PET/CT Based EGFR Mutation Status Classification of NSCLC Using Deep Learning Features and Radiomics Features

Area of Science:

• Medical Imaging
• Oncology
• Artificial Intelligence

Background:

• Accurate prediction of Epidermal Growth Factor Receptor (EGFR) mutation status is crucial for personalized treatment of non-small cell lung cancer (NSCLC).
• PET/CT imaging offers valuable data for non-invasive assessment of tumor characteristics.

Purpose of the Study:

• To compare the performance of radiomics and deep learning models in predicting EGFR mutation status in NSCLC patients using PET/CT images.
• To develop an optimized hybrid model for enhanced prediction accuracy.

Main Methods:

• 194 NSCLC patients' PET/CT images were analyzed.
• Radiomics (4306 features) and deep learning (2048 features) models were developed and compared.
• A hybrid model integrating deep learning scores and clinical data (smoking status) was created.

Main Results:

• The hybrid model demonstrated superior performance in both training (AUC: 0.91) and validation (AUC: 0.85) sets.
• The hybrid model outperformed individual radiomics (AUC: 0.82, 0.68) and deep learning (AUC: 0.90, 0.79) models.
• The hybrid model showed better diagnostic accuracy in predicting EGFR mutation status.

Conclusions:

• Deep learning models show promise in predicting EGFR mutation status from PET/CT images.
• A hybrid model combining deep learning features and clinical data (smoking) offers superior predictive performance for EGFR mutations in NSCLC.
• This approach can aid in selecting personalized treatment strategies for NSCLC patients.