Topic Modeling in Medical Data Analysis. Case Study Based on Medical Records Analysis

The paper proposes a method to assist a physician in establishing a diagnosis for a patient based on preliminary analysis of patient's medical records. Medical records for patients contain all the information related to the health state of each patient. The registered information in a medical records for a patient is useful and usually gives relevant feedback for future health issues based on medical history. To overcome this objective, the paper presents a particular method of text analysis. This method is based on topic modeling and document clustering in order to automatize the process of extracting relevant information. The main idea of the experiment is to extract the relevant content of a topic, meaning clustering in groups works that are interconnected. To establish the relevance of a topic we use the metric topic coherence score. This metric is one of the most important method used to estimate the number of topics. We intend to use these conclusions as an input for a physician, in order to make decisions for a treatment and take future steps in improving a patient's health.