Pulmonary Function Prediction Method Based on Convolutional Surface Modeling and Computational Fluid Dynamics...

Area of Science:

• Medical Imaging and Computational Modeling
• Respiratory Physiology
• Biomedical Engineering

Background:

• Pulmonary function tests (PFTs) are crucial for assessing respiratory diseases but are limited by patient compliance and inability to predict future health trends.
• Existing computational fluid dynamics (CFD) airway models lack accuracy due to geometric defects and failure to represent patient-specific structural features.

Purpose of the Study:

• To develop an individualized airway modeling method using CT data and convolutional surface technology to overcome current model limitations.
• To predict patient lung function using CFD simulations based on the constructed airway models.
• To explore the application of this method for preoperative prediction of airway expansion extent in patients with large airway stenosis.

Main Methods:

• Personalized airway models were constructed from CT-derived airway skeletons using convolutional surface technology.
• Airway models were simulated using patient-specific pulmonary function data to generate simulation data.
• A regression equation was established between measured pulmonary function values and airway simulation data for predictive modeling.

Main Results:

• A proof-of-concept study accurately predicted FEV1 in patients with large airway stenosis using a linear regression model (RMSE = 0.0246, R² = 0.9822).
• The method calculated a 72.86% airway radius expansion needed at the stenotic site to achieve normal FEV1.
• Preoperative predictions for the required airway dilation were successfully made.

Conclusions:

• A novel method for predicting patient pulmonary function using CFD and convolutional surface technology was established.
• This approach mitigates limitations associated with patient compliance and accuracy in traditional PFTs.
• The study offers a valuable tool for the preoperative evaluation of airway dilation therapy.