viTBI-BERT: A Vietnamese Language Model for Prediction of Traumatic Brain Injury

  • Duc-Khiem Doan School of Electrical and Electronic Engineering, Hanoi University of Science and Technology, Hanoi, Vietnam
  • Thanh-Hai Tran School of Electrical and Electronic Engineering, Hanoi University of Science and Technology, Hanoi, Vietnam
  • Trung-Kien Tran Military Insitute of Science and Technology, Hanoi, Vietnam
  • Thi-Lan Le School of Electrical and Electronic Engineering, Hanoi University of Science and Technology
  • Hai Vu School of Electrical and Electronic Engineering, Hanoi University of Science and Technology
  • Huu-Khanh Nguyen Military Hospital 103, Hanoi, Vietnam
  • Van-Mao Can Vietnam Military Medical University, Hanoi, Vietnam
  • Thanh-Bac Nguyen Military Hospital 103, Hanoi, Vietnam
Keywords: Traumatic Brain Injury, Large Language Model, Transformer, Classification

Abstract

Artificial Intelligence (AI) is increasingly utilized for disease prediction by analyzing large and complex datasets. Traditionally, AI-based traumatic brain injury (TBI) prediction relies on multimodal data, including structured
information like test results and unstructured data such as CT and/or MRI scans. However, physician conclusions in text form at admission and discharge represent an underutilized source of valuable insights into a patient’s condition. This paper proposes a novel approach to classifying TBI severity using these physicians’ written conclusions. We introduce viTBI-BERT, a model built on the ViHealthBERT backbone, incorporating pre-processing techniques such as acronym normalization, special character removal, word segmentation, and textual data augmentation. Following pre-processing, the data is passed through the pre-trained ViHealthBERT model, augmented with a fully connected layer and a softmax layer for classification. Evaluated on three sets of medical notes from a self-collected dataset of 503 Vietnamese patients, viTBI-BERT achieved a sensitivity of 71% in classifying four levels of injury severity. These findings demonstrate the potential of language based analysis, which, when integrated with structured clinical and subclinical data, could further improve TBI classification outcomes.

Author Biographies

Duc-Khiem Doan, School of Electrical and Electronic Engineering, Hanoi University of Science and Technology, Hanoi, Vietnam

Duc-Khiem Doan is currently a final-year undergraduate student in Biomedical Engineering at Hanoi University of Science and Technology. His main research interests include medical data processing and the application of machine learning and deep learning models to biomedical data.

Thanh-Hai Tran, School of Electrical and Electronic Engineering, Hanoi University of Science and Technology, Hanoi, Vietnam

Thanh-Hai Tran graduated with an engineer’s degree in Information Technology from Hanoi University of  Science and Technology (HUST) in 2001. She holds an M.S. degree and a Ph.D. degree in Imagery Vision Robotics from Grenoble INP, France, in 2002 and 2006 respectively. Currently, she is a lecturer/researcher at the School of Electrical and Electronics Engineering and International Research Institute in Multimedia, Information, Communication, and Application, HUST. Her main research interests are visual object recognition, video understanding, human-robot interaction, and text detection for applications in Computer Vision.

Trung-Kien Tran, Military Insitute of Science and Technology, Hanoi, Vietnam

Trung-Kien Tran graduated from Hanoi University of Science and Technology in 2004 and worked at the Military Information Technology Institute, AMST, as a researcher in the field of cybersecurity until 2014. After that, he began pursuing a Master´ s degree and Ph.D. in Machine Learning at the Artificial Intelligence Laboratory, National Defense Academy of Japan, and graduated in March 2020. Currently, he is an AI specialist at the Military Information Technology Institute, AMST, with primary research interests in Edge Artificial intelligence, Human-Robot Interaction, and Unmanned Vehicle.

Thi-Lan Le, School of Electrical and Electronic Engineering, Hanoi University of Science and Technology

Thi-Lan Le graduated in Information Technology from Hanoi University of Science and Technology (HUST), Vietnam. She obtained an MS. degree in Signal Processing and Communication from HUST, Vietnam. In 2009, she received her Ph.D. degree at INRIA Sophia Antipolis, France in video retrieval. She is currently an associate professor at the School of Electrical and Electronic Engineering (SEEE), HUST, Vietnam. Her research interests include image processing, computer vision, content-based indexing and, retrieval, video understanding, and human-robot interaction.

Hai Vu, School of Electrical and Electronic Engineering, Hanoi University of Science and Technology

Hai Vu received a PhD in Computer Science from Osaka University, Japan, in 2009. Currently, the author is a lecturer at the School of Electrical and Electronic Engineering (SEEE), HUST, Vietnam. His research interests include computer vision-based techniques in smart surveillance camera networks, medical image analysis for diagnostic assistance and Human-Machine Interaction.

Huu-Khanh Nguyen, Military Hospital 103, Hanoi, Vietnam

Huu-Khanh Nguyen is a neurosurgery resident at 103 Hospital in Hanoi, Vietnam. He graduated from Vietnam Military Medical University in 2022, where he completed seven years of comprehensive medical training. Since then, he has been specializing in neurosurgery, developing his skills and knowledge at one of Vietnam’s leading medical institutions.

