source: Georgia State University
New research from Georgia State University’s TReNDS Center may lead to early diagnosis of devastating conditions such as Alzheimer’s disease, schizophrenia and autism — in time to help prevent and treat these disorders more easily.
In a new study published in Scientific Reports A team of seven scientists from Georgia has built a sophisticated computer program that has combed through vast amounts of brain imaging data and discovered new patterns associated with mental health conditions.
Brain imaging data came from scans using functional magnetic resonance imaging (fMRI), which measures dynamic brain activity by detecting small changes in blood flow.
“We built artificial intelligence models to interpret large amounts of information from fMRI,” said Sergey Bliss, associate professor of computer science and neuroscience at Georgia State, and lead author on the study.
Compare this type of dynamic imaging to a film—as opposed to a snapshot like the more common X-ray or structural MRI—and note that “the available data is much larger, and much richer than a blood test or a regular MRI. But that is the challenge—the vast amount of The data is difficult to interpret.”
In addition, fMRI in these specific conditions is expensive, and it is not easy to obtain. Using an artificial intelligence model, data can be extracted from normal fMRI. Those are available in large numbers.
“There are large data sets available for individuals without a known clinical disorder,” explains Vince Calhoun, founding director of the TReNDS Center and one of the study’s authors. The use of large but unrelated available data sets improved model performance on smaller specific data sets.
“New patterns have emerged that we can definitively link to each of the three brain disorders,” Calhoun said.
The AI models were trained for the first time on a data set of more than 10,000 individuals to learn how to understand basic functional MRI and brain function. The researchers then used multi-site data sets of more than 1,200 individuals including those with autism spectrum disorder, schizophrenia and Alzheimer’s disease.
How it works? It’s a bit like Facebook, YouTube or Amazon learning about you from your online behavior, and starting to be able to predict future behaviour, likes and dislikes. The computer program was even able to go home at the “moment” when the brain-imaging data was most likely related to the psychiatric disorder in question.
To make these findings clinically useful, they must be applied prior to the onset of the disorder.
“If we can find signs and predict Alzheimer’s risk in a 40-year-old, we might be able to do something about it,” Calhoun said.
Likewise, if schizophrenia risks can be predicted before actual changes in brain structure occur, there may be ways to deliver better or more effective treatments.
“Even if we knew from other tests or from a family history that someone was at risk of developing a disorder like Alzheimer’s, we still wouldn’t be able to predict exactly when it might occur,” Calhoun said.
“Brain imaging can narrow that time window, by capturing relevant patterns as they appear before the onset of clinical disease.”
“The vision is that we collect a large imaging data set, and our AI models delve into it, showing us what they have learned about some of the perturbations,” Bliss said. “We build systems to discover new knowledge that we couldn’t discover on our own.”
“Our goal is to connect big worlds and big data sets to small worlds and disease-specific data sets and move toward markers that are relevant to clinical decisions,” said Muhammad Mahfuz-ur-Rahman, first author of the study and a doctoral student in computer science at Georgia State.
Financing: This study was supported by start-up funds to SMP and in part by NIH grants R01EB006841, R01MH118695, RF1MH121885, and NSF 2112455.
About this AI and mental health research news
author: Noel Ritz
source: Georgia State University
Contact: Noel Ritz – Georgia State University
picture: The image is in the public domain
original search: open access.
“Interpreting models that explain brain dynamics” by Sergey Bliss et al. Scientific reports.
Interpreting models that explain brain dynamics
Brain dynamics is very complex and yet it holds the key to understanding brain function and dysfunction.
The dynamics captured by resting-state fMRI data are noisy, high-dimensional and not easily explained. The typical approach to reducing this data to low-dimensional features and focusing on features that are more predictive comes with strong assumptions and can miss fundamental aspects of fundamental dynamics.
In contrast, introspection of discriminatively trained deep learning models may reveal perturbation-related components of the signal at the level of individual time points and spatial locations. However, the difficulty of reliable training on data sets with low, high-dimensional sample size and the unclear significance of the resulting predictive markers prevent the widespread use of deep learning in functional neuroimaging.
In this work, we provide a deep didactic framework for learning from dynamic, high-dimensional data while maintaining stable and environmentally valid interpretations.
The results successfully demonstrate that the proposed framework enables learning of resting-state fMRI dynamics directly from small data and captures compact and stable interpretations of predictive features of function and dysfunction.