Related Experiment Video
Updated: Aug 3, 2025
04:16
Author Spotlight: Advancing Lung Transplant Immunology Through Intravital Imaging
Published on: April 19, 2024
1.1K
CT-based Machine Learning for Donor Lung Screening Prior to Transplantation.
Sundaresh Ram1,2, Stijn E Verleden3,4,5, Madhav Kumar2
1Department of Radiology, University of Michigan, Ann Arbor, MI, United States.
Medrxiv : the Preprint Server for Health Sciences
|April 10, 2023
Summary
A new machine learning algorithm uses ex vivo CT scans to objectively screen donor lungs, improving transplant success. This tool identifies lungs at higher risk for complications like chronic lung allograft dysfunction (CLAD).
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Pulmonary Medicine
Background:
- Donor lung assessment is subjective and lacks standardized criteria.
- Current donor lung selection methods are not optimized for the evolving donor pool.
- Objective screening tools are needed to improve lung transplant outcomes.
Approach:
- A prospective clinical trial collected CT scans and clinical data from 100 donor lung cases.
- A supervised machine learning algorithm (dictionary learning) was trained on CT images to classify lung suitability.
- The algorithm identified specific image patterns indicative of pulmonary abnormalities.
Key Points:
- The machine learning algorithm detected pulmonary abnormalities on ex vivo CT scans.
- It identified donor lungs associated with increased risk of prolonged ICU stay.
- The algorithm flagged lungs linked to a 19-fold higher risk of developing chronic lung allograft dysfunction (CLAD) within two years.
Conclusions:
- A CT-based machine learning strategy enables objective ex vivo donor lung screening.
- This approach assists physicians in identifying high-risk lungs and recipients.
- Objective screening is crucial for mitigating post-transplant complications with increasing use of suboptimal lungs.
