Interview with Prof. Vince Calhoun, 2024 Winner of the Glass Brain Award
Author: Ashley Tyrer
Editor: Audrey Luo, Elisa Guma, Kevin Sitek , Megan Sheppard, Simon R. Steinkamp
Dr. Vince D. Calhoun is the recipient of the Organization for Human Brain Mapping (OHBM) Glass Brain Award at the 2024 annual meeting held in Seoul, South Korea. Dr. Calhoun is a Distinguished Professor and the Founding Director of the tri-institutional Center for Translational Research in Neuroimaging and Data Science (TReNDS): a joint effort between Georgia State, Georgia Tech, and Emory University. Dr. Calhoun received his Bachelor’s degree in Electrical Engineering from the University of Kansas in 1991, Master’s degrees in Biomedical Engineering and Information Systems from Johns Hopkins University in 1993 and 1996, respectively, and his Ph.D. in Electrical Engineering from the University of Maryland Baltimore County in 2002. From 1993 to 2002 he also worked as a research engineer in the Psychiatric Neuroimaging laboratory at Johns Hopkins University. Between 2002 and 2006, he served as the Director of medical image analysis at the Olin Neuropsychiatry Research Center and as an Associate Professor at Yale University (until 2005). Most recently, he was appointed Distinguished Professor at the University of New Mexico and the President of the Mind Research Network.
Dr. Calhoun’s primary research focus involves developing techniques for elucidating complex brain imaging data. As each brain imaging modality has inherent limitations, integrating data of multiple modalities is required to understand both healthy and disordered human brains. To this end, Dr. Calhoun has generated algorithms that map dynamic networks of brain function, structure and genetics, and investigated how these are affected in different task-based contexts or in individuals with mental illness such as schizophrenia.
The Glass Brain Award recognizes his leading role in the field and his lifetime achievements. We had the honor to interview Dr. Calhoun, reflecting on his scientific journey. If you want to hear more from Dr. Calhoun, he also appeared on Neurosalience together with Dr. Kulman in 2021 and in a recent double episode in 2024 (part 1 & part 2).
1. Which of your scientific contributions are you most proud of?
I am proudest of my contributions to the development of methods to analyze complex, noisy brain imaging data using flexible approaches that relax assumptions and allow the data to have a stronger voice in the resulting conclusions. My work has led to data-driven algorithms that map the dynamic networks of brain function, and has also enabled multimodal fusion of brain function and structure as well as non-imaging data such as genomics, behavior, and environmental data to leverage and integrate shared information. An emphasis on brain-based biomarkers of health and disorder has been a major goal of my group. A focus on algorithm development included pioneering the group Independent Component Analysis (ICA) approach in 2001, which is deployed in the GIFT toolbox and widely used and incorporated into various other software tools. One key thing to note about group ICA is that it is not at its core about groups, but rather the coupling of flexible modeling with inter-individual correspondence that allows us to faithfully estimate individual variation. This approach has proven extremely fruitful and also informs our approach to newer approaches including more flexible multi-layered neural networks.
In 2009, we introduced the chronnectome framework to capture whole-brain connectivity over time, recognized by the NIMH as a top contribution. My group was also among the first to validate and apply deep learning to neuroimaging data in 2014, with a focus on advancing mental health research through machine learning. Our work in multimodal data fusion integrates data from multiple imaging types (EEG, fMRI, dMRI, sMRI), enabling true joint analysis without assuming cross-modal alignment, embodied in the FIT toolbox.
In imaging genetics, we developed the parallel ICA approach to identify links between genetic and imaging data, allowing for hypothesis-driven analysis while uncovering unknown factors. Recent innovations include embedding group ICA within parallel ICA for detailed subgroup analysis and blending joint source separation approaches with deep learning architectures (including convolution neural networks, transformers, and latent diffusion modeling). My decade-long work at Johns Hopkins University deepened my expertise in data-driven approaches for studying mental disorders and applying predictive models to imaging and genetics. My contributions to initiatives such as a schizophrenia classification challenge and early review papers on single subject prediction emphasized mitigating bias in machine learning to maximize generalizable results.
Finally, I also have worked for years on neuroinformatics initiatives by creating systems to ensure collected data is effectively archived, shared, and utilized, fostering collaboration and maximizing the impact of neuroimaging research. This includes a focus on decentralized or federated analysis of neuroimaging data, to broaden the data ecosystem beyond that which is sharable.
2. What has helped you most when you have faced career challenges or obstacles?
I’ve found that concentrating on my research has been the most effective way to navigate career challenges, whether it’s feeling disheartened by negative feedback, a lack of recognition, questioning the importance of my work after seeing impressive new findings (a bit of imposter syndrome!), or facing slow progress in certain areas. Engaging with the data reignites my enthusiasm and helps me see new possibilities, reinforcing my belief in the importance of my work I’m doing. I experienced a major shift during my PhD when I changed direction and did not follow through with my original plan (my first journal paper was never submitted!). Instead, I pivoted to a full-time research engineering role in a psychiatric neuroimaging group with a new mentor, where I thrived and began focusing on human MRI and fMRI acquisition and analysis. This taught me the value of finding an environment that appreciates what you enjoy doing and that has an ‘impedance match’ with your gifts, skills, and interests. While perseverance is necessary to overcome challenges, long-term resilience comes from genuinely loving the work and feeling that you are supporting those around you and contributing to the broader field. Great mentors who you can share ideas with and who can support you are essential. Additionally, my faith as a Christian has been a significant source of guidance and strength, influencing how I engage with others and respond to challenges. While this perspective may not be universally shared, it remains an important part of my journey.
3. What recent advances in the field are you most excited about?
We are getting better at predicting future onset of mental illness or dementia from neuroimaging data. This type of work coupled with brain stimulation has the potential to make a real difference clinically. I think the work teasing apart dimensional measures of mental illness will be really important as we move beyond binary, noisy categories, for labeling disorders. Recent MRI studies showing rich acquisition protocols such as echo planar time-resolved imaging and extensions are opening up new windows into brain function that dovetail well with advanced machine learning approaches. Similarly, detailed studies of cell types, such as those from the Allen Brain Institute, can help us further unravel the complexity of the brain, though much more work and data are needed. Brain decoding studies are also still in the early stages, but show great promise to help us understand how the brain is working and how it may represent information. I’m also excited about the recent layer-specific studies, as well as studies focusing on multimodal brain networks. The new work on foundational models for the brain are interesting, but still have a long way to go. I think it will pay off in certain cases as models improve.
4. How has the field changed during your career? And how do you see the field evolving in the next few years?
The field has dramatically changed in so many ways: new imaging technology (faster imaging, new types of contrast, mobile MRI, and so much more), brain stimulation efforts, the use of big data and machine learning approaches, data sharing efforts, the focus on replicability, the onset of wide (large N) and deep (extensive phenotyping or extended data collection) studies, and the availability of tools for automated preprocessing and analysis. I would have liked to see more advances in the clinical applications; we are starting to see some of this potential manifest in a few groundbreaking studies and ongoing clinical trials, but the complexity of brain disorders and the macroscopic scale of human neuroimaging has made this challenging.
5. Do you have any advice for young researchers?
I would encourage them to embrace innovation and not shy away from challenging established norms. It's important to recognize and question our own biases and assumptions so as not to limit potential research avenues just because others believe they have been thoroughly explored. Additionally, keeping an eye out for creative ways to merge existing research areas—integrating different disciplines and methodologies can lead to unexpected and exciting breakthroughs. Lastly, find opportunities to be helpful to those around you while pursuing your interests and advancing your research. Though there’s no definitive path to success, collaborating and working synergistically can yield surprising and rewarding results.
6. If you could start a whole new research program right now, what would be the most exciting new direction for you?
It is hard to pick, there are a lot of exciting directions that could be pursued. On the development side, I think there is a lot of untapped potential in the combination of MRI pulse sequences with flexible modeling. I’m also excited about the potential for higher spatial resolution, we are not quite there yet, but if we can perform whole brain functional imaging with less than 0.5mm voxels, this opens some new possibilities for studying neural activity. I also think multimodal approaches are still in their infancy, or perhaps pre-adolescent stage; we still have a lot to learn about how to combine and use the complementary information we gain from combining both static and dynamic modalities. On the application side, we have so much data now, but only for certain easily scalable modalities. Nonetheless, there is great potential in leveraging these data to better inform brain mechanisms and develop biomarkers. A major focus on improving our ability to visualize and interpret flexible nonlinear models could make a huge impact.