*WINNER* Diagnostic Prediction using Clinical text analysis


  • Sharanya Aavunoori


Interpreting symptoms plays an important role in determining whether your medical diagnosis is accurate. Learning and assessing the skill of writing patient notes requires feedback from other doctors, a time-intensive process that could be improved with the addition of machine learning. Clinical Skills exam is an important part of United States Medical Licensing Examination (USMLE). In this the test takers are required to interact with standardized patients and write a patient note. Trained physicians scores them based on a rubric that are outlined for each case's important concepts. Approaches using natural language processing have been created to address this problem, but patient notes can still be challenging to interpret computationally because features may be expressed in many ways. In this project we will develop an automated method to map clinical concepts like "diminished appetite" to various ways in which these concepts are expressed in clinical patient notes written ("eating less", "clothes fit looser"). we aim at predicting the diagnostics, which helps doctors in diagnosing a patient's ailment using the anamnesis provided. We will use NLP methods to classify and label the data based on the clinical concepts.





Computer Science