Artificial intelligence (AI) tools have become increasingly popular in recent months, with people using them to do different things, ranging from writing stories to planning travel itineraries. What if there was potential to use this technology in neuroscience?

The goal of the Dynamic Brain Circuits in Health and Disease Research Excellence Cluster’s 2023 hackathon was to explore how large language models (LLMs) such as ChatGPT can aid brain health research.

Using AI algorithms, LLMs have the potential to help researchers in brain research in a variety of ways. This includes writing code for data analysis, generating synthetic data, and assisting in generating text for research proposals, questionnaires and ethics applications.

Over the weekend of March 24-26, 34 participants on six teams came together from different institutions across the Cascadia region, participating both in person and virtually.


First place winner: PsycheCloak

Daniel Ramandi, a PhD student in Dr. Lynn Raymond’s lab, and his team wanted to use LLMs for analyzing psychological interviews but couldn’t find any open-source datasets, making collaboration with other researchers in the field challenging due to a lack of proper anonymization and ethical concerns. As a result, they decided to pivot their project towards finding a way to anonymize the interviews instead.

“We trained PsycheCloak on synthetic interview data from ChatGPT for anonymizing in-depth psychology interview transcripts while preserving data integrity and protecting participant privacy.” Daniel explained.

Unlike traditional ‘find and replace’ methods that lead to generic placeholders and hinder natural language processing analysis, PsycheCloak is unique as it retains the richness and context of the original transcripts, ensuring that the data remains meaningful and easy to analyze.

“The most challenging aspect of the hackathon was the limited time we had to develop our project,” said Daniel. “However, the most rewarding part was the opportunity to collaborate with participants from different groups and backgrounds, who were all willing to help each other despite being in competition.”

The next steps for the project include fine-tuning the model, conducting rigorous evaluations and preparing a manuscript for publication.

“We are committed to making PsycheCloak openly available to all researchers, psychologists and psychiatrists,” Daniel said. “We hope that everyone can benefit from our innovative solution for anonymization.”


Second place winner: Brain Chat: A Brain Diagnostic Assistant Tool

A few days before the hackathon, Tanmay Agarwal, a UBC Master of Data Science student, was chatting with his brother, a neurologist, about ChatGPT. He mentioned that it performed poorly in certain specific use cases, such as identifying brain localization. During the hackathon’s brainstorming session, Tanmay proposed this idea to his team and it was well-received by everyone.

The result was Brain Chat, a dashboard that combines a chatbot and a 3D plot to assist doctors in diagnosing and treating patients with neurological disorders. The chatbot component provides information on which parts of the brain may be affected based on the symptoms provided by the user. The 3D plot component allows the user to locate the affected areas of the brain, view the cross-section of the brain and understand the location of the lesion in a visual format.

“We only had two days to implement this idea into a proof of concept so we had to divide the tasks efficiently and work diligently,” said Tanmay. “In the end, I am glad that we successfully developed the tool we initially aimed to create.”

While most of his team members are currently preoccupied with their accelerated Data Science program, Tanmay hopes they will be able to present their findings at a conference in the near future.


Second place winner: MoCA-BOT: AI-Powered Self-Administered Cognitive Assessment

MoCA-BOT is a chatbot capable of administering selective portions of the Montreal Cognitive Assessment (MoCA), like the Language Task, Memory and Delayed Recall Task, Abstraction Task, Naming Task, Attention Task and Orientation Task. It scores the user based on their responses, evaluates the final score against the grading rubric and recommends additional medical interventions and resources for substandard scores.


Third place winner: ReMind

ReMind is geared to improve the emotional well-being and independence of people with dementia and their caregivers by initiating adaptive conversations with the patient and creating a positive emotional atmosphere, helping them remember important information (like names of their loved ones) and providing feedback to caregivers on their interactions with the patient.

Congratulations to all the winning teams! Special thanks to Dynamic Brain Circuits cluster staff Jeff LeDue and Amy Wang for their hard work in planning and organizing the event.

Learn more about the other projects that were part of the Hackathon.