Use data augmentation for a deep learning classification model with chest X-ray clinical imaging featuring coal...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Pulmonary Medicine

Background:

• Coal workers' pneumoconiosis (CWP) presents unique challenges in chest X-ray diagnosis.
• Accurate identification of CWP imaging features is crucial for early detection and management.

Purpose of the Study:

• To develop a deep learning model for discovering unique chest X-ray imaging features of CWP.
• To enhance diagnostic accuracy using data augmentation techniques.

Main Methods:

• A prospective cohort study included 149 CWP patients and 68 dust-exposed workers.
• Chest X-ray images were analyzed using a deep learning model (ShuffleNet V2-ECA Net) with data augmentation.
• Model performance was evaluated using Receiver Operating Characteristic (ROC) curves, Area Under the Curve (AUC), accuracy, and loss curves.

Main Results:

• The ShuffleNet V2-ECA Net model demonstrated superior performance.
• The model achieved an average AUC of 0.98 for CWP feature classification.
• All clinical imaging feature classifications exceeded an AUC of 0.95.

Conclusions:

• A deep learning model (ShuffleNet V2-ECA Net) effectively identified unique CWP chest X-ray features.
• The model, enhanced by data augmentation, achieved 98% average accuracy.
• This approach provides valuable reference material for clinical applications in diagnosing CWP.