Approved Research

Explainable and interpretable machine learning solutions in computational medicine

Lay summary

In an increasingly data-driven world, artificial intelligence is expected to be a key tool for converting big data into tangible benefits and health care is no exception to this. Machine learning, a subfield of artificial intelligence, aims to identify complex patterns in multi-dimensional clinical data and use these uncovered patterns to classify new unseen patients or make predictions about disease progression in patients. However, one major issue preventing the wide-spread application of machine learning models in health care is the black-box nature of these models. The main objective of our proposed research using UK Biobank data is to develop machine learning models that can explain their decision-making process to patients and health care professionals in intuitive, understandable ways. Having access to the vast imaging and clinical data curated by UK Biobank will allow us to develop and evaluate novel explainable machine learning solutions for many highly relevant health-care problems. This will improve the trust of physicians and patients in diagnostic decision support tools developed using artificial intelligence methods.

In this research project, we will develop and evaluate novel methods that can help exploring the complex relationships in clinical data. Previously, we have, for example, adapted these solutions to problems in brain aging that not only answer "How to improve the prediction of brain age?", but also allow us to explore "What morphological changes are observed by this model for a 60-year-old male who has been smoking for past 20 years look like?". Beyond healthy brain aging, we have employed these methods to improve our understanding of neurological diseases and genetic syndromes. Using the UK Biobank data, we hope to extend our research to other diseases while also improving our methods to explain the decision making of artificial intelligence models.

Since we are an actively growing research group and the nature of this research is largely explorative and iterative, we request a rolling three year timeframe to conduct this research. This will allow our research lab to build a strong research program on explainable artificial intelligence for health care.