Artificial Intelligence-Based Radiology Results for Lung Disease Detection

This research applies and tests Artificial Intelligence with the Deep Learning method to classify Chest X-Ray (CXR) of the lung images to determine lung disease with the imbalance data. Furthermore, this research also explained how Self-Supervised Learning (SSL) can help Deep Learning models during pre-training, and Transfer Learning (TL) can be used to train models, resulting in models that are more resistant to data imbalances. The Swapping Assignment between Views (SwAV) algorithm, in particular, has been widely recognized for its outstanding performance in improving the accuracy of CNN models for classification tasks following TL.

The AIRDC research team has successfully deployed the system for classification of CXR for determination of lung disease and reached the 0.952 AUROC with 0.821 macro-averaged F1 score. Here is the screen capture of the system that has been deployed by the research team.