A Novel Hybrid Model for Automatic Non-Small Cell Lung Cancer Classification Using Histopathological Images.

Area of Science:

• Oncology
• Medical Imaging
• Computational Pathology

Background:

• Lung cancer, particularly non-small cell lung cancer (NSCLC), is a leading cause of cancer mortality.
• Accurate NSCLC subtype classification is crucial for effective treatment strategies.
• Manual histopathological image analysis is time-consuming and expertise-dependent.

Purpose of the Study:

• To develop a hybrid model for automated NSCLC subtype classification from histopathological images.
• To improve the accuracy and efficiency of NSCLC diagnosis.

Main Methods:

• A hybrid model integrating EfficientNet-B0 (deep features), Local Binary Pattern (LBP, textural features), and Vision Transformer (ViT, contextual features) was employed.
• Extracted features were combined into a comprehensive vector for machine learning classifiers.
• Support Vector Machine (SVM), Logistic Regression (LR), LightGBM, and XGBoost were evaluated.

Main Results:

• The highest classification accuracy of 99.87% was achieved using the hybrid model with EfficientNet-B0, LBP, ViT Encoder, and SVM.
• The proposed model significantly improved NSCLC subtype classification accuracy compared to traditional methods.
• Multiple training scenarios and classifiers were tested to validate the model's performance.

Conclusions:

• The hybrid model effectively integrates diverse features for robust NSCLC classification.
• This approach enhances diagnostic accuracy, reduces misdiagnosis risk, and supports better treatment planning.
• Automated classification of NSCLC subtypes shows significant promise in clinical pathology.