Dr Charlotte James

Research Fellow, University of Exeter

Methods Optimisation Working Group Lead, DEMON Network

Title: Using machine learning to predict dementia incidence in memory clinic patients

Abstract: Machine learning algorithms may be used as clinical decision making aids to enhance clinical practice. Using a cohort of memory clinic patients from the US, we will demonstrate that machine learning algorithms can predict incident dementia within 2 years more accurately than existing clinical decision making aids, require relatively few predictors, and may reduce misdiagnosis by up to 94%.

Dr Razvan Marinescu

Postdoctoral Researcher in Medical Image Computing, Massachusetts Institute of Technology

Title: Transfer learning from typical Alzheimer's disease to rare dementias using Disease Knowledge Transfer

Abstract: Population studies in rarer neurodegenerative diseases often suffer from limited datasets. While training sophisticated Machine Learning models on such data is difficult, transfer learning is a key potential solution. We use Disease Knowledge Transfer to infer robust multimodal biomarker trajectories in rare neurodegenerative diseases even when only limited, unimodal data is available, by transferring information from larger multimodal datasets of common neurodegenerative diseases. This allows the estimation of plausible, multimodal biomarker trajectories related to amyloid, tau, glucose metabolism and white-matter degradation in Posterior Cortical Atrophy given only unimodal MRI data.

Click here to join the seminar

DEMON Network Seminars are available for DEMON Network members only. Joining the Network is quick and free. If you are not yet a member of our Network, please visit www.demondementia.com/join to sign up before registering.

Machine Learning and Disease Knowledge Transfer