Continuous imputation of blood gas and metabolic panel laboratory values in the intensive care unit using machine learning
Behrooz Mamandipoor, Martin Krause, Pragnya Korti, Chun-Nan Hsu, Rodney A. Gabriel · 2026
Read Continuous imputation of blood gas and metabolic panel laboratory values in the intensive care unit using machine learning by Behrooz Mamandipoor, Martin Krause, Pragnya Korti, Chun-Nan Hsu, Rodney A. Gabriel (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/PMC13562130.txt.
Description
Behrooz Mamandipoor and four colleagues teach machine learning models to fill in hourly blood gas and metabolic panel values for intensive care patients between real laboratory draws. Their 2026 study feeds prior labs, medications, clinical events, fluid balance, hemodynamics, and ventilator settings into the estimators. A downstream task, predicting who will need kidney replacement therapy, tests whether the filled values help.
The gradient boosting estimator cut imputation error by about a third for blood gas targets and about a fifth for metabolic ones, against simply carrying the last value forward. A bidirectional memory network fed the machine-filled series predicted dialysis need better than the same network fed carried-forward values, on both discrimination measures reported. The authors argue continuous imputation turns sparse lab data into a live signal for deep learning at the bedside.
Catalog record
| CREATOR | Behrooz Mamandipoor, Martin Krause, Pragnya Korti, Chun-Nan Hsu, Rodney A. Gabriel |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 44596 |
| LICENSE | https://creativecommons.org/licenses/by/4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |