FVC-NET: An Automated Diagnosis of Pulmonary Fibrosis Progression Prediction Using Honeycombing and Deep Learning.

Area of Science:

• Medical Imaging
• Artificial Intelligence in Medicine
• Pulmonary Medicine

Background:

• Pulmonary fibrosis is a chronic lung disease characterized by irreversible lung scarring and loss of lung capacity.
• Forced Vital Capacity (FVC) is a key metric for assessing disease prognosis in pulmonary fibrosis.
• Accurate prediction of disease progression is crucial for effective patient management.

Purpose of the Study:

• To develop and evaluate a deep learning model, FVC-Net, for predicting pulmonary fibrosis progression.
• To integrate computed tomography (CT) scan data and patient metadata for enhanced predictive accuracy.
• To compare the performance of FVC-Net against existing state-of-the-art deep learning models.

Main Methods:

• A novel deep learning architecture, FVC-Net, was designed.
• The model utilizes a combined input of an image score (derived from CT scan honeycombing assessment) and patient metadata.
• A 3-layer neural network processes the input to predict disease progression.

Main Results:

• The FVC-Net model achieved superior performance compared to contemporary deep learning models on a pulmonary fibrosis progression dataset.
• The model demonstrated significant improvements in prediction accuracy, as indicated by a modified Laplace Log-Likelihood of -6.64.
• The image score, derived from segmented lung CT scans, effectively contributed to the predictive power of the model.

Conclusions:

• FVC-Net shows significant promise for accurately predicting pulmonary fibrosis progression.
• The integration of CT-derived image scores and patient metadata offers a powerful approach for disease prognostication.
• Future research should explore further refinements and applications of FVC-Net in clinical settings.