What Is Functional Magnetic Resonance Imaging?

Functional magnetic resonance imaging (fMRI) is a neuroimaging technique that measures brain activity by detecting the changes in blood oxygenation and flow that accompany neural activity. It is noninvasive and requires no ionizing radiation, radioactive tracers, or injection of contrast agents, which has made it the dominant method for mapping brain function in humans since its introduction in the early 1990s. fMRI sits at the intersection of physics, physiology, biomedical engineering, and neuroscience, and it draws on magnetic resonance physics for signal acquisition, vascular biology for interpreting the signal, and signal processing and statistics for extracting meaningful results from large, noisy datasets.

The technique rests on the coupling between neural activity and local blood flow, a relationship formalized as neurovascular coupling. When a region of the brain becomes active, local blood flow increases over the following seconds, delivering more oxygenated hemoglobin than the tissue consumes. This shift in the ratio of oxygenated to deoxygenated hemoglobin is detectable by MRI because the two forms differ in their magnetic properties.

The BOLD Signal and Hemodynamic Response

The primary contrast mechanism in fMRI is blood-oxygen-level dependent (BOLD) contrast, first identified by Seiji Ogawa and colleagues at Bell Labs in 1990. Deoxygenated hemoglobin is paramagnetic and distorts the local magnetic field, shortening the T2* relaxation time of nearby water protons and reducing signal in gradient-echo MRI sequences. When neural activity triggers a local increase in blood flow, deoxygenated hemoglobin is swept out and replaced by oxygenated hemoglobin, which is diamagnetic and causes less field distortion. The result is a small increase in the MRI signal, typically 1 to 5 percent at clinical field strengths of 1.5 to 3 Tesla.

According to an overview of functional magnetic resonance imaging published in PMC, the BOLD hemodynamic response to a brief stimulus peaks approximately 5 to 6 seconds after onset and has a full width of roughly 3 seconds, imposing a fundamental limit on the temporal resolution with which rapid neural events can be resolved. Echo-planar imaging (EPI) sequences acquire a full brain volume in approximately 2 seconds, providing the temporal sampling needed for most experimental designs.

Data Acquisition and Experimental Design

fMRI experiments are organized around experimental contrasts. In block designs, the brain alternates between sustained periods of task and rest, producing large signal changes that are easy to detect statistically. Event-related designs interleave brief individual trials, allowing estimation of the hemodynamic response to each trial type and randomization of trial order to prevent anticipation effects. The repetition time (TR), ranging from 500 milliseconds to 3 seconds in typical protocols, sets the temporal resolution of the data. Spatial resolution is determined by voxel size, typically 2 to 3 millimeters on a side, which represents a compromise between signal-to-noise ratio and anatomical specificity.

Functional MRI methods reviewed by NIH-funded researchers describe the growing use of resting-state fMRI, in which subjects lie still without performing a task, allowing the measurement of intrinsic low-frequency fluctuations in the BOLD signal that reflect spontaneous neural activity and functional network organization.

Image Processing and Analysis

Raw fMRI data require extensive preprocessing before analysis. Standard steps include slice-timing correction to account for the fact that different slices are acquired at slightly different times within each TR, motion correction to realign images across time points, spatial normalization to a standard brain atlas, and spatial smoothing to improve signal-to-noise ratio. Statistical analysis commonly uses a general linear model (GLM) that regresses the measured BOLD time series against a predicted response based on the convolution of the stimulus timing with the hemodynamic response function. The resulting statistical maps identify voxels where neural activity is significantly modulated by the experimental condition. Independent component analysis (ICA) provides an alternative data-driven decomposition that identifies spatially coherent networks without requiring a predefined model. Biomedical image processing methods for fMRI cover how segmentation and registration algorithms are applied to co-register functional data with high-resolution anatomical MRI volumes for precise localization.

Applications

Functional magnetic resonance imaging has applications in a range of fields, including:

• Cognitive neuroscience, mapping functions such as language, memory, and perception to brain regions
• Presurgical planning to localize eloquent cortex before tumor resection or epilepsy surgery
• Psychiatric and neurological research as a biomarker for Alzheimer's disease, depression, and schizophrenia
• Brain-computer interfaces using decoded fMRI signals to infer intended actions or imagined speech
• Pharmaceutical research assessing how drugs modulate brain circuit