Physics-Informed In-Silico Dynamic Computed Tomography of Human Lungs: Generation, Evaluation, and Refinement

Area of Science:

• Medical Imaging
• Computational Biology
• Pulmonary Medicine

Background:

• Lung injuries cause uneven ventilation, and current lung function tests lack regional detail.
• Dynamic medical imaging and image registration (IR) can assess lung movement, but validation is challenging.
• In-silico images from finite element (FE) simulations offer a way to verify IR results.

Purpose of the Study:

• To use in-silico CT images from FE simulations to evaluate the accuracy of an IR method for lung parenchyma.
• To compare displacement and volumetric strain estimations between FE simulations and IR on actual and in-silico CT images.

Main Methods:

• Reconstructed lungs from human 4DCT images to create an FE mesh.
• Performed in-silico simulations using the lung FE mesh to generate in-silico dynamic CT images.
• Executed IR on both actual and in-silico images and compared results to FE simulations.

Main Results:

• FE simulations and IR showed good agreement in lung displacement estimation.
• The greatest displacement difference between IR and FE simulations was 2.7 mm at the main bronchi.
• Higher resolution in-silico images improved volumetric strain contour agreement.

Conclusions:

• FE simulation-derived in-silico CT images effectively validate IR methods for lung analysis.
• This approach can optimize medical imaging techniques for studying lung diseases.
• The method holds potential for improved regional lung function assessment in clinical settings.