Technical Evaluations of AI Algorithms

Full course description

This module is the second module within the learning Data science in Healthcare, and builds on module 1 ‘Data and technology in healthcare’. The module is organised around a real-life clinical example/ problem. Students are trained to obtain and process "rich data" based on relevant (medical) data and data sources in healthcare: to choose and implement machine learning algorithms: to solve clinical or health problems and challenges. Given a specific question, students will first look at what types of data are needed and what requirements must be set to assure good data quality. Based on the fundamentals from Module 1, AI algorithms are discussed and their specific advantages, disadvantages and most appropriate use cases are presented. The focus will be on the thorough understanding of methods as opposed to an exhaustive list of all available algorithms.

Students learn how to select an appropriate artificial intelligence algorithm. Particularly for healthcare settings, limitations and usefulness must be carefully weighted to ensure the best possible outcome while preserving trust by clinicians and patients. Students learn details of various principles of data processing and various tests to determine data quality (such as dealing with missing data and various forms of bias). Concepts such as repeatability, generalisability, transferability, accuracy, reliability, sensitivity and specificity and the difference between training, testing and validating algorithms will become part of the students’ vocabulary. Students learn various validation methods to determine the internal and external validity of AI algorithms and determine the performance on the basis of international reporting standards.