Analyzing Lung Disease Using Highly Effective Deep Learning Techniques.

Area of Science:

• Medical Imaging
• Artificial Intelligence in Medicine
• Radiology

Background:

• Image processing and computer-aided diagnosis (CAD) systems aid radiologists in detecting lung disease from chest X-rays.
• Early and accurate detection of lung lesions is crucial for effective treatment, but some cases present diagnostic challenges.

Purpose of the Study:

• To evaluate the performance of deep learning models for lung lesion detection.
• To develop an optimized convolutional neural network (CNN) model for enhanced computer-aided diagnosis of lung diseases.

Main Methods:

• Trained and validated MobileNet, Densenet-121, and Resnet-50 models on 5810 chest X-ray images.
• Utilized a rotational technique to augment the lung disease dataset for CNN training.
• Employed the Mish activation function and Nadam optimizer for model performance enhancement.

Main Results:

• The Densenet-121 model, optimized with Mish activation and Nadam, achieved high performance metrics.
• Overall accuracy, recall, precision, and F1 scores reached 98.88% during validation.
• Testing on unseen data yielded an accuracy rate of 98.97%.

Conclusions:

• The Densenet-121 model demonstrates superior performance for lung lesion detection.
• This study provides a robust foundation for developing advanced computer-aided diagnosis systems for lung diseases.
• Optimized deep learning approaches significantly enhance the accuracy of medical image analysis.