Configurable In-Database Similarity Search of Electronic Medical Records.

With the development of technology, Electronic Medical Re-cords (EMRs) are widely used for medical analysis through methods such as similarity search. Typical EMRs contain attributes of different data types including string, enumeration and numeric data, and are commonly stored in a database. However, many EMR similarity search algorithms neither separate different data types nor conduct search directly in the database. In addition, for researchers and doctors who need similarity search but do not have strong programming background, a user-friendly interface is missing. Therefore, we design a tool “SIR” to solve the aforementioned problems. SIR can conduct configurable similarity search in high dimensions (within 0.0931 and 0.7824 s respectively using its basic and advanced version), can be embedded directly in the database, and has an intuitive interface.