This repository supports the JCO-CCI paper "Natural language processing for patient selection in phase I/II oncology clinical trials", Delorme et al.
The tool aims at predicting the successful screening and dose-limiting toxicity period completion (SSD) from automatic analysis of consultation reports of patients with cancer. A random forest model uses frequencies of semantic clusters in a document to perform its predictions.