Skip to content

NLP pipeline with a semi-supervised LSTM model, to automate fine-grained medical report classification and cohort generation with a small training set. Work conducted at MGH & BWH Center for Clinical Data Science.

Notifications You must be signed in to change notification settings

neild0/Semi-Supervised-NLP-for-MedReport-Classification

Repository files navigation

Semi Supervised NLP for Fine-Grained Medical Report Classification

This folder contains the most useful notebooks to train/eval/run the semi-supervised model.

  • quick_training is functionally the same as detailed_training, but without the complex documentation and methods defined in the notebook. It pulls the methods from the architecture folder, which has scripts of predefined functions.
  • interpret allows for model evaluation, through a variety of metrics, saliency scores of words in test strings, and encoding generation tools.
  • inference runs the given models on any given data, outputting a folder of cohort files, with relevant model predictions.

About

NLP pipeline with a semi-supervised LSTM model, to automate fine-grained medical report classification and cohort generation with a small training set. Work conducted at MGH & BWH Center for Clinical Data Science.

Resources

Stars

Watchers

Forks