Guest Lecture by Maarten De Vos (Oxford University, UK)Deep learning for longitudinal neurological monitoring
Many neurological applications suffer from infrequent and subjective monitoring, leaving big questions unanswered around day to day variability and late detection of deterioration. Big data, collected on brain and behaviour, could make neurological monitoring more objective and enrich disease quantification. Analysing those big datasets is however beyond the realm of classical medicine. Machine learning techniques and in particular deep neural networks have revolutionalised applications such as image classification and speech recognition where large quantities of annotated data are readily available.
We aim to leverage the potential of deep learning. to generate novel insights and improve clinical management of neurological and sleep disorders. We will discuss how we can address current challenges around high levels of noise present due to collecting data under variable environmental conditions, how to deal with missing data and focus in particular on providing benchmark algorithms for automated longitudinal sleep staging. By abandoning the traditional formulation of sleep staging as a simple epoch by epoch classification problem, but treating it as asequence- to-sequence classification problem that receives a sequence of multiple epochs as input and estimates all of the labels at once, we can significantly improve performance. Applications in longitudinal RBD monitoring will be discussed.
about Prof. De Vos: https://www.bnc.ox.ac.uk/about-brasenose/academic-staff/549-professor-maarten-de-vos
(Copy and paste into your host application - further infos here)