VER-Net: a hybrid transfer learning model for lung cancer detection using CT scan images

Area of Science:

• Medical Imaging Analysis
• Artificial Intelligence in Oncology

Background:

• Lung cancer is a leading global malignancy, necessitating early detection for improved patient outcomes.
• Computer-aided detection systems, particularly those using machine learning and deep learning, are transforming clinical decision-making.
• Transfer learning is increasingly favored for image-based disease detection due to its efficiency.

Purpose of the Study:

• To develop and evaluate a novel transfer learning model (VER-Net) for accurate lung cancer detection using CT scans.
• To enhance the efficacy of the model through various optimization techniques.

Main Methods:

• A novel transfer learning model, VER-Net, was constructed by integrating three distinct transfer learning architectures.
• VER-Net was trained for multiclass classification of lung cancer using chest CT images.
• Techniques including image preprocessing, data augmentation, and hyperparameter tuning were employed to optimize VER-Net's performance.

Main Results:

• VER-Net demonstrated superior performance over eight other transfer learning models.
• The model achieved high scores: 91% accuracy, 92% precision, 91% recall, and 91.3% F1-score.
• VER-Net exhibited improved accuracy compared to existing state-of-the-art methods.

Conclusions:

• VER-Net provides an effective solution for lung cancer detection from CT images.
• The model's potential extends to the detection of other diseases identifiable through CT scans.