Multi-Label Classification of Chest X-ray Abnormalities Using Transfer Learning Techniques.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computer Vision

Background:

• Deep neural networks have revolutionized image classification.
• Convolutional Neural Network (CNN) techniques are increasingly applied to medical imaging.
• Chest X-ray analysis is crucial for diagnosing various thoracic diseases.

Purpose of the Study:

• To classify 14 different diseases using chest X-ray images.
• To develop a deep learning model that outperforms existing state-of-the-art methods.
• To address challenges in medical imaging datasets, such as imbalance and multi-label classification.

Main Methods:

• Utilized the EfficientNet model architecture for feature extraction.
• Employed a custom data split to manage dataset imbalances.
• Implemented binary cross-entropy loss for multi-label classification.
• Leveraged transfer learning and deep learning engineering techniques.

Main Results:

• Achieved an average area under the receiver operating characteristic curve (AUC-ROC) score of 84.28%.
• The proposed solution demonstrated superior performance compared to previous deep learning models.
• The model successfully classified 14 distinct diseases from chest X-ray images.
• Effective results were obtained using consumer-grade graphics processing units (GPUs).

Conclusions:

• The EfficientNet-based approach offers a highly effective solution for multi-label chest X-ray classification.
• Transfer learning and standard deep learning techniques enable high performance on accessible hardware.
• This study advances the application of deep learning in medical image analysis for disease detection.