Key takeaways
The PhD project will explore if and how Natural Language Processing and Advanced Statistical Methods can improve prediction accuracy regarding the outcome of medical treatment. The ultimate aim is to use these approaches to develop reliable clinical decision support systems. The problem of early prediction of complications specific to the recovery after esophageal resection surgery will be considered the first application. The explored approaches are expected to be generalizable also to other clinical settings.
The challenge
This research project focuses on developing natural language robots to support clinical decisions. These language robots are computer programs with the ability to understand spoken and written human language. Exploiting textual information sources in an automated and systematic manner enables additional expert knowledge to be represented explicitly. The work aims to improve predictions regarding the probability of a patient successfully recovering from treatment. Such predictions, particularly in fields with a limited number of patients, may become more accurate, enabling improved clinical decisions, when textual observations by clinical staff can be incorporated fully. Three specific questions in this research are
- How to quantify the accuracy of the extracted expert knowledge when predicting patient-specific outcomes?
- How to build prediction models based on robotic language processing?
- How to incorporate these prediction models into clinical decision support?