Sebastian Cohn | MD-PhD Student in Biomedical Engineering
Computational methods for clinically meaningful interpretation of 4D Flow MRI.
Sebastian Cohn is an MD-PhD student at Northwestern University developing computational approaches for extracting structured, clinically meaningful information from difficult cardiovascular MRI data. His current work focuses on automated detection and quantification of hemodynamically active tears in chronic aortic dissection using 4D Flow MRI.
Current flagship project: a tear detector that localizes and quantifies true-lumen and false-lumen communication sites from time-resolved flow data.
Research Focus
Four ideas define the current work
The current work is easiest to understand through four linked ideas: an important cardiovascular problem, a technically difficult imaging modality, an original computational method, and a clear future trajectory toward structured intelligent interpretation.
Important health problem
Chronic aortic dissection remains dangerous because persistent true-lumen and false-lumen communication can sustain false-lumen pressurization, drive aneurysmal degeneration, increase rupture risk, and contribute to later death or major intervention.
Technically difficult modality
4D Flow MRI provides time-resolved, volumetric velocity fields in vivo, but the relevant signal is distributed across anatomy, direction, and the cardiac cycle.
Original computational contribution
The tear detector reframes dissection analysis as a localization-and-quantification problem, producing tear-level objects and flow descriptors rather than relying on manual plane placement and qualitative visualization alone.
Strong future trajectory
Validated tear-level outputs can be carried forward into agentic, cohort-scale interpretation and future multi-agent patient risk models.
Flagship Project
Flagship project: automated tear detection in chronic aortic dissection
This project ties together a life-threatening vascular disease, a technically difficult imaging modality, a novel computational detector, and structured outputs that can support future intelligent interpretation.
Why the method matters
The detector turns a difficult hemodynamic representation into standardized tear-level outputs with location, size, and exchange measurements.
That is the key research move: not simply showing complex flow, but extracting structured computational objects that can be interpreted clinically and carried into future models.
- Training context
- Northwestern MD-PhD
- Biomedical engineering with a cardiovascular MRI focus
- Current disease focus
- Chronic aortic dissection
- Persistent false-lumen perfusion through TL/FL communication
- Flagship method
- Automated tear detector
- Tear-level localization, size estimation, and flow readouts
- Future direction
- Agentic interpretation
- Structured tear outputs carried into cohort-scale and patient-level models
Input
4D Flow magnitude and velocity data
The detector begins with volumetric magnitude images and velocity fields before orthogonal-flow projection.
Representative 4D Flow input data showing magnitude and velocity-component views.
Detector output
Localized tear plus structured flow readout
The flagship contribution is the transformation from imaging signal to standardized tear-level measurement.
Representative detector output showing localized tear geometry with corresponding flow readout.