Accuracy of machine learning in preoperative identification of genetic mutation status in lung cancer: A systematic...

Area of Science:

• Oncology
• Radiology
• Bioinformatics

Background:

• Non-small cell lung cancer (NSCLC) treatment is increasingly personalized based on genetic mutation status.
• Accurate detection of these mutations is crucial for guiding therapy.
• Machine learning (ML) offers potential for non-invasive mutation detection.

Purpose of the Study:

• To systematically review and meta-analyze the performance of ML models in detecting genetic mutation status in NSCLC patients.
• To evaluate the diagnostic accuracy of radiomics-based ML models using various imaging modalities.

Main Methods:

• Systematic literature search of PubMed, Cochrane, Embase, and Web of Science up to July 2023.
• Meta-analysis of 128 original studies, focusing on ML models (primarily radiomics) for detecting EGFR, ALK, KRAS, and BRAF mutations.
• Analysis of models based on clinical features, CT, MRI, and PET-CT radiomics.

Main Results:

• ML models, particularly radiomics-based, demonstrated significant performance in detecting genetic mutations in NSCLC.
• For EGFR mutation detection, aggregated c-indexes in validation sets ranged from 0.750 to 0.822, with MRI-based radiomics showing higher accuracy (0.816).
• Combined clinical and radiomics models also showed high performance, with c-indexes exceeding 0.80 for CT, MRI, and PET-CT.

Conclusions:

• Radiomics-based ML methods show high accuracy for early discrimination of EGFR mutation status in NSCLC.
• Clinical variables play a significant role and should be considered alongside radiomics.
• Future research should explore radiomics' accuracy for other gene mutations in NSCLC.