TetraCVD: A Temporal-Textual Transformer based Model for Cardiovascular Disease Diagnosis

This repository contains the official implementation of TetraCVD, a multimodal deep learning model for cardiovascular disease (CVD) diagnosis and subtype classification, as presented at the 2023 IEEE International Conference on Bioinformatics and Biomedicine (BIBM).

Paper Information

• Title: TetraCVD: A Temporal-Textual Transformer based Model for Cardiovascular Disease Diagnosis
• Authors: K. Lu, F. Zhao, P. Gu, H. Wang, T. Zang, H. Wang
• Conference: 2023 IEEE International Conference on Bioinformatics and Biomedicine (BIBM), Istanbul, Turkey
• Date: December 5–8, 2023
• Pages: 2129–2132
• DOI: 10.1109/BIBM58861.2023.10385761

Abstract

Cardiovascular disease (CVD) is one of the leading causes of death globally. There is considerable clinical significance and an emerging need to assist doctors in diagnosing cardiovascular disease and identifying its subtype, from which doctors can provide different treatments and medications to increase the cure rate.

The goal of this paper is to develop a deep learning model to predict cardiovascular disease and classify its subtype by handling data from two modalities: time-series vital signs and text reports. We propose TetraCVD, a temporal-textual transformer based model, to address the challenges of irregular temporal feature extraction and medical long-text feature extraction, respectively.

TetraCVD is a multimodal deep learning model consisting of two networks:

• cvdGNN – a time-series backbone that leverages knowledge from temporal vital signs
• cvdHierBERT – a language backbone that processes text reports

Our results show that TetraCVD achieves promising performance in predicting subtypes of cardiovascular disease using the P18-ECER dataset and obtains state-of-the-art results. This study is among the first efforts that use both time-series vital signs and text report data to predict cardiovascular disease and its subtype.

Model Architecture

TetraCVD is a multimodal architecture that integrates two specialized backbones:

Component	Modality	Function
cvdGNN	Time-series vital signs	Extracts irregular temporal features
cvdHierBERT	Text reports	Extracts features from medical long-text

The two networks work in parallel and their outputs are combined for final CVD diagnosis and subtype classification.

📝 Citation

If you find this work useful for your research, please cite it using the following formats:

APA 7th (text):

Lu, K., Zhao, F., Gu, P., Wang, H., Zang, T., & Wang, H. (2023). TetraCVD: A Temporal-Textual Transformer based Model for Cardiovascular Disease Diagnosis. In Proceedings - 2023 IEEE International Conference on Bioinformatics and Biomedicine (BIBM) (pp. 2129–2132). IEEE. https://doi.org/10.1109/BIBM58861.2023.10385761

BibTeX (for LaTeX users):

@inproceedings{Lu2023TetraCVD,
  author    = {Lu, K. and Zhao, F. and Gu, P. and Wang, H. and Zang, T. and Wang, H.},
  title     = {TetraCVD: A Temporal-Textual Transformer based Model for Cardiovascular Disease Diagnosis},
  booktitle = {Proceedings - 2023 IEEE International Conference on Bioinformatics and Biomedicine (BIBM)},
  year      = {2023},
  pages     = {2129--2132},
  doi       = {10.1109/BIBM58861.2023.10385761},
  note      = {Conference held in Istanbul, Turkey, December 5-8, 2023}
}