NA-MIC Project Weeks
Back to Projects List
New Slicer Module for Visual Assessment of Pulmonary Congestion from Ultrasound
Key Investigators
- Mike Jin (Centaur Labs, Brigham and Women's Hospital, USA)
- Tamas Ungi (Queens's University, Canada)
- Fahimeh Fooladgar (University of British Columbia, Canada)
- Tina Kapur (Brigham and Women's Hospital, USA)
Project Description
This work is part of an NIH Trailblazer R21 grant to our team to develop and validate computational methods for quantifying pulmonary congestion using B-lines in heart failure patients from bedside lung ultrasound in emergency settings. Tools for automated quantification could help emergency department physicians more rapidly and frequently examine patients to assess progress and adjust treatment, resulting in improved care and patient outcomes.
Objective
- Add a new public module for annotation of pulmonary congestion in ultrasound
- Add new feature to existing public Anonymizer module: AI-assisted detection of image fan boundaries in ultrasound to streamline anonymization, followed by OCR in output which produces warning if any text is detected in image
Approach and Plan
We will spend Project Week developing the software to support these features and hopefully release the modules publicly.
Progress and Next Steps
- We have added a new AnnotateUltrasound module to the Ultrasound extension that allows for easy annotation of sectors representing B-lines (indicating pulmonary congestion).
-
As part of the public Anonymizer module, we have added a new button that uses an AI model to auto-detect the boundaries of the ultrasound image fan.
Next steps: we will work on adding OCR text detection to add an additional check that anonymized images don’t contain any remaining PHI text prior to export.
Illustrations
Background and References
- Asgari-Targhi et al. (2024). Can Crowdsourced Annotations Improve AI-based Congestion Scoring For Bedside Lung Ultrasound? MICCAI 2024. (link to paper)