Building a Smarter Cancer Center
We develop and deploy artificial intelligence from large language models to wearable sensors to improve how we take care of patients with cancer.
Turning research into care
Our work moves from publications into clinics — catching complications before they happen, helping teams respond to patient needs faster, guiding treatment decisions, and improving outcomes.
Watching out for patients
Artificial intelligence and wearable sensors identify symptoms before they become complications so care teams can intervene early. Our WATCH trial now evaluates these models in patients undergoing cancer therapy.
AI that makes data usable
Large language models synthesize the expanse of data in cancer care to empower clinicians in caring for the patient in front of them and drive the next generation of cancer research.
Leaders who carry the mission
Lab alumni now lead their own programs across academia and industry. Our trainees go on to push for progress in cancer care everywhere.
Multiple threads, one goal
Our work spans methods and clinical domains, connected by a single question: how can data and AI make cancer care better?
Predicting & Preventing Crises
SHIELD-RT, among the first randomized controlled studies of healthcare machine learning, directed supportive care using electronic health record-based prediction of emergency visits and hospitalizations, demonstrating reductions in acute care requirements and costs (NEJM AI 2024; JCO 2020). Ongoing NIH R01 extends this nationally in the VISION study.
Wearables & Symptom Science
One of the first computational models to predict hospitalizations from wearable step-count data (JAMA Oncol 2024), now implemented under study in the open WATCH trial and the NRG Oncology NRGF-001 cooperative-group trial.
LLMs for Population Oncology
Reviews and tooling for using large language models to collect, aggregate, and analyze cancer data at scale, plus LLM-identified symptom trajectories for quality-of-life outcomes in the BEAQON study (NIH R01 supplement).
Tools for Patients
Applying LLMs and NLP to reach patients, understanding patient portal messages and partnering with patient advocacy helping patients understand their treatment decisions in the open COUNSEL trial.
AI That Doesn't Break
SHIFT and related frameworks (ICML; PCORI-funded) detect when clinical AI degrades as data changes and how to intervene, including calibrating models with social and clinical context (Convergence Award).
Precision Prostate Cancer Care
PSMA PET-guided management of oligometastatic prostate cancer (JNCI Spectrum), accelerating clinical-trial readouts in metastatic prostate disease.
The Hong Lab is a part of
Weill Cancer Hub West
Latest from the lab
WATCH trial is open
Wearable Activity Tracking to Curb Hospitalizations evaluates prediction models for unplanned acute care during cancer therapy.
Weill Cancer Hub West announced
UCSF and Stanford unite with IMPACT-AI; Julian joins as UCSF Co-PI. Coverage in the SF Examiner.
LLMs meet the patient portal
Ji Hyun Chang leads our team applying multiple LLMs to understand what patients message about — paving the way for AI-assisted triage.