Related Experiment Video
Updated: Jan 9, 2026
09:39
A Microfluidic Model of Biomimetically Breathing Pulmonary Acinar Airways
Published on: May 9, 2016
8.3K
Artificial intelligence-assisted multiscale lung modeling to predict alveolar septal wall stress
Sunder Neelakantan1, Mostafa Ismail2, Nikhil Kadivar3
1Department of Biomedical Engineering, Texas A&M University, College Station, TX, USA.
Acta Biomaterialia
|November 29, 2025
Summary
This study introduces a machine learning method to estimate lung tissue stress, aiding early diagnosis of radiation-induced lung injury (RILI) and fibrosis. The approach improves accuracy by using synthetic data, revealing biomechanical changes indicative of lung damage.
Area of Science:
- Biomedical Engineering
- Computational Biology
- Pulmonary Medicine
Background:
- Lung injuries like radiation-induced lung injury (RILI) cause ventilation issues and reduced function.
- The biomechanical factors driving lung injury progression are not well understood.
Purpose of the Study:
- To develop a machine learning-based method for estimating in-vivo alveolar septal wall stress.
- To understand the role of parenchymal biomechanics in lung injury onset and progression.
Main Methods:
- Reconstructed lung tissue elements from X-ray microtomography.
- Utilized a generative adversarial network (GAN) to create synthetic tissue data.
- Trained an artificial neural network (ANN) using finite element simulations on real and synthetic data.
Main Results:
- ANN accuracy improved from 61.6% to 84.0% with synthetic data.
- Observed reduced stress (pneumonitis) at 3 months and elevated stress (fibrosis) at 5 months post-radiation in a rodent model.
- Detected stress heterogeneity indicating volutrauma at 5 months.
Conclusions:
- Regional biomechanical markers can aid early diagnosis of subclinical lung injuries.
- The ML method accurately captures pneumonitis and fibrosis in RILI rodent models.
- This approach offers insights into lung injury progression missed by global measures.
