Deep Residual Xception Network-Based Lung Cancer Detection Using CT Images.

Area of Science:

• Medical Imaging and Diagnostics
• Artificial Intelligence in Healthcare
• Oncology

Background:

• Lung cancer (LC) is a leading global cause of mortality, necessitating advancements in early detection.
• Current Computer-Aided Diagnosis (CAD) systems for lung cancer detection using Computed Tomography (CT) face challenges in processing time and diagnostic accuracy.
• Improving early diagnosis of lung cancer is critical for enhancing patient survival outcomes.

Purpose of the Study:

• To introduce and evaluate a novel Deep Residual Xception Network (DRX-Net) for effective and precise lung cancer identification from CT images.
• To address the limitations of existing CAD systems in terms of speed and accuracy for lung cancer diagnosis.
• To develop an advanced deep learning model for enhanced early detection of lung nodules.

Main Methods:

• CT images were preprocessed, including denoising with a Wiener filter.
• Lung nodule segmentation was performed using Pyramidal Attention-based Y Net (PAY-Net) with a hybrid loss function.
• Feature extraction and data augmentation were applied before classification using the proposed DRX-Net, integrating Xception and Deep Residual Network (DRN).

Main Results:

• The DRX-Net model demonstrated high performance in lung cancer detection.
• Achieved an accuracy of 93.988%.
• Reported a True Positive Rate (TPR) of 95.567% and a True Negative Rate (TNR) of 91.432% on the K Group 8 dataset.

Conclusions:

• The proposed DRX-Net offers a significant advancement in lung cancer detection using CT images.
• The model shows potential for improving diagnostic precision and efficiency in clinical settings.
• This deep learning approach contributes to the development of more effective Computer-Aided Diagnosis systems for lung cancer.