Rosie Cowell and David Huber, neuroscientists in the Department of Psychological and Brain Sciences, have been awarded a $2.36 million grant from the National Institutes of Health to develop a mathematical model of the blood-oxygen-level dependent (BOLD) signal used in fMRI. Researchers from UC San Diego and MIT will also collaborate on the project. Funding comes as part of the BRAIN initiative, striving to create new technologies that will enhance our understanding of the human brain. This research project seeks to bridge across different scales of neurobiological mechanism.
Functional Magnetic Resonance Imaging (fMRI) is a non-invasive experimental technique which measures brain activation with a spatial resolution of around 2mm, such that the signal recorded with fMRI averages over the responses of very many individual neurons in the brain. The BOLD signal can tell you many things, such as which parts of the brain are reacting during a task. In the visual cortex, it can tell you what kinds of visual input the brain likes to respond to, by examining which visual stimuli produce a stronger signal.
Cowell and Huber will endeavor to model that signal and how it changes, developing a procedure to understand the way in which fine-grained neural responses are mapped onto the coarse-grained signals recorded with fMRI. The plan is that this new mathematical model will eventually be provided to other neuroscientists, allowing them to more accurately and precisely infer the properties of neural-level responses from fMRI data.
Cowell explains, “The ultimate goal of neuroscience is to relate what neurons do to how humans perceive the world, record memories, or how we behave. We relate single neurons to behavior. When you measure a person’s BOLD signal in a scanner, you’re looking at massive numbers of neurons aggregated into a very sluggish blood flow signal. What we really want to record is more fine-grained activity coming from individual neurons. Neuroscientists want to measure in humans, non-invasively, the relationship between neurons and behavior.”
The smallest chunk of the brain from which the fMRI scanner can read out a signal is a voxel, or cube measuring about 2mm on all sides. This cube likely contains millions of neurons. To find out what is really going on at the smallest scale researchers need to figure out a way to calculate what subpopulations of cells within the voxel are doing. Cowell and Huber want to find out how the neural sub-populations in this area are responding to visual stimuli. They will use data analysis techniques to untangle a great mixture of different activity among cells.
In the first year, Cowell and Huber hope to take a simple test case that involves displaying simple visual stimuli to human subjects, and compare their mathematical model to what they already know the brain should be doing in this case. This previous brain data comes from decades of research by the scientific community on non-human primates. John Serences of UC San Diego, one of two collaborators in the project, will provide fMRI consulting for the project, providing advice on scanner operations and the application of computational models to fMRI data. Earl Miller, a neurophysiologist at MIT, will provide a source of non-human primate data acquired from past studies. His team will reanalyze data to give Cowell and Huber a ‘ground truth’ against which to compare their computational models.
Huber notes, “In animals, scientists can record from one cell in isolation, which responds to some things but not others. The temptation is to believe that when a voxel is activated or not, it is performing in the same way as a single cell. But we know there are actually millions of cells within each voxel that have different response profiles.”
Further, “When we record from many voxels at the same time, our model assumes that different voxels differ in terms of how much each neural sub-population contributes to the voxel. Using the simultaneous constraint across voxels and stimuli, we can infer the nature of these sub-populations.”
Cowell and Huber’s team will analyze data coming from many healthy adults to see how the brain responds to visual stimuli under different conditions, and how these responses are linked to behavior. fMRI scanning will be done on-campus at the Human Magnetic Resonance Center within the Institute for Applied Life Sciences. Participants in the study will be inside the scanner while viewing images projected into their view through a mirror. They will also perform button presses on a controller, responding to simple questions or cues.
The researchers hope to provide a tool that the whole neuroscience community can use. If they are successful, they will be able to link fMRI imaging and an accurate neural-level of measurement to the behavior associated with visual perception.