This research analyzed the clinical notes of epilepsy patients using techniques from corpus linguistics and machine learning and predicted which patients are can- didates for neurosurgery, i.e. have in- tractable epilepsy, and which are not. Information-theoretic and machine learn- ing techniques are used to determine whether and how sets of clinic notes from patients with intractable and non- intractable epilepsy are different. The re- sults show that it is possible to predict from an early stage of treatment which pa- tients will fall into one of these two cate- gories based only on text data. These re- sults have broad implications for develop- ing clinical decision support systems.

Click Here to View the Paper In XML Format