Persistent Homology; Topological Data Analysis; Alzheimer's Disease; Time-varying Data; Connectivity Analysis; Machine Learning
Brüningk Sarah C., Hensel Felix, Jutzeler Catherine R., Rieck Bastian (2020), Image Analysis for Alzheimer's Disease Prediction: Embracing Pathological Hallmarks for Model Architecture Design, in
`Machine Learning for Healthcare' Workshop at NeurIPS, N/A, N/A.
Rieck Bastian, Yates Tristan, Bock Christian, Borgwardt Karsten, Wolf Guy, Turk-Browne Nick, Krishnaswamy Smita (2020), Uncovering the Topology of Time-Varying {fMRI} Data using Cubical Persistence, in
Advances in Neural Information Processing Systems~(NeurIPS), 33, 6900-6912, Curran Associates, Inc., Red Hook, NY, USA 33, 6900-6912.
Hensel Felix, Moor Michael, Rieck Bastian, A Survey of Topological Machine Learning Methods, in
Frontiers in Artificial Intelligence.
Alzheimer’s disease (AD) is the sixth-leading cause of death for Americans ages 65 years and older. It is an irreversible, progressive brain disorder that slowly destroys memory and, eventually, an individual’s ability to perform even the simplest tasks, such as bathing, feeding, and dressing. Beyond the immediate health consequences, the societal costs are of epidemic proportion, thus making AD a pressing public health and medical problem. With disease-modifying treatment trials still unsuccessful at the present time and only medications to treat symptoms available, an emerging research initiative is to identify approaches to intervene before the damage begins, making it potentially possible to prevent AD.As AD is known to affect brain connectivity, analysing said connectivity can lead to a better understanding of how the disease progresses. The proposed project is in a unique position: on the one hand, advances in imaging techniques have resulted in a large number of data sets available. The Alzheimer’s Disease Neuroimaging Initiative (ADNI), for example, has recruited cohorts that are monitored over a prolonged period of time, using a variety of imaging techniques and reporting numerous clinical scores. On the other hand, recent advances in data analysis methods resulted in the development of topological data analysis (TDA), a new domain that employs methods from algebraic topology and differential topology to study the connectivity of complex data sets in a variety of different application areas. We propose using these methods to analyse brain connectivity based on neuroimaging data. This will necessitate the development of new TDA methods for handling time-varying data-a challenge that is hitherto only partially addressed by previous research in the context of low-dimensional signal processing or time series analysis. Moreover, making use of recent advances in topology-based machine learning, we also aim to analyse time-varying topological changes in a supervised and unsupervised setting in order to (i) build predictive models (for predicting AD severity and progression), and (ii) to detect patient subgroups (with a similar progression of AD).This project will generate new knowledge on AD progression, which will help guide future research into the topic. Since this project will be performed under the auspices of clinicians that are experts in AD, it the results of the predictive models and the patient subgroups can also be linked back to clinical scores and lab measurements, for example. This can lead to insights about (clinical) markers that predict AD progression. Finally, we will make our code and analyses publicly available as an open source project to share them with the research community at large and fostering new collaborations.