Summary: Researchers have developed a revolutionary model called Brain Language Model (BrainLM) using generative artificial intelligence to map brain activity and its implications for behavior and disease. BrainLM leverages 80,000 scans from 40,000 subjects to create a fundamental model that captures the dynamics of brain activity without the need for specific disease-related data.
This model significantly reduces the cost and scale of data required for traditional brain studies, providing a robust framework capable of predicting conditions such as depression, anxiety, and PTSD more effectively than other tools. BrainLM demonstrates powerful application in clinical trials, potentially cutting costs in half by identifying patients most likely to benefit from new treatments.
Highlights:
- Generative AI model: BrainLM uses generative AI to analyze brain activity patterns from numerous datasets, learning the underlying dynamics without specific patient details.
- Cost and effectiveness of research: The model reduces the need for large-scale patient recruitments in clinical trials, which could significantly reduce costs by using its predictive capabilities to select appropriate candidates for studies.
- Wide applicability: Tested on different scanners and demographics, BrainLM showed superior performance in predicting various mental health conditions and shows promise in facilitating future research and treatment strategies.
Source: Baylor College of Medicine
A team of researchers from Baylor College of Medicine and Yale University incorporated generative artificial intelligence (AI) to create a fundamental model of brain activity. The Brain Language Model (BrainLM) was developed to model the brain in silico and determine how brain activities relate to human behavior and brain diseases.
The research was published as a conference paper at ICLR 2024.
“We have known for a long time that brain activity is linked to a person’s behavior and to many diseases like seizures or Parkinson’s disease,” said Dr. Chadi Abdallah, associate professor in the Menninger Department of Psychiatry and Science. of Behavior at Baylor and co-corresponding author of the article.
“Functional brain imaging or functional MRI allows us to observe brain activity throughout the brain, but previously we could not fully capture the dynamics of these activities in time and space at the same time. using traditional data analysis tools.
“More recently, people have started using machine learning to capture the complexity of the brain and how it relates to specific diseases, but this has proven to require the recruitment and in-depth examination of thousands of patients presenting a particular behavior or illness, a very expensive process.”
The power of new generative AI tools lies in their use to create fundamental models independent of a particular task or a specific patient population. Generative AI can act as a detective uncovering hidden patterns within a data set.
By analyzing data points and the relationships between them, these models can uncover the underlying dynamics: how and why things change or evolve.
These fundamental models are then refined to understand a range of topics. Researchers have used generative AI to capture how brain activity works regardless of a particular disorder or disease.
This can be applied to any population without needing to know the subject’s behavior, information about their disease, history or age. All it takes is brain activity to teach the computer and AI model how brain activity changes in space and time.
The team performed 80,000 scans on 40,000 subjects and trained the model to understand how these brain activities related to each other over time, establishing the fundamental BrainLM model of brain activity. Now researchers can use BrainLM to refine a specific task and pose questions for further studies.
“If you want to run a clinical trial to develop a drug for depression, for example, it could cost hundreds of millions of dollars because you have to recruit a large number of patients and treat them for a long time.
“With the power of BrainLM, we can potentially cut this cost in half by recruiting only half the subjects by using the power of BrainLM to select individuals most likely to benefit from treatment. So, BrainLM can apply the knowledge gained from the 80,000 analyzes to apply it to these specific study topics,” said Abdallah.
The first step, preprocessing, summarized the signals and removed noise that is irrelevant to brain activity. The researchers fed the summaries into a machine learning model and masked a percentage of the data for each person. When the model learned the dynamics, they tested it on an excluded test group.
They also tested this on different samples to understand how well the model could generalize to data acquired with different scanners and in different populations, such as older adults and young adults.
They found that BrainLM performed well in various samples. They also found that BrainLM predicted depression, anxiety, and PTSD severity better than other machine learning tools that don’t use generative AI.
“We have found that BrainLM works very well. It predicts brain activity in a new sample that was hidden from it during training and works well with data from the new scanners and the new population,” Abdallah said.
“These impressive results were obtained with scans of 40,000 subjects. We are now working to significantly increase the training dataset.
“The stronger the model we can build, the more we can do to facilitate patient care, such as developing new treatments for mental illnesses or guiding neurosurgery for seizures or DBS. »
The researchers plan to apply this model to future research aimed at predicting brain-related diseases.
About this research news in AI and neuroscience
Author: Homa Warren
Source: Baylor College of Medicine
Contact: Homa Warren – Baylor College of Medicine
Picture: Image is credited to Neuroscience News
Original research: Results will be presented at ICLR 2024
News Source : neurosciencenews.com
Gn Health
