Comprehensive Perspective for Lung Cancer Characterisation Based on AI Solutions Using CT Images.

Area of Science:

• Oncology
• Radiology
• Artificial Intelligence
• Medical Imaging

Background:

• Lung cancer remains a leading global cause of cancer mortality, necessitating advanced diagnostic tools.
• Current computer-aided diagnosis (CAD) primarily analyzes tumor features from thoracic computed tomography (CT) scans.
• Radiomics traditionally focuses on intratumoral characteristics, often overlooking extratumoral information relevant to genotype classification.

Purpose of the Study:

• To highlight the critical need for integrating information from both lung tumors and surrounding lung structures in next-generation CAD systems.
• To explore how a holistic analysis can improve tumor genotype classification, specifically for epidermal growth factor receptor (EGFR) mutations.
• To underscore the potential impact on targeted therapies and personalized medicine in lung cancer management.

Main Methods:

• This perspective paper proposes a comprehensive analysis of combining tumor and non-tumor lung structure data.
• It advocates for the development of artificial intelligence (AI)-based approaches capable of holistic analysis of CT images.
• The focus is on leveraging interpretable AI models to identify novel imaging biomarkers.

Main Results:

• Integrating information from diverse lung structures alongside tumor features can provide a more comprehensive understanding of lung cancer.
• This integrated approach has the potential to significantly enhance the accuracy of tumor genotype classification, including EGFR mutations.
• The identification of novel biomarkers through AI can lead to deeper insights into cancer development pathways.

Conclusions:

• Future AI-driven CAD systems should adopt a holistic analytical approach, incorporating data from the entire lung.
• This comprehensive analysis is crucial for improving diagnostic accuracy and facilitating personalized medicine strategies.
• Enhanced diagnostic capabilities will directly support the selection of optimal treatment plans, particularly targeted therapies for specific mutations like EGFR.