Van-Mao Can, Vietnam Military Medical University, Hanoi, Vietnam

Van Mao Can graduated from the Military Academy of Medicine, Hanoi, Vietnam, in 2000. He earned his  Master’s Degree in Medicine in 2003 and a Doctoral Degree in System Emotional Science from Toyama University, Japan in 2009. He was appointed Associate Professor in Medicine in March 2018. Since 2018, he has been the Head of the Pathophysiology Department at the Military Medical University, contributing significantly to education, research, and medical advancements.

Thanh-Bac Nguyen, Military Hospital 103, Hanoi, Vietnam

Thanh Bac Nguyen graduated from the Military Academy of Medicine, Hanoi, in 2000 and completed his residency in neurosurgery at Military Hospital 103 in 2005. He earned his Ph.D. in 2020. Since then, he has been the Head of the Department of Neurosurgery at Military Hospital 103. In December 2024, he was appointed As sociate  Professor. He is dedicated to advancing neurosurgery and mentoring the next generation of medical professionals in Vietnam.

References

H. Khalili, M. Rismani, M. A. Nematollahi, M. S. Masoudi, A. Asadollahi, R. Taheri, H. Pourmontaseri, A. Valibeygi, M. Roshanzamir, R.Alizadehsani et al., “Prognosis prediction in traumatic brain injurypatients using machine learning algorithms,” Scientific reports, vol.13,no.1,p.960,2023.

D. Yeboah, H. Nguyen, D. B. Hier, G. R. Olbricht, and T. Obafemi Ajayi, “A deep learning model to predict traumatic brain injury severity and outcome from mr images,” in 2021 IEEE Conference on Computational Intelligence in Bioinformatics and Computational Biology (CIBCB), Oct 2021, pp. 1–6.

M. P. Nguyen, V. H. Tran, V. Hoang, T. D. Huy, T. H. Bui, and S. Q. H. Truong, “ViHealthBERT: Pre-trained Language Models for Vietnamese in Health Text Mining,” 13th Edition of its Language Resources and Evaluation Conference, 2022.

K. Matsuo, H. Aihara, T. Nakai, A. Morishita, Y. Tohma, and E. Kohmura, “Machine learning to predict in-hospital morbidity and mortality after traumatic brain injury,” Journal of Neurotrauma, vol. 37, no. 1, pp. 202–210, 2020. [Online]. Available: https://doi.org/10.1089/neu.2018.6276

R. Bruschetta, G. Tartarisco, L. F. Lucca, E. Leto, M. Ursino, P. Tonin, G. Pioggia, and A. Cerasa, “Predicting outcome of traumatic brain injury: Is machine learning the best way?” Biomedicines, vol. 10, no. 3, 2022. [Online]. Available: https: //www.mdpi.com/2227-9059/10/3/686

S. M. Adil, C. Elahi, D. N. Patel, A. Seas, P. I. Warman, A. T. Fuller, M. M. Haglund, and T. W. Dunn, “Deep learning to predict traumatic brain injury outcomes in the low-resource setting,” World Neurosurgery, vol. 164, pp. e8–e16, 2022.

M. Pease, D. Arefan, J. Barber, E. Yuh, A. Puccio, K. Hochberger, E. Nwachuku, S. Roy, S. Casillo, N. Temkin, D. O. Okonkwo, S. Wu, , N. Badjatia, Y. Bodien, A.-C. Duhaime, V. R. Feeser, A. R. Ferguson, B. Foreman, R. Gardner, S. Gopinath, C. D. Keene, C. Madden, M. McCrea, P. Mukherjee, L. B. Ngwenya, D. Schnyer, S. Taylor, and J. K. Yue, “Outcome prediction in patients with severe traumatic brain injury using deep learning from head ct scans,” Radiology, vol. 304, no. 2, pp. 385–394, 2022. [Online]. Available: https://doi.org/10.1148/radiol.212181

S. Ding, J. Ye, X. Hu, and N. Zou, “Distilling the knowledge from large-language model for health event prediction,” medRxiv, pp. 2024–06, 2024.

M. Jin, Q. Yu, D. Shu, C. Zhang, L. Fan, W. Hua, S. Zhu, Y. Meng, Z. Wang, M. Du et al., “Health-llm: Personalized

retrieval- augmented disease prediction system,” arXiv preprint arXiv:2402.00746, 2024.

L. Rasmy, Y. Xiang, Z. Xie, C. Tao, and D. Zhi, “Med-bert: pretrained contextualized embeddings on large-scale structured electronic health records for disease prediction,” NPJ digital medicine, vol. 4, no. 1, p. 86, 2021.

S. Hegselmann, A. Buendia, H. Lang, M. Agrawal, X. Jiang, and D. Sontag, “Tabllm: Few-shot classification of tabular data with large language models,” in International Conference on Artificial Intelligence and Statistics. PMLR, 2023, pp. 5549–5581.

Z. Li, Y. Li, Q. Li, P. Wang, D. Guo, L. Lu, D. Jin, Y. Zhang, and Q. Hong, “Lvit: language meets vision transformer in medical image segmentation,” IEEE transactions on medical imaging, 2023.

C. Le Barbey, “Evaluation of artificial language model chatgpt4 to predict 6-month outcome after traumatic brain injury,” Ph.D. dissertation, 2023.

C. Vo, T. Cao, and B. Ho, “Abbreviation detection in vietnamese clinical texts,” VNU Journal of Science: Computer Science and Communication Engineering, vol. 34, no. 2, 2018. [Online]. Available: //jcsce.vnu.edu.vn/index.php/jcsce/article/view/211

A. Tabassum and D. R. R. Patil, “A survey on text pre-processing & feature extraction techniques in natural language processing,” 2020. [Online]. Available: https://api.semanticscholar.org/CorpusID: 235211496

D. Q. Nguyen, D. Q. Nguyen, T. Vu, M. Dras, and M. Johnson, “A fast and accurate vietnamese word segmenter,” CoRR, vol. abs/1709.06307, 2017. [Online]. Available: http://arxiv.org/abs/1709. 06307

T. Vu, D. Q. Nguyen, D. Q. Nguyen, M. Dras, and M. Johnson, “VnCoreNLP: A Vietnamese natural language processing toolkit,” in Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Demonstrations. Association for Computational Linguistics, 2018, pp. 56–60. [Online]. Available: https://aclanthology.org/N18-5012

J. Wei and K. Zou, “EDA: Easy data augmentation techniques for boosting performance on text classification tasks,” in Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP). Association for Computational Linguistics, nov 2019, pp. 6382–6388. [Online]. Available: https://aclanthology.org/D19-1670

T. P. Nguyen, V.-L. Pham, H.-A. Nguyen, H.-H. Vu, N.-A. Tran, and T.-T.-H. Truong, “A two-phase approach for building Vietnamese WordNet,” in Proceedings of the 8th Global WordNet Conference (GWC). Global Wordnet Association, 27–30 2016, pp. 261–266. [Online]. Available: https://aclanthology.org/2016.gwc-1.38

M. Xia, X. Kong, A. Anastasopoulos, and G. Neubig, “Generalized data augmentation for low-resource translation,” in Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics. Association for Computational Linguistics, jul 2019, pp. 5786–5796. [Online]. Available: https://aclanthology.org/P19-1579

A. Sugiyama and N. Yoshinaga, “Data augmentation using back-translation for context-aware neural machine translation,” in Proceedings of the Fourth Workshop on Discourse in Machine Translation (DiscoMT 2019). Association for Computational Linguistics, nov 2019, pp. 35–44. [Online]. Available: https://aclanthology.org/D19-6504

J. Devlin, M.-W. Chang, K. Lee, and K. Toutanova, “Bert: Pre-training of deep bidirectional transformers for language understanding,” in North American Chapter of the Association for Computational Linguistics, 2019. [Online]. Available: https: //api.semanticscholar.org/CorpusID:52967399

A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, L. Kaiser, and I. Polosukhin, “Attention is all you need,” 2023. [Online]. Available: https://arxiv.org/abs/1706.03762

M. Sundararajan, A. Taly, and Q. Yan, “Axiomatic attribution for deep networks,” in International conference on machine learning. PMLR, 2017, pp. 3319–3328.

Z. Zhao, R. Anand, and M. Wang, “Maximum relevance and mini mum redundancy feature selection methods for a marketing machine learning platform,” in 2019 IEEE International Conference on Data Science and Advanced Analytics (DSAA), 2019, pp. 442–452.

F. Gaillard, “Rotterdam CT score of traumatic brain injury Radiology Reference Article | Radiopaedia.org — radiopaedia.org,” https://radiopaedia.org/articles/rotterdam-ct-score-of-traumatic-brain-injury.

N. Kokhlikyan, V. Miglani, M. Martin, E. Wang, B. Alsallakh, J. Reynolds, A. Melnikov, N. Kliushkina, C. Araya, S. Yan, and O. Reblitz-Richardson, “Captum: A unified and generic model interpretability library for pytorch,” 2020.

Q. T. Nguyen, T. L. Nguyen, N. H. Luong, and Q. H. Ngo, “Fine-tuning BERT for sentiment analysis of vietnamese reviews,” CoRR, vol. abs/2011.10426, 2020. [Online]. Available: https://arxiv.org/abs/2011.10426

D. Q. Nguyen and A. T. Nguyen, “Phobert: Pre-trained language models for vietnamese,” CoRR, vol. abs/2003.00744, 2020. [Online]. Available: https://arxiv.org/abs/2003.00744

L. Phan, T. Dang, H. Tran, T. H. Trinh, V. Phan, L. D. Chau, and M.-T. Luong, “Enriching biomedical knowledge for low-resource language through large-scale translation,” 2022. [Online]. Available: https://arxiv.org/abs/2210.05598

Published
2025-03-04