Artificial Intelligence for Biomedical Data

Brief Description

The course focuses on developing models to aid clinical decision-making by applying artificial intelligence methods to biomedical data of various types. It starts with traditional machine learning models based on specific features derived from biomedical data (feature engineering) and then moves on to more current approaches based on deep learning that use “raw” biomedical data (signals or images) to learn the descriptors or features of interest. As a final stage, different methods of Explainable Artificial Intelligence (XAI) are studied, with the aim of identifying the information that most influences the model’s decision. This allows for a justified interpretation of the results for clinical decision-making, error control, model improvement, and the discovery of new knowledge about underlying biological problems.

Objectives

Upon completing the course, the student should be able to:

• Create complex data structures with information on the health or clinical condition of individuals/patients (biomedical signals and images, clinical data) as input for various types of artificial intelligence algorithms.
• Implement, optimize, and apply automated methods from different artificial intelligence approaches and architectures to address needs and problems related to the prevention, diagnosis, and treatment of diseases.
• Determine the most appropriate artificial intelligence approach and algorithm for solving various types of problems in the clinical field, from diagnosis and treatment to disease prevention.
• Deepen their understanding of the fundamental principles of design and operation of artificial intelligence techniques to obtain optimal, robust, and generalizable automatic processing models/systems.
• Integrate the designed automatic models into clinical environments, updating and adapting them to real clinical practice.
• Delve into the reasons that lead automatic models to make their decisions, in order to identify the most relevant regions (signals and images) and patterns (biomedical data) and act on them for the benefit of patients.