Some bases and applications of topological data analyses in exploring structure in fMRI data
Vaibhav Diwadkar
Brain Imaging Research Division
Dept. of Psychiatry & Behavioural Neurosciences
Wayne State University School of Medicine
DEIB - Alpha Room (Building 24)
May 27th, 2022
10.30 am
Contacts:
Eleonora Maggioni
Monica Soncini
Brain Imaging Research Division
Dept. of Psychiatry & Behavioural Neurosciences
Wayne State University School of Medicine
DEIB - Alpha Room (Building 24)
May 27th, 2022
10.30 am
Contacts:
Eleonora Maggioni
Monica Soncini
Sommario
On May 27th, 2022 at 10.30 am, Prof. Vaibhav Diwadkar, Brain Imaging Research Division, Dept. of Psychiatry & Behavioural Neurosciences, Wayne State University School of Medicine, will hold a seminar on "Some bases and applications of topological data analyses in exploring structure in fMRI data" in DEIB Alpha Room.
fMRI is the preeminent method for collecting signals from the human brain in vivo.
These signals are then used in the service of functional discovery. However, extracting information from fMRI data is a difficult process, in part because of the high dimensionality and the relatively low information content that is embedded in fMRI signals.
In this presentation, I will provide a working overview of some ideas from Topological Data Analyses (TDA) that position TDA as an important tool for extracting structure from fMRI data. I will discuss the use of persistent homology and persistent landscapes to reveal and describe how complex structures can be extracted from data sets generally. Finally, I will present a working application to a specifically collected fMRI data set (simply motor control with multiple task conditions). I will show that TDA-summarized data can be used to distinguish between different task conditions with a significantly higher degree of classification accuracy (using SVMs) than non-TDA summarized data. I suggest that the programmatic application of TDA to biomedical data can provide an important conceptual bridge between mathematics and medicine.
fMRI is the preeminent method for collecting signals from the human brain in vivo.
These signals are then used in the service of functional discovery. However, extracting information from fMRI data is a difficult process, in part because of the high dimensionality and the relatively low information content that is embedded in fMRI signals.
In this presentation, I will provide a working overview of some ideas from Topological Data Analyses (TDA) that position TDA as an important tool for extracting structure from fMRI data. I will discuss the use of persistent homology and persistent landscapes to reveal and describe how complex structures can be extracted from data sets generally. Finally, I will present a working application to a specifically collected fMRI data set (simply motor control with multiple task conditions). I will show that TDA-summarized data can be used to distinguish between different task conditions with a significantly higher degree of classification accuracy (using SVMs) than non-TDA summarized data. I suggest that the programmatic application of TDA to biomedical data can provide an important conceptual bridge between mathematics and medicine.
Biografia
Vaibhav Diwadkar, PhD is Professor of Psychiatry and Behavioural Neurosciences in Wayne State University’s School of Medicine. He received his Bachelor’s Degree in Computer Science and Psychology (Coe College, Iowa) and his PhD in Psychology (Vanderbilt University, Tennessee). Following Fellowships in Neuroimaging and Clinical Neuroscience at Carnegie Mellon University and the University of Pittsburgh, he joined the faculty in the Dept. of Psychiatry at the University of Pittsburgh in 2004 before moving to Wayne State University in 2006. His research lies at the intersection of several fields including psychiatry, psychology, neuroscience, computer science and mathematics.
Scientific area: Image-Signal Process
Scientific area: Image-Signal Process
