Artificial Intelligence in Healthcare: Computer-aided diagnosis of dementia

Dementia refers to a range of diseases that affect memory, communication, behavior, and the ability to perform daily activities. Making the diagnosis of dementia in an early stage of the disease is very challenging as symptoms may be unclear. On average it takes two years after a patient’s first visit to a memory clinic before the definitive diagnosis is made. In this talk, I will show how artificial intelligence can improve the diagnosis of dementia.

Dementia is a brain disease in which nerve cells are damaged. With MRI, the changes in the brain of dementia patients can be measured and quantified. To make the diagnosis of an individual patient on the basis of measures derived from MRI, artificial intelligence can be used; this is called ‘computer-aided diagnosis’. Computer-aided diagnosis algorithms learn from examples by using machine-learning or other multivariate data-analysis techniques. A model (classifier) is trained to categorize groups (e.g., patients and controls) based on data measurements (features). This model can be applied to new data for making the diagnosis. Such methods have shown to diagnose dementia with an accuracy of 80-90%. These techniques can potentially lead to a more objective and accurate diagnosis than when using clinical criteria, as potentially group differences are used that are not noted when the MRI scans are inspected qualitatively. My research focused on quantitative measurements from MRI scans and combining these with artificial intelligence to make a computer-aided diagnosis system for dementia. I will show you the potential of different MRI techniques, the differentiation between different types of dementia and the results of a competition that I organized to objectively compare diagnosis algorithms from international research groups.

Esther Bron is post-doc researcher at the Biomedical Imaging Group Rotterdam (Erasmus MC, Rotterdam, Netherlands). Her main research interest is advanced analysis of brain MRI for improving diagnostics. Currently she is mainly working on image analysis pipelines for multi-center studies. In addition, Esther is working on disease progression modeling in collaboration with the Progression Of Neurodegenerative Disease (POND) group at University College London. Interested in an MSc project at the Erasmus MC? Check www.bigr.nl or http://openlab.bigr.nl to register for the BIGR Open Lab Day on June 9th.