High-Performance Detection of Pulmonary Diseases in CT Images Using PFC-DenseNet121 and Binary Robust Independent Elementary Features

A serious interstitial lung disease called pulmonary fibrosis (PF) causes the lung tissue to gradually scar, impairing respiratory function irreversibly and posing serious health risks. CT scans play a pivotal role in diagnosing lung diseases by providing detailed cross-sectional images with offering a more comprehensive view of lung structures compared to conventional imaging methods. This study introduces an enhanced DenseNet architecture combined with BRIEF (Binary Robust Independent Elementary Features) for effectively detecting pulmonary fibrosis in CT scan images. The model identifies critical regions of interest by integrating BRIEF for keypoint detection and local feature extraction while maintaining computational efficiency. The modified DenseNet, named PFC-DenseNet, incorporates additional layers and attention mechanisms to improve feature extraction, mitigate the vanishing gradient problem, and ensure robust performance. Preprocessing techniques, including resizing and normalization and further enhance the network reliability by reducing overfitting. A pulmonary fibrosis dataset with tagged CT images is used to train and assess the model. It demonstrates improved performance in terms of accuracy, sensitivity, specificity, F1 score, and precision. Additionally, it effectively identifies affected areas in CT images associated with pulmonary fibrosis.