research connections: 12 November 2026Key findings
Key findings
Generative imputation recovered additional physiological signals in incomplete wearable records; clinical effectiveness remains untested.
Study design & setting
Study design and methods
Population, setting or evidence base
Retrospective COVID-19-to-malaria transfer study; Kenya: 300 enrolled, 290 usable; external COVID-19 subset: 2,124. Overall monitoring duration unspecified.
Methods
COVID-trained GAN imputation followed by retrospective anomaly detection.
Limits to interpretation
Limitations
No prospective clinical evaluation; Imputation can suppress real anomalies; Three-week RDT timing and mixed evaluation units complicate accuracy.
Research relevance
Research relevance
Informs data-quality research, not a validated diagnostic service.
People, places & projects
Research connections
Labels identify published coauthors, study settings, planned protocol settings and wider methods or network connections.
Coauthors with research profiles
Sites & research settings
Projects & methods connections
Published sources & review scope
Sources
This page summarises the publication. Sources and editorial scope ↗