Back

Automated Medical Chart Review for Breast Cancer: A Novel Natural Language Processing Software System

Chen, Y.; Hao, L.; Zou, V. Z.; Hollander, Z.; Ng, R. T.; Isaac, K. V.

2021-05-04 health informatics
10.1101/2021.05.04.21256134 medRxiv
Show abstract

The incoming health records to the BC Cancer Registry are processed between two to three years behind real-time. In response, we developed a Natural Language Processing (NLP) software to automate the electronic chart review workflow. For the same task that costs hundreds of hours of trained labour, our pipeline extracts data within minutes. During preliminary evaluation, an MD student yielded 93.0% and 98.2% accuracies on a sample of operative and pathology breast cancer documents (for a total number of 2,563 data points processed). In comparison, our prototype achieved 89.6% and 91.4% accuracies, respectively. Future plans include improving the performance of the pipeline and eventually adapt it to accepting a more comprehensive range of electronic health records across cancer types and diseases. In the context of BCs digital healthcare transformation initiatives, this customized software may provide time and cost savings for both the Registry and cancer researchers.

Matching journals

The top 4 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.