Synapse
⌘+K
Synapse
PulseExploreClubsResearchersJournals
Instagram
HomeClubsExplore
July 4, 2026BMJ Health & Care InformaticsOpen Access

Detection of cancer recurrence from Thai-English electronic medical records using sentence embeddings

View Full Paper
Ask AI
Bookmark
Share

Authors

ESEkapob SangariyavanichWPWanchana PonthongmakNTNawanan Theera-Ampornpunt

Discussion

Loading...

Member takes

Overview

Randomized trial developed bilingual sentence embeddings to detect cancer recurrence, indicating potential for clinical integration.

Key Points

  • This study aims to develop and validate sentence embedding models for identifying cancer recurrence in Thai-English medical records.
  • Developed monolingual and bilingual sentence-bidirectional encoder representations from transformers (SBERT) models.
  • Utilized a multicentre dataset of 32,436 documents from 1,250 patients for model development and an external dataset of 9,244 documents for validation.
  • Compared model performance against a fine-tuned PubMedBERT (MetBERT).
  • MetBERT achieved the highest AUPRC for locoregional versus no recurrence at 11.1% and for locoregional versus distant recurrence at 91.7%.
  • Bilingual-SBERT performed best in external validation for distant versus no recurrence with AUPRC of 17.55%–24.39%.
  • Fine-tuned MetBERT showed the highest robustness in distinguishing locoregional from distant recurrence, AUPRC between 88.30%–94.70%.

Cite This Study

Sangariyavanich et al. (2026) studied this question.

synapsesocial.com/papers/6a48a57289561a0c2d78e187https://doi.org/10.1136/bmjhci-2025-101997
View Full Paper
Ask AI
Bookmark
Share

Also Consider

Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context:

  1. 1Named Entity Recognition for Chinese Cancer Electronic Health Records—Development and Evaluation of a Domain-Specific BERT Model: Quantitative Study2025
  2. 2Applying natural language processing to patient messages to identify depression concerns in cancer patients2024 · 2 citations
  3. 3Domain and Language adaptive pre-training of BERT models for Korean-English bilingual clinical text analysis2025 · 2 citations
  4. 4Multimodal BEHRT: Transformers for Multimodal Electronic Health Records to predict breast cancer prognosis2024 · 1 citations
  5. 5Comparative Evaluation of Pre-Trained Language Models for Biomedical Information Retrieval2024