Research
Research organized around clinically meaningful computation.
The overarching goal is to extract structured, mechanistically relevant information from complex 4D Flow MRI data and carry it forward into more informative models of cardiovascular disease activity.
Research Thesis
Computational methods should make difficult flow data more interpretable, more reproducible, and more clinically useful, not less.
See the flagship tear detector project4D Flow MRI
Time-resolved flow structure
Representative streamline visualization from 4D Flow MRI.
Research Snapshot
- Current disease focus
- Aortic dissection
- Especially chronic dissection with persistent false-lumen perfusion
- Core modality
- 4D Flow MRI
- Time-resolved, volumetric hemodynamic imaging
- Computational emphasis
- Tear-level structure
- Detection, quantification, and downstream aggregation
Current Arc
Current work follows a three-stage trajectory
The current framing is not a generic imaging workflow. It is a progression from detector development to cohort translation to multi-agent modeling.
Aim 1
Build and validate automated tear detection
Develop a reproducible detector that localizes hemodynamically active true-lumen and false-lumen communication sites and returns interpretable tear-level descriptors from 4D Flow MRI.
Aim 2
Translate tear-level readouts across cohorts
Apply the pipeline across dissection cohorts to study how tear count, location, and exchange severity relate to remodeling and clinically important outcomes beyond anatomy alone.
Aim 3
Move toward multi-agent patient modeling
Use validated tear outputs as structured building blocks in a future patient-level model that aggregates tear-specific, anatomic, and clinical information into risk scoring.
Flagship
The tear detector page is the clearest place to see the full research story
It combines the disease framing, data challenge, algorithmic contribution, and future direction in one focused narrative.
Automated tear detection for chronic aortic dissection
The project microsite explains the clinical problem, why 4D Flow MRI is technically difficult, how the detector works, what its outputs look like, how evaluation is structured, and how the work can grow into cohort-scale agentic analysis.