AI Trained With More Than 10,000 Brain Scans Can Identify Patterns Linked to Autism, Alzheimer's, Schizophrenia, and More Before Symptoms Appear

Researchers developed a complex AI system that can spot mental health conditions by sifting through a massive amount of brain imaging data to locate new patterns linked to autism, Alzheimer's, and schizophrenia before symptoms set in.

AI Spots Patterns of Mental Health Conditions

The AI model was first trained with over 10,000 brain images from healthy adults. Then it was shown brain images of adults diagnosed with mental health issues, allowing the system to identify changes in the structure that often go unnoticed. The brain imaging data originated from scans using functional magnetic resonance imaging that measured dynamic brain activity via changes in blood flow.

The system was developed by a research team led by scientists from Georgia State who notes that the AI model could one day detect Alzheimer's in patients as young as 40 years old, roughly 25 years before symptoms of the disease begin to appear, reports DailyMail.

By catching early signs of the disease, patients would be able to receive the right treatment that could lessen the effects and strain of the mental illness.

The AI system was demonstrated in the study published in the journal Nature Scientific Reports, titled "Interpreting models interpreting brain dynamics."

Sergey Plis, lead author and an associate professor of computer science and neuroscience, says that the team built the artificial intelligence model to interpret large data from fMRI. He compares dynamic imaging to movies instead of shots like x-rays or common structural MRI.

Plis notes that the available data is larger and richer than blood tests and regular MRIs. However, that is where the challenge arises; the huge data set makes it difficult to interpret.

Once the AI read basic fMRI, the team fed the system datasets of no less than 1,200 scans of people diagnosed with different mental illnesses. The system was able to spot patterns for three mental diseases.

Predicting Mental Disease Patterns

Authors of the study note that fMRI for identifying mental illness can be expensive, wherein humans must manually comb through large data sets. This is why an AI system for the sole purpose of identification not only cuts costs but also increases the efficiency and accuracy of the process.

Vince Calhoun, a co-author of the study and founding director of the TReNDS Center, says in a statement with Futurity that even if other testing or daily history of a person at risk of the disorders are known, it is virtually impossible to predict when the disorders will occur.

Brain imaging could narrow the time window by catching valuable patterns when they appear before the disease becomes apparent.

Researchers hope that the team can collate large imaging datasets to pore over the AI model. Plis says that the team is building systems to discover new knowledge regarding mental health disorders that would not be possible without the system's help.

Md Mahfuzur Rahman, the first author, says that the team's goal is to bridge the gap between big worlds and big datasets to move toward markers relevant to clinical situations.

Check out more news and information on Tech & Innovation in Science Times.