⚡ Quick Summary

This systematic scoping review analyzed 22 external validation studies of artificial intelligence (AI) pathology models for diagnosing lung cancer. The findings revealed that subtyping models performed exceptionally well, with average AUC values ranging from 0.746 to 0.999, highlighting the need for more robust external validation to enhance clinical adoption.

🔍 Key Details

• 📊 Dataset: 22 studies reviewed
• 🧩 Tasks performed: Classification of malignant vs. non-malignant tissue, tumor growth pattern classification, subtyping of adenocarcinomas vs. squamous cell carcinomas
• ⚙️ Technology: AI pathology models
• 🏆 Performance: Average AUC values from 0.746 to 0.999

🔑 Key Takeaways

• 📊 AI pathology models show promise in lung cancer diagnosis.
• 💡 Subtyping models are the most common and perform highly.
• 🏆 Average AUC values indicate strong diagnostic capabilities.
• 🔍 Methodological issues include small and non-representative datasets.
• 📝 Most studies were retrospective or case-control without real-world validation.
• 🌍 Increased external validation is essential for clinical adoption.
• 📅 Study period: January 2010 to October 2024.

📚 Background

The integration of artificial intelligence in healthcare, particularly in pathology, has the potential to revolutionize the diagnosis of diseases such as lung cancer. However, the clinical adoption of these digital pathology-based AI models has been limited, primarily due to a lack of robust external validation. This review aims to address this gap by providing a comprehensive overview of existing AI models and their performance in diagnosing lung cancer.

🗒️ Study

The systematic scoping review conducted by Arun et al. involved a thorough search of medical, engineering, and grey literature databases for external validation studies published between January 2010 and October 2024. A total of 22 studies were included, focusing on various AI models designed for lung cancer diagnosis, including tasks such as tissue classification and tumor subtyping.

📈 Results

The review found that subtyping models, which differentiate between adenocarcinomas and squamous cell carcinomas, were the most prevalent and exhibited high performance. The average AUC values for these models ranged from 0.746 to 0.999, indicating a strong ability to accurately classify lung cancer types. However, many studies faced methodological challenges, including the use of small and non-representative datasets, which could limit the applicability of the findings in real-world settings.

🌍 Impact and Implications

The findings of this review underscore the critical need for more rigorous external validation of AI pathology models to facilitate their clinical adoption. By addressing the methodological issues identified, researchers can enhance the reliability and applicability of these models, ultimately improving lung cancer diagnosis and patient outcomes. The potential for AI to transform pathology practices is immense, but it requires a concerted effort to validate these technologies in diverse clinical settings.

🔮 Conclusion

This systematic scoping review highlights the significant promise of AI pathology models in lung cancer diagnosis, particularly in subtyping tumors. However, the need for robust external validation remains paramount to ensure these models can be effectively integrated into clinical practice. Continued research and validation efforts will be essential in realizing the full potential of AI in enhancing diagnostic accuracy and patient care in oncology.

💬 Your comments

What are your thoughts on the integration of AI in lung cancer diagnosis? Do you believe that improved validation methods could enhance clinical adoption? 💬 Share your insights in the comments below or connect with us on social media:

• 🐦 X
• 📘 Facebook
• 🔗 LinkedIn