JoVE - Journal of Visualized Experiments
JoVE Visualize
Contact Us

Tumor Cell Proportion Assessment in Advanced Non-Squamous Non-Small Cell Lung Cancer Tissue Samples in Real-World Settings in Japan: The ASTRAL Study

  • 0Center for Development of Advanced Diagnostics, Hokkaido University Hospital, Sapporo 060-8648, Japan.
Diagnostics (basel, Switzerland) +

|

September 13, 2025

|

September 13, 2025

View abstract on PubMed

Summary

This summary is machine-generated.

Accurate tumor cell proportion assessment is vital for non-small cell lung cancer (NSCLC) treatment. An AI algorithm showed moderate agreement with a central pathology committee, suggesting potential utility in improving NSCLC diagnostics.

Area Of Science

  • Oncology
  • Pathology
  • Medical Diagnostics

Background

  • Accurate driver gene alteration identification is crucial for non-small cell lung cancer (NSCLC) treatment selection.
  • Precise tumor cell proportion assessment is essential for reliable gene alteration detection in NSCLC.
  • The ASTRAL study evaluated inter-rater agreement in tumor cell proportion assessments for advanced NSCLC.

Purpose Of The Study

  • To investigate the agreement in tumor cell proportion assessments among local pathologists, a Central Pathology Committee (CPC), and an artificial intelligence (AI) algorithm.
  • To determine the reliability of AI in estimating tumor cell proportion compared to human expert assessments.
  • To assess the clinical utility of AI in improving diagnostic accuracy for NSCLC.

Main Methods

  • Prospective, observational, multicenter study (ASTRAL) involving 204 advanced NSCLC patients.
  • Tumor tissues assessed by local pathologists (H&E slides), CPC (digitized slides), and an AI algorithm (digitized slides).
  • Intraclass correlation coefficient (ICC) used to measure agreement between raters.

Main Results

  • Poor to moderate agreement (ICC=0.588) between local pathologists and the CPC.
  • Moderate agreement (ICC=0.652) between the AI algorithm and the CPC.
  • Poor to moderate agreement (ICC=0.465) between the AI algorithm and local pathologists.

Conclusions

  • The AI algorithm demonstrated the highest numerical agreement with the CPC, indicating potential usefulness in clinical practice.
  • Current agreement levels highlight the need for continued efforts to refine AI algorithms for accurate tumor cell proportion estimation.
  • Integrating AI tools may enhance the consistency and accuracy of diagnostic assessments in real-world NSCLC management.

Keywords: