Data-driven classification and explainable-AI in the field of lung imaging

Area of Science:

• Medical Imaging Analysis
• Artificial Intelligence in Healthcare
• Radiology

Background:

• Accurate lung disease detection in medical images is challenging, even for experienced radiologists.
• Complex biomarkers and unseen patterns can lead to diagnostic inaccuracies.
• Machine learning offers potential solutions for improving diagnostic accuracy.

Purpose of the Study:

• To review datasets and machine learning techniques for lung disease classification, focusing on pneumonia detection in chest X-rays.
• To compare conventional machine learning, deep learning models (CNNs), and ensemble methods.
• To emphasize the role of Explainable AI (XAI) in enhancing model transparency and trust.

Main Methods:

• Literature review of datasets, preprocessing, feature extraction, and classification techniques.
• Analysis of conventional machine learning, pre-trained deep learning models, customized CNNs, and ensemble methods.
• Inclusion of machine vision, machine learning, deep learning, and XAI approaches.

Main Results:

• Transfer learning-based methods utilizing CNNs demonstrated superior performance in lung disease classification.
• Ensemble models and features also showed significant effectiveness.
• XAI techniques are increasingly important for understanding model decisions and improving clinical trust.

Conclusions:

• Transfer learning and ensemble methods are highly effective for lung disease classification using chest X-rays.
• XAI integration is vital for transparent and trustworthy AI in medical diagnostics.
• This review provides valuable insights for researchers in medical imaging and AI.