Real-time Analysis of Patient Cohorts

Abstract

In today’s medicine, many complex diseases, such as cancer, are treated with a standardized therapy plan. However, this standard plan will usually not work in the same way for everybody. Researchers assume that this might be due to certain variants in patients’ genes. Currently, researchers analyze large groups of patients with the statistical language R to investigate such cases further. But R is only a language and cannot store data, so that the data has to be loaded and often needs to be converted into the right format. In-memory databases could provide an alternative since they can store and handle a huge amount of data and allowanalyzing them using the Structured Query Language. In this work, I propose to use the integration of statistical analysis in the database in a cohort analysis environment. I will show how an in-memory database can be used to analyze patient groups on the basis of k-means and hierarchical clustering. Therefore, I will compare the performance of clustering algorithms executed within an in-memory database and the statistics environment R. To give researchers the possibility to use the algorithms I developed a prototype, which provides a visualization of the clustering results and parallel clustering with several genes. In the future, it could be possible to perform a statistical analysis in the patient cohort analysis field using an in-memory database and save time for loading and preparing data. Thus, patient cohorts could be analyzed directly in the database without a system change.