Research agenda

Reliable and Fair Medical AI Across Clinical Settings

I study how medical foundation models can be audited, monitored, and responsibly deployed across clinical settings—knowing when, where, and for whom a model can be trusted.

Research Focus

01

Reliability auditing

Evaluating medical vision-language and foundation models across public clinical domains, findings, and deployment contexts. My work examines localized reliability, calibration, high-confidence errors, and the limits of pooled performance.

Medical foundation modelsCalibrationClinical shift
02

Dual fairness in medical imaging

Studying fairness at two levels in federated medical imaging: collaboration fairness across institutions and group fairness across patient populations. This line builds on my systematic review of the dual-fairness lens.

Federated learningGroup fairnessInstitutional fairness
03

Context-aware deployment

Developing post-hoc monitoring, context-aware calibration, and fair deferral strategies that act on unreliability. The long-term goal is to support safer decisions without requiring large-scale model pretraining.

MonitoringFair deferralDeployment

Interactive Research Map

Drag a node, hover to see its connections, or click it to explore how my research themes, methods, and projects fit together.

Core theme Research theme Method / setting Project / paper

Current Research Programme

My current programme connects a fairness foundation, a localized reliability audit, and future methods for acting fairly on unreliability.

  1. Fairness foundationDual-fairness review in federated medical imaging.
  2. Reliability auditHD-Cal: few-shot reliability estimation for cross-domain hallucination heterogeneity in chest X-ray vision-language models.
  3. Act fairly on unreliabilityContext-adjusted fair deferral and label-free subgroup reliability monitoring as future directions.

Previous Work

My earlier work includes lightweight medical image segmentation and deployable medical AI, providing a technical foundation for my current research on trustworthy clinical AI.