AUTOMATED DIAGNOSIS OF ACUTE APPENDICITIS BASED ON CLINICAL NOTES

Abstract

Medical diagnosis is a very important task which requires high accuracy and efficiency, especially for patients admitted to the accident and emergency (A&E) department. It is extremely challenging for the attending doctors to perform quick and accurate diagnosis. Most of the relevant and useful information are in the form of free text notes entered by medical doctors. The objective of this thesis is to develop a system to aid in the diagnosis of appendicitis during A&E admissions by giving diagnosis recommendations to doctors. We have developed a novel neural network model that is able to learn from free texts and additional real-valued features without any feature engineering. The performance of our model is close to that of emergency department (ED) doctors. Visualization shows that the model is able to meaningfully learn important features, signs, and symptoms of patients from unstructured free-text ED notes, which helps doctors to make better diagnosis.