Selen Bozkurt
Emory University, GA·Medicine & Clinical Research
Hospitals record a huge amount about their patients, but much of it is written as free-text notes that computers cannot easily use. Important facts are buried in those notes.
- Fields
Medicine & Clinical Research
Computer Science
Public Health & Epidemiology
- Topics
clinical informatics
biomedical informatics
electronic health records
- Grants
- 2 known grants, through May 2030
What they work on
- Translate clinical trial protocol eligibility text into computable phenotypes and query packages tied to standard vocabularies, and validate them against expert references and prescreening logs.
- Build modular large language model pipelines with schema-constrained output and quantified uncertainty to extract criteria such as cancer progression, performance status and biomarkers from clinical notes.
- Apply the ECLIPSE framework to trials in kidney cancer, heart failure and opioid use disorder at three health systems, and measure cohort overlap, effect-estimate stability in target-trial emulations, and time saved.
- Develop and validate an LLM-based tool that detects language signaling loneliness and social isolation in the clinical notes of adults aged 55 and older.
- Integrate loneliness and isolation markers into a late-life suicide risk prediction model, and evaluate the model for bias and interpretability.
More about the research
These include whether a cancer has come back, how well a patient functions day to day, and signs that an older adult is lonely. Bozkurt's work builds artificial intelligence (AI) tools, including large language models (LLMs) that read and interpret text, to turn electronic health records (EHRs) into reliable data. One project uses that data to judge clinical trial eligibility and to emulate trials. Another uses it to predict suicide risk in older adults.
Why it matters
Making trial eligibility computable could let health systems check faster whether a trial can find enough patients. It could also test whether trial results apply to real-world patients in kidney cancer, heart failure and opioid use disorder, and help make trial populations more representative. Detecting loneliness and isolation in clinical notes could improve suicide risk prediction for older adults. This may enable earlier support during the healthcare visits that often come before a suicide.
Known grants
From NIH and NSF research grants. Labs often have other funding we can't see.
Closest labs
Allison Summers
Oregon Health & Science University, OR·Medicine & Clinical Research
Develop large language model pipelines that extract amblyopia treatment, adherence and visual outcome information from free-text ophthalmology clinical notes, where these concepts are documented inconsistently and not in structured EHR fields.
Yili Zhang
Georgetown University, DC·Computer Science
Develop a large language model–based natural language processing pipeline that extracts immune-related adverse event evidence, type and grade from de-identified clinical notes, creating a reusable annotated corpus.
Danielle Haley
Boston University, MA·Public Health & Epidemiology
Determine whether cannabis use disorder is associated with new falls, fractures, delirium and dementia, and with HIV viral suppression and antiretroviral therapy adherence, in about 180,000 Veterans Aging Cohort Study patients.
The school
Emory University: 70 labs on PI Pages, $1.2B a year on research, 286 research doctorates a year.
Reach out
A start for your email. Fill in the brackets.