EchoLabeler

Set of classes to perform label extraction training/inference based on echocardiogram reports.

Span labelling

• Approximate list lookup using regular expressions
• MedCAT + biLSTM
• Spacy SpanCategorizer

The identification can be followed up by an aggregation over the whole document to perform document classification.

Document labelling

• BOW, using TF-IDF, latent Dirichlet allocation, and topic modelling enrichment, followed by a standard gradient-boosted classifier.
• SetFit using RobBERTv2
• MedRoBERTa.nl
• biGRU
• CNN

fixes:

the current toml will not directly work.

• in gensim/matutils.py -> change from scipy.linalg import triu to from numpy import triu

Reference

Arxiv:

@misc{arends2024diagnosisextractionunstructureddutch,
      title={Diagnosis extraction from unstructured Dutch echocardiogram reports using span- and document-level characteristic classification}, 
      author={Bauke Arends and Melle Vessies and Dirk van Osch and Arco Teske and Pim van der Harst and René van Es and Bram van Es},
      year={2024},
      eprint={2408.06930},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2408.06930}, 
}

The models are on Huggingface