a VALIANT Effort | January 2025

January 2025

                               

Friends,

2025 is off to an adventurous start. This past weekend, NEURDY Lab hosted the second annual BrainHack Vanderbilt 2025, a meeting designed to foster interdisciplinary collaboration to advance open, reproducible neuroscience research here at Vanderbilt (and beyond).  
 
The initiative successfully engaged with underrepresented groups in science, with over 70% of participants coming from traditionally underrepresented backgrounds in scientific research. The event received support from 11 Vanderbilt departments and institutes, with VALIANT serving as the largest contributor and providing crucial accelerated computing resources as a key partner.* 
 
The event serves as a hub for collaboration, experimentation, and idea-sharing. In this attendee-driven event, 9 unique Vanderbilt labs pitched 10 projects. Over the weekend more than 100 attendees (both in-person and online) teamed up and worked on projects ranging from improving diffusion data visualization to brain state decoding. You can find all the projects pitched and their final presentations on the event website.  
 
Notable efforts included Foundation Model Adapters for Spectrogram Latent Features Extraction that leveraged pre-trained transformer models to differentiate alertness states from EEG and fMRI spectrogram images. Trial and Error: A Bayesian Approach, which leverages Bayesian models to uncover how past decisions shape behavior. Teams also explored advanced neuroprosthetic control with Dynamic Neural Decoding, harmonized multi-site imaging data with Multi-site fMRI Harmonization, and simplified diffusion-weighted imaging workflows in Less Confusion in Diffusion.  
 
Emerging technologies took center stage as well, with projects like DeepMReye Validation, estimating eye gaze patterns from functional MRI data, and OpenDIVE, a software package improving accessibility to diffusion data visualization. The All Optical Brain-Computer Interface and Foundation Model Adapters for Spectrogram Latent Features Extraction brought exciting new approaches to neural decoding and spectrogram analysis. 
 
VALIANT’s support for BrainHack Vanderbilt 2025 underscores our commitment to fostering collaboration and innovation in neuroscience. We are proud to see Alumni Hall transformed into a space for bold ideas and meaningful progress. Thank you to all the participants and mentors who made this weekend a success!Best wishes,
– Bennett
P.s. The solution to last month’s brain teaser is that the solar day (the time that it takes the sun to complete one loop around the sky) is not constant. Solar days are shorter around the solstices and longer around equinoxes. For a great explanation with figures, see this Conversation.
* BrainHack 2025 Sponsors include VALIANT, CS, DSI, VBI, VKC, VUSE, BME, ECE, Biological Sciences, Psychology & The Wond’ry

Expanding Boundaries in Neuroimaging and OpenScience

Dr. Roza Gunes Bayrak, Senior Research Engineer and Research Assistant Professor of Electrical and Computer Engineering at Vanderbilt, embodies the spirit of curiosity and innovation that defines VALIANT. With a background in electrical engineering and a passion for brain research, her journey has taken her from hardware research to spearheading transformative initiatives in neuroimaging and open science.

Dr. Bayrak’s fascination with brains was sparked when she took an extracurricular neurobiology class in graduate school and came across the concept neuroplasticity—how the brain adapts and rewires itself even in adulthood—. This newfound inspiration led her to pursue a PhD at Vanderbilt. “Neuroplasticity gave me a sense of wonder about the brain’s ability to change,” she reflects. “Vanderbilt gave me a home to explore that.” During her PhD she explored brain-body interactions through functional neuroimaging and physiological data. Dr. Bayrak developed individual specific computational approaches to improve precision in brain mapping with Dr. Catie Chang in the NEURDY lab.

Dr. Bayrak is also the driving force behind BrainHack Vanderbilt, a hub for open and reproducible neuroscience. The event, part of the global BrainHack initiative, connects from students, to faculty, all brain researchers across disciplines to tackle pressing problems in neuroscience. “Vanderbilt has a big interdisciplinary neuroscience community,” she explains. “BrainHack fosters collaboration by breaking down silos and creating opportunities for dialogue and innovation.” The 2025 BrainHack event, which attracted over 100 participants from 24 departments and institutes, emphasized inclusivity with hybrid participation options and doubled its active participants from the previous year.

In her current role, Dr. Bayrak is focused on developing a comprehensive library for time series data analysis, tailored to neuroimaging. Her team is integrating state-of-the-art pre-trained foundation models and methods for analyzing diverse datasets such as fMRI, EEG, and physiological recordings. “We are leading the initial effort, but his library will be a community-driven effort, designed to evolve and expand beyond a single lab,” she shares. It’s a project that highlights her commitment to both innovation and accessibility in science.

Dr. Bayrak’s connection to Vanderbilt extends beyond the lab. She praises the university’s collaborative environment and resources, crediting them with enabling her to write her first NIH grant and advance her research. “The support here has been transformative for my career,” she says and I’m excited to see how we can push the boundaries of what’s possible in the years ahead.
 

Outside of work, she enjoys exploring Nashville’s natural beauty and indulging in local gems like the vegetarian Indian restaurant Woodlands. She also recommends tubing in Tennessee’s rivers and visiting the state’s many parks—a pastime she enthusiastically recommends to colleagues and students alike.


Alumni Lookout

This story from the field is brought to us by Parth Datar (MS/BS’22, Vanderbilt Computer Science), Software Engineer at Veeva Systems. 
In May of 2022, I completed my Master’s in Computer Science alongside a Bachelor’s in Computer Science and Mathematics at Vanderbilt, conducting my research in Professor Maizie Zhou’s VCBL lab. My research was focused on using techniques in Natural Language Processing to improve the sequencing of human genomes, tackling the problem of Unphased Reads in Haplotype Phasing. The work was, in-effect, creating a proverbially small language around DNA using FastText to discern between different regions and haplotypes. Almost all of the work was performed on ACCRE, Vanderbilt’s computing cluster, which meant that, during Spring Break, I was constantly checking the output files for my jobs from home. Since graduating, I now work as a software engineer at Veeva Systems, a cloud provider for life sciences companies, with my main professional connection to AI being Veeva’s metrics and anomaly detection. Alongside official work, I enjoy working with PyTorch in my spare time, designing models for interesting use cases.
The English language contains potentially a million words, at least according to Merriam-Webster, and it is ever-expanding; whether the words arise from social media platforms, scientific terminology, or even other languages, the utility of word creation is in describing new phenomena. As such, one would imagine there to be a cat and mouse game between static LLMs and human language generation. Dynamic LLMs can attempt to bridge the gap by perusing the web, integrating new phenomena into their context spaces. However, the more outdated the training data for any particular LLM becomes, the less cutting-edge fields can rely on said LLM, so as to not miss approaches which have been formulated in the intervening time.
Software, for instance, requires a subconscious concern for version management, whether the code be internal or external. While software versions can be autonomously managed, code
debt can accumulate through widespread use of APIs and libraries which can be considered deprecated or even vulnerable. While manmade systems have accumulated these debts incognizant of AI, it appears that the AI trained on manmade code is not free of this concern.
Individual models may vary in this regard; in dealings with GPT-4 for example, one can easily run into deprecated or even hallucinatory APIs. While it is clear that many of these models are general models, the unsightly corollary appears that the further one specializes in work, the more distant the utility of general models becomes.
I recommend that students at Vanderbilt immerse themselves not simply in the use of AI, but in the underlying mathematics; models may be written without mathematics, but I’m unsure whether they can be understood without it. As someone who graduated months before ChatGPT was released, I can only wonder what the future of AI that Vanderbilt holds; I am certain it will be illuminating.

VALIANT Ventures

Our scientists are breaking new ground:
  • Dr. Huo’s team was selected to lead the Nashville CVPR Medical Computer Vision workshop in June 2025.
  • Led by NVIDIA, Dr. Landman’s team was selected to lead the Nashville CVPR Multi-modality and Medicine
    workshop
    in June 2025.
  • Dr. Maizie Zhou received Scaling Success award for “Next generation of spatial transcriptomics analysis tools to understand somatosensation development.”
  • Dr. Landman was elevated to Fellow of the IEEE for  contributions to MRI harmonization—the integration of diverse imaging data acquired across multiple sites—and data-driven modeling in medical image processing.
  • Dr. Landman joined Vanderbilt-Ingram Cancer Center (VICC) as a member.

A VALIANT Attempt at… Cookies & Crochet for VAL-entine’s Day!

Mark your calendars for February 11th at 5:30 PM for Cookies & Crochet for VAL-entine’s Day in FGH 203! Learn to crochet a coaster and enjoy some cookies! No experience required – we will supply you with a hook and yarn and teach you how to get started!

AI Fellow Tkaczyk awarded Presidential Early Career Award for Scientists and Engineers (PECASE)

Eric Tkaczyk, MD, PhD,leverages biophotonics, 3D imaging, and advanced AI techniques to pioneer noninvasive diagnostic technologies that integrate patient care and clinical innovation across dermatology and other specialties.

Diving Deep

Nancy Newlin continued our chalk series with discussion on tractography/connectomics using diffusion MRI

VALIANT Applied AI in Industry and Business Workshop

Save the date: March 27th from 3-4 pm. Join represenatives from theCenter for Entrepreneurship(Owen School of Business)  Center for Technology Transfer and Commercialization(CTTC) for key tips on navigating intellectual property, disclosures, and internships. More details to follow. 

VALIANT Student Group Recognition

VALIANT Attempt has received recognition as the Vanderbilt organization for the undergraduate and graduate affiliates of VALIANT. Join today through AnchorLink

 

 

 


Alchemists’ Corner

We have a lot going on right now. Here is what has hit Scopusfrom our community since last month.

valiant@vanderbilt.edu | https://vanderbilt.edu/valiant | @VandyValiant | LinkedIN

“Vanderbilt” and the Vanderbilt logo are registered trademarks and service marks of Vanderbilt University.

Explore Story Topics