Lung Region Segmentation in Chest X-Ray Images using Deep Convolutional Neural Networks.

Area of Science:

• Medical Imaging Analysis
• Artificial Intelligence in Healthcare
• Radiology

Background:

• Lung cancer is a leading cause of cancer mortality globally.
• Accurate lung segmentation in medical images is crucial for early cancer detection.
• Deep Convolutional Neural Networks (DCNNs) show potential for medical image segmentation.

Purpose of the Study:

• To evaluate DCNN architectures for lung region segmentation in chest X-ray images.
• To compare the effectiveness of different regularization and optimization methods for segmentation accuracy.
• To identify the optimal DCNN configuration for improving Computer-Aided Diagnosis (CAD) systems.

Main Methods:

• Three DCNN architectures were assessed for lung segmentation on the JSRT database.
• Evaluated regularization techniques included Dropout, L2, and combined Dropout + L2.
• Compared optimization methods: Stochastic Gradient Descent with Momentum (SGDM), RMSprop, and ADAM.

Main Results:

• The combination of Dropout + L2 regularization and the ADAM optimizer yielded the best performance.
• Achieved a high Jaccard Coefficient of 0.97967 ± 0.00232, surpassing existing state-of-the-art methods.
• The optimized DCNN approach demonstrated superior lung segmentation accuracy.

Conclusions:

• The proposed DCNN method significantly improves lung segmentation accuracy in chest X-rays.
• This advancement can enhance the effectiveness of Computer-Aided Diagnosis for lung cancer.
• The optimized segmentation reduces analysis time for radiologists, improving clinical workflow efficiency.