A Hybrid Framework for Lung Cancer Classification.

Area of Science:

• Oncology
• Medical Imaging
• Artificial Intelligence

Background:

• Lung cancer is a leading cause of cancer-related mortality globally.
• Early detection is crucial for improving patient outcomes.
• Deep learning models show promise for computer-aided lung cancer diagnosis but are prone to overfitting, hindering performance.

Purpose of the Study:

• To address the overfitting issue in deep learning-based lung cancer classification.
• To propose a novel hybrid framework for enhanced lung cancer image analysis and classification.

Main Methods:

• Developed a hybrid framework named LCGANT.
• Incorporated a lung cancer deep convolutional generative adversarial network (LCGAN) for synthetic image generation.
• Utilized a regularization-enhanced transfer learning model (VGG-DF) for image classification into three classes.

Main Results:

• Achieved high performance metrics: 99.84% accuracy, 99.84% precision, 99.84% sensitivity, and 99.84% F1-score.
• The LCGANT framework demonstrated superior performance compared to state-of-the-art methods on the dataset.
• Successfully resolved the overfitting problem in lung cancer classification tasks.

Conclusions:

• The LCGANT framework offers a robust solution for overcoming overfitting in lung cancer classification.
• This hybrid approach significantly enhances the accuracy and reliability of computer-aided lung cancer detection.
• The proposed method represents a significant advancement in deep learning applications for early lung cancer diagnosis.