A comprehensive framework studying the performance of simple and complex stroke-specific Natural Language Processing (NLP) and Machine Learning (ML) methods to determine presence, location, and acuity of ischemic stroke from radiographic text.
##Folders Included
- Classifier: code for training across all outcomes and methods.
- DataFromStructuredFiles: exploratory analysis code from structured patient data.
- GloveTraining: main outputs and code related to the creation of neurology specific embeddings. We directly used this code for training the algorithm.
- Ngrams_WordCounts: exploratory analysis with n-grams and word counts.
- SensitivityAnalysis: our script for the sensitivity analysis we did on our results.