Measuring what matters
While brain imaging and other biological tools have attracted significant attention, several speakers said that understanding behavior may be just as important as measuring the brain itself.
Halassa, who is also a professor of psychiatry and behavioral medicine with the Virginia Tech Carilion School of Medicine, emphasized a lesson emerging from both neuroscience and artificial intelligence research.
“It’s impossible to interpret the brain outside of behavioral context,” he said.
Researchers can reliably interpret neural activity when an animal or person is engaged in a structured task, he explained. Without understanding behavior, biological measurements become far more difficult to interpret.
“The challenge is to develop a task structure that is well controlled enough but also along the line of how we think psychosis works,” Halassa said.
Chiu echoed that perspective, arguing that future breakthroughs will likely require integrating multiple kinds of information rather than relying on a single biomarker.
“We can’t take any single stream of data, but we need to combine it,” she said.
From measurement to treatment
A common thread during discussions was how to turn objective measurements of brain function and behavior into information that can guide treatment.
Read Montague, director of the institute’s Center for Human Neuroscience Research, said that although speakers approached the problem from different angles, computation emerged as a common theme.
“You can’t get precise until you first learn how to compute something that gives you a precise estimate,” Montague said.
For decades, psychiatry has relied heavily on clinical observations and patient reports. While those assessments remain essential, researchers increasingly hope to complement them with biological measurements, computational models, brain imaging, and other quantitative data.
“When your readout measures are qualitative assessments by clinicians, there’s going to be disagreement,” Montague said.
The goal, he said, is to build objective measures and computational frameworks capable of identifying patterns that may be difficult for any individual clinician to detect.
“Going into a psychiatrist and having a quantitative measure, biological measure, and then having a computation sitting around it to interpret it and sort it through,” he said, “that’s where the field is headed.”
Montague said one of the biggest challenges ahead will be gathering the large, high-quality datasets needed to train those systems.
“The hurdle is collecting the data at a scale that’s relevant to what we want to solve,” he said.