Multi-class deep learning architecture for classifying lung diseases from chest X-Ray and CT images

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Radiology

Background:

• Medical imaging, including chest X-rays and CT scans, is crucial for diagnosing lung diseases.
• Early and accurate detection aids in disease prevention and management.
• Developing automated systems can improve the efficiency and accuracy of lung disease diagnosis.

Purpose of the Study:

• To propose an automated system for detecting multiple lung diseases using chest X-ray and CT scan images.
• To develop and evaluate a novel image enhancement technique for medical scans.
• To compare the performance of a customized convolutional neural network (CNN) and pre-trained models for lung disease classification.

Main Methods:

• A two-step approach involving pre-processing and deep learning classification.
• Image pre-processing utilizes a k-symbol Lerch transcendent function model for enhancement based on pixel probability.
• Classification employs a customized CNN, AlexNet, and VGG16Net models.

Main Results:

• The system achieved high classification accuracy, sensitivity, and specificity on public datasets.
• For X-ray datasets: 98.60% accuracy, 98.40% sensitivity, 98.50% specificity.
• For CT scan datasets: 98.80% accuracy, 98.50% sensitivity, 98.40% specificity.

Conclusions:

• The proposed automated system demonstrates significant potential for multi-lung disease detection.
• The integrated image enhancement model significantly benefits the overall processing and classification accuracy.
• The study validates the effectiveness of deep learning models in medical image analysis for pulmonary conditions.