Keynote


Dr. Tianming Liu

Title: Discovering Hierarchical Organizational Brain Architecture via Deep Learning

Abstract: After decades of active research, there has been mounting evidence that the human brain function emerges from and is realized by the interaction of multiple concurrent neural networks. To faithfully locate and reconstruct the brain regions and networks, a variety of model-driven approaches or data-driven approaches have been proposed for fMRI data modeling and analysis. However, a fundamental limitation of existing fMRI data modeling approaches is that they only build shallow models, and they are under the strong assumption that fMRI signal could be linearly decomposed into time series components with their corresponding spatial maps. As growing evidence shows that human brain function is hierarchically organized, new computational approaches that can infer and model the hierarchical organizational structure of brain function are much needed. This talk will introduce and discuss a variety of recent deep learning models that reveal the hierarchical organizational brain architectures using fMRI data. These deep learning models, including Deep Convolutional Auto-Encoder (DCAE), Volumetric Sparse Deep Belief Network (VS-DBN), and Deep Sparse Recurrent Auto-encoders (DSRAE), offer great promises to produce both theoretic understanding and concrete representation of the hierarchical organizational architecture of brain function.

Biography: Dr. Tianming Liu is a Distinguished Research Professor (since 2017) and a Full Professor of Computer Science (since 2015) at University of Georgia (UGA). Dr. Liu is also an affiliated faculty (by courtesy) with UGA Bioimaging Research Center (BIRC), UGA Institute of Bioinformatics (IOB), UGA Neuroscience PhD Program, and UGA Institute of Artificial Intelligence (IAI). Before he moved to UGA in 2008, Dr. Liu was a faculty member of Weill Medical College of Cornell University (Assistant Professor, 2007-2008) and Harvard Medical School (Instructor, 2005-2007). Dr. Liu was a postdoc in neuroimaging in the University of Pennsylvania (2002-2004) and Harvard Medical School (2004-2005). Dr. Liu received PhD in computer science from Shanghai Jiaotong University in 2002. Dr. Liu is the recipient of the Microsoft Fellowship Award (2000-2002), the NIH Career Award (2007-2012) and the NSF CAREER Award (2012-2017). He is a Fellow of AIMBE (inducted in 2018) and was the General Chair of MICCAI 2019.


Dr. Kilian M. Pohl

Title: Modelling Confounding Effects within End-to-End Learning

Abstract: The presence of confounding effects (or biases) is one of the most critical challenges in using deep learning to advance discovery in medical imaging studies. Confounders affect both the input data (e.g., brain MRIs) and output variables (e.g., diagnosis). Improper modelling of those relationships often results in data-driven inference identifying spurious and biased associations. Traditional machine learning and statistical models minimize the impact of confounders by, for example, matching data sets, stratifying data, or residualizing imaging measurements. Alternative strategies are needed for state-of-the-art deep learning models that use end-to-end training to automatically extract informative features from images. While invariant feature learning approaches are trying to account for biases in natural 2D images, their application to 3D MRIs has been of limited impact. In this talk, we review the needs of brain MRI studies and propose strategies for learning and visualizing discriminative features from MRIs that are invariant to confounding factors. We review findings of the proposed deep learning approaches on large publicly available data sets (such as ABCD study, > 10K samples) and smaller in-house studies (< 100 samples). The corresponding code is publicly available and can be easily integrated into existing deep learning approaches.

Biography: Dr. Kilian M. Pohl is an Associate Professor of Psychiatry and Behavioral Sciences at Stanford University, has a secondary appointment as Program Director of Biomedical Computing at SRI International, and is the Director of the Computational Neuroscience Laboratory (http://cnslab.stanford.edu). Over 15 years ago, Kilian joined the Artificial Intelligence Laboratory at the Massachusetts Institute of Technology to advance neuroscientific research through the use of machine learning technology and has continuously published in that domain since then. He now focuses on computational science aimed at identifying biomedical phenotypes improving the mechanistic understanding, diagnosis, and treatment of neuropsychiatric disorders.  Kilian is a multiple Principal Investigator of the Data Analysis Resource of the National Consortium on Alcohol and Neurodevelopment in Adolescence (NCANDA) and his research is supported by funding from multiple NIH institutions (in particular NIAAA), Stanford Institute for Human-centered Artificial Intelligence (HAI) AWS Cloud Credit, and private donations.