[Brainmap]: Wenzhu Mowrey, PhD - Sparse k-means clustering with resampling: A method developed to define amyloid-positivity among cognitively normal elderly

Wednesday, July 19, 2017 - 12:00 to 13:00
149 13th Street (Building 149), room 2204

 

Abstract:

High dimensional datasets are commonly generated in a large variety of scientific disciplines such as genetics and imaging resulting from the difficulties of subject recruitment and/or the financial burden of the actual data collection.  Since some datasets arise from the development of new technologies, for example, the development of the Pittsburgh Compound-B (PiB) Positron Emission Tomography (PET) imaging agent allowed for the novel in vivo measurement of amyloid plaque deposition in living human brain, there is little knowledge of groups and/or subsets of subjects within the population that may be of interest, requiring the use of unsupervised learning techniques such as clustering methods.  We propose a method to add resampling onto sparse k-means clustering to improve upon the current clustering methodology. The addition of resampling methods to sparse k-means results in variable selection that is more accurate.  The method is also used to assign a "cluster membership probability" to each observation, providing a new metric to measure membership certainty.  The performance of the method is studied via simulation and illustrated in the motivating imaging data example where the goal was to identify amyloid positive subjects among the cognitively normal elderly people based on their PiB PET brain scan images.

 

About the Speaker:

Dr. Mowrey received a Bachelor's degree in Electrical Engineering from Shanghai Jiao Tong University, a Master's degree in Computational Mathematics from Duquesne University, and a PhD degree in Biostatistics from University of Pittsburgh.  She also received training in neuroscience as a PhD add-on program from the Center for Neural Basis of Cognition, jointly by Carnegie Mellon University and University of Pittsburgh.  She works as an assistant professor at Albert Einstein College of Medicine at New York City.  Her statistical methodology interests include analysis of neuroimaging data from all modalities (PET, MRI, fMRI, DTI, EEG, MEG and optical imaging), sparse clustering methods, dimension reduction of high dimensional data, survival and longitudinal data analysis.  She collaborates on projects to study aging, Alzheimer's disease, epilepsy, Rett syndrome, rheumatology and infectious diseases.