Robust Artificial Intelligence System for Predicting EBV Status in Gastric Cancer Diagnosis

Strategic Insights -

Research from Kangbuk Samsung Hospital in South Korea has developed a deep learning-based classifier to detect Epstein-Barr virus (EBV) positivity in gastric cancer biopsy and resection specimens. The system, called EBV-TRACER, uses stain normalization to improve accuracy and has been validated on a dataset of 2684 gastric specimens. The results show that EBV-TRACER outperforms other models in detecting EBV status, with area under the receiver operating curve (AUC) ranging from 0.6596 to 0.8414 in internal validation cohorts. The system has the potential to improve decision-making in gastric cancer diagnosis and treatment planning.

Key Takeaways:

  • The EBV-TRACER system uses a two-stage stain normalization-based robust artificial intelligence classifier to detect EBV positivity in gastric cancer biopsy and resection specimens.
  • The system was validated on a dataset of 2684 gastric specimens collected from January 1, 2011 to December 31, 2023.
  • EBV-TRACER yielded AUCs ranging from 0.6596 to 0.8414 in internal validation cohorts using three prediction scores: EBV positive cancer-to-tissue ratio, EBV positive cancer-to-tumor ratio, and EBV positive cancer size.
  • In the external validation cohort, AUCs of 0.7644, 0.7652, and 0.7221 were obtained for the three scores, respectively.
  • EBV-TRACER significantly outperforms models without stain normalization and those using conventional stain normalization.
  • The system has the potential to improve decision-making in gastric cancer diagnosis and treatment planning.

Statistics:

  • 2684 gastric specimens were collected from January 1, 2011 to December 31, 2023 and used to validate the EBV-TRACER system.
  • In internal validation cohorts, AUCs ranged from 0.6596 to 0.8414.
  • In the external validation cohort, AUCs of 0.7644, 0.7652, and 0.7221 were obtained for the three scores, respectively.

Sources:

  • A robust artificial intelligence system for predicting EBV status in gastric cancer biopsy and resection specimens. Scientific Reports, 2025;15(1):35100. Published by Nature Portfolio.
  • Kangbuk Samsung Hospital, Dept. of Pathology, Seoul, 03181, South Korea.