dPFT (Santibanez, V., Pisano, T.J., et al. 2024) are generated using an automated lung analysis pipeline that takes raw dynamic digital radiography (DDR) videos and outputs DDR-based Pulmonary Function Test (dPFT) data. This is accomplished using convolutional neural networks for serial anatomical detection across frames.
An overview of dPFT is shown here:
Please see INSTALLATION.md for installation instructions.
Please see dPFT_example.ipynb for basic dPFT inference and analysis example.
Trained neural networks can be downloaded below: All models (recommended). Individual models. Annotated SLEAP training sets, to use for transfer learning (Advanced).