Scientific Publications & Research Highlights

State-of-the-art medical imaging benchmark to comprehensively assess multi-modal capabilities of foundation models in radiology

Machine learning based pipeline to accurately predict mortality in heart failure patient cohorts.

Review covers advances and limits of generative AI for biomedical time-series data.

Model predicts patient-specific future MRI for Alzheimers over multi-year horizons.

Selective inference allows valid statistical testing for generative AI-based anomaly detection in Alzheimers.

TFT-multi predicts multiple vital sign trajectories in ICU, leveraging correlations for accuracy.

VISTA improves SV detection accuracy robust across organisms/coverages with precision/recall tuning.

Systematic review of packaging/containerization for omics tools. Highlights usability/reproducibility advances.

Covers ML for genomics, microbiome, and EHR prediction model fairness and translational impact.

Developed models using UCLA EHR data to predict perinatal depression in 18,000 pregnancies, assessing fairness across ethnoracial groups and social determinants.

Benchmarks reproducibility of WGS mapping tools. Shows mapping accuracy variation and key pipeline advice.

ML modeling to predict urosepsis risk using UCLA EHR data. We show that Random forest outperforms baseline APR.