Predicting EGFR mutation status in lung adenocarcinoma on computed tomography image using deep learning

Area of Science:

• Oncology
• Radiology
• Artificial Intelligence

Background:

• Epidermal growth factor receptor (EGFR) genotyping is crucial for guiding lung adenocarcinoma treatment, particularly with tyrosine kinase inhibitors.
• Current methods like biopsy and sequence testing are invasive and face challenges with tissue accessibility.

Purpose of the Study:

• To develop and validate a deep learning model for predicting EGFR mutation status in lung adenocarcinoma non-invasively using computed tomography (CT) images.

Main Methods:

• Retrospective collection of pre-operative CT images, EGFR mutation data, and clinical information from 844 lung adenocarcinoma patients across two hospitals.
• Development of an end-to-end deep learning model trained on 14,926 CT images to predict EGFR mutation status.

Main Results:

• The deep learning model achieved high predictive performance in both primary (AUC 0.85) and validation (AUC 0.81) cohorts, significantly outperforming previous methods.
• The model's deep learning score showed significant differentiation between EGFR-mutant and EGFR-wild type tumors (p<0.001).

Conclusions:

• The proposed deep learning model provides a non-invasive, accurate, and user-friendly method for predicting EGFR mutation status in lung adenocarcinoma.
• This CT-based approach can aid in treatment selection, complementing routine lung cancer diagnostic procedures.