Multi-task deep learning model for predicting EGFR mutation status in NSCLC

Area of Science:

• Oncology
• Radiology
• Artificial Intelligence

Background:

• Epidermal growth factor receptor (EGFR) mutation status is crucial for non-small cell lung cancer (NSCLC) treatment selection.
• Current methods for determining EGFR mutation status can be invasive.
• Deep learning (DL) offers potential for non-invasive predictive biomarkers.

Purpose of the Study:

• To develop and evaluate a multi-task deep learning (MTDL) model for predicting EGFR mutation status using CT images.
• To assess the association of the MTDL model's predictions with patient survival and tumor characteristics.

Main Methods:

• A multi-task deep learning (MTDL) model was developed to analyze CT images.
• The model was trained to predict EGFR mutation status.
• Performance was evaluated, and associations with survival and tumor microenvironment were analyzed.

Main Results:

• The MTDL model demonstrated promising accuracy in predicting EGFR mutation status.
• The MTDL score showed significant association with survival in patients undergoing EGFR-targeted therapy.
• The score also correlated with gene expression patterns and tumor microenvironment characteristics.

Conclusions:

• The developed MTDL model shows potential as an accurate, non-invasive biomarker for EGFR mutation status prediction in NSCLC.
• This approach can aid in personalized treatment strategies for NSCLC patients.
• Further validation is warranted to integrate this method into clinical practice.