Mei Liu
University of Florida, FL·Medicine & Clinical Research
Acute kidney injury (AKI) is a sudden loss of kidney function. It affects 10–15% of hospitalized patients and more than half of critically ill patients.
- Fields
Medicine & Clinical Research
Computer Science
Public Health & Epidemiology
- Topics
acute kidney injury
electronic health records
machine learning
- Grants
- 2 known grants, through Mar 2027
What they work on
- Develop methods to build the best individualized cohort of similar past patients from EHR data, as the training set for a personalized AKI risk model for each new hospital patient.
- Design personalized transfer learning techniques that avoid negative knowledge transfer and deliver equitable AKI prediction across heterogeneous patient subgroups.
- Create transfer learning solutions for three types of non-matching EHR data across health systems, using records of more than 21 million patients from 12 healthcare systems.
- Build a privacy-preserving federated learning framework with dynamic weighting of each site's contribution to train a robust, interpretable gradient-boosting model for early AKI prediction.
More about the research
Patients with AKI are 6.5 times more likely to die in the hospital. They also face a higher risk of chronic kidney disease and heart disease. Spotting risk early allows doctors to act before damage occurs. Machine learning on electronic health records (EHR) can help, but AKI has many causes and courses, and data from one hospital often differ from data at another. Both projects build models that predict AKI risk while handling these differences, across individual patients and across hospitals.
Why it matters
More reliable early warnings of acute kidney injury could help hospital teams start prevention for high-risk patients, including critically ill patients and subgroups that standard models misjudge. Privacy-preserving methods would let many health systems learn from each other's data without sharing records. Interpretable predictors could guide prevention and management plans. The methods could extend to other diseases.
Known grants
From NIH and NSF research grants. Labs often have other funding we can't see.
Closest labs
Javier Neyra
University of Alabama at Birmingham, AL·Medicine & Clinical Research
Develop and validate deep learning models that predict survival without ongoing dialysis at clinically actionable time points for ICU patients on continuous renal replacement therapy, using EHR and CRRT machine time-series data from 9 institutions.
Andrew Bishara
University of California-San Francisco, CA·Medicine & Clinical Research·Early-career
Implement machine learning models that predict perioperative acute kidney injury in real time during surgery and alert clinicians while kidney-protective treatments can still be given.
Tianxi Cai
Harvard University, MA·Mathematics & Statistics
Develop semi-supervised risk prediction methods that combine noisy electronic health record surrogates, such as billing codes, with a small set of chart-reviewed outcome labels, even when predictors are numerous.
The school
University of Florida: 65 labs on PI Pages, $1.3B a year on research, 795 research doctorates a year.
Reach out
A start for your email. Fill in the brackets.