The study of biomarkers in the brain, powered by cutting-edge machine learning techniques, could redefine how we classify and diagnose mental health conditions, leading to more effective and personalized treatments.
That’s the goal of Yu Zhang, assistant professor of bioengineering and electrical and computer engineering at Lehigh University’s PC Rossin College of Engineering and Applied Sciences. He recently received significant support from the National Institute of Mental Health (NIMH), a division of the National Institute. Department of Health (NIH). Two grants totaling nearly $4 million will fund two projects using brain imaging and machine learning (ML) to explore biomarkers to improve diagnosis and treatment outcomes for patients with mental health disorders. offer.
A biomarker is essentially some type of measurable indication of a medical condition.
The first studies aim to improve treatment of depression. According to the World Health Organization, approximately 280 million people worldwide suffer from this condition. Antidepressants are the main treatment, but only about half of people who take antidepressants benefit from them, says Zhang, who heads Lehigh University’s Brain Imaging Computational Laboratory (BIC Lab). says Mr.
“Traditionally, medical professionals combine behavioral and clinical symptoms to diagnose depression, but those symptoms are highly subjective, leading to great heterogeneity among patients,” he said. say. “Our goal is to use brain imaging and machine learning to build objective biomarkers that better capture brain dysfunction. will be able to predict whether they will respond to drugs based on brain circuitry, helping guide personalized interventions.” ”
Zhang and his team include collaborators from the University of Texas at Austin’s Dell School of Medicine (Dell Med), the University of Pennsylvania’s Perelman School of Medicine (PSOM), and the Stanford University School of Medicine to help establish biomarkers. A double-blind, randomized, placebo-controlled clinical trial. Data collected from patients before treatment, such as functional magnetic resonance imaging (fMRI) and electroencephalography (EEG), are used to train machine learning models to identify biomarkers in the brain.
The biomarkers we are looking for are not characterized by single brain regions, but by interactions between different regions and brain imaging modalities. We study large brain networks associated with various psychiatric disorders, primarily related to cognitive working memory and emotional regulation. We hypothesize that interactions between these unique brain networks may reveal informative biomarkers that can predict treatment response on an individual level. ”
Yu Zhang, Assistant Professor of Bioengineering and Electrical and Computer Engineering, PC Rossin College of Engineering and Applied Sciences, Lehigh University
Fundamentally, the degree of interaction between networks may indicate how well a person responds to a drug, he says.
Once the team builds the model, it will be tested by conducting an independent clinical trial. Researchers at Dell Med recruit about 50 people diagnosed with depression, prescribe antidepressants, and measure changes in symptoms.
“Then we plan to collect pretreatment brain imaging data and use that data to validate and optimize biomarker discovery,” Zhang said.
He envisions a future where models that can be easily installed on any computer work in concert with portable EEG devices.
Patients in clinics and hospitals have their brains scanned by electroencephalography, and the data is fed into the model. Models use these brain signals to assess connections between brain regions, or strengths and weaknesses of biomarkers, and produce an output that tells a doctor or clinician how likely a person is to respond. increase. Administer antidepressants based on those biomarkers.
Zhang and his team are only looking at selective serotonin reuptake inhibitors (SSRIs), but the ultimate goal, he says, is to fine-tune the model enough to predict how people will respond to other compounds. says.
Their AI-guided biomarkers will not only provide a personalized treatment approach, but will replace current trial-and-error treatment strategies that waste both time and money, he says.
“Time is often more important to patients than money,” says Zhang. “So combining cutting-edge artificial intelligence with brain imaging has the potential to really drive new treatment solutions that help people faster and give them greater confidence in their treatment. It could be a form of precision mental health care that gives hope for
Zhang’s second newly funded study will also use brain imaging data to identify biomarkers, this time redefining the classification of mental disorders.
Currently, mental health conditions are classified based on subjective behavioral and clinical assessments and self-reported questionnaires, says Zhang. As a result, even within one of his diagnostic categories, such as autism, the range of symptoms can be enormous.
“Some patients have very different or heterogeneous symptoms compared to others within that autism category,” he says. “At the same time, across categories such as autism, attention-deficit/hyperactivity disorder, and depression, we find considerable overlap in symptoms, or comorbidities. I think we lack a deep understanding.”
Redefining the classification system could help develop more effective treatments for patients, Zhang said. Currently, patients diagnosed with specific diseases are generally treated with a one-size-fits-all approach. Some patients respond well, others do not respond at all, and some even experience side effects. That’s because their brains work very differently. The more fine-tuned the system, the better it will be able to tailor treatments such as medication, psychotherapy and neuromodulation to specific needs.
Zhang and his team plan to feed brain imaging data and behavioral assessment data into machine learning models to identify brain connectivity patterns. These biomarkers help describe mental health conditions in a more continuum.
“Diagnosis is like a hard label at this point, but we believe that describing these conditions along a spectrum can help identify subpopulations within clinical samples,” he said. increase. “Once we identify these subtypes, we will be able to further study the common brain abnormalities that are unique to them and better understand the treatments that are most useful for that particular subtype.”
Ultimately, the doctor would first collect both brain imaging data and behavioral data from the patient, feed it into the model to learn which subtype the patient belongs to, and then prescribe the appropriate treatment for that subtype. The idea is to move on.
“This research has the potential to redefine the mental health condition and represents a major advance in the field,” Chan said. “This could help us establish more effective treatments for individual patients, something that conventional clinical diagnostics cannot achieve.”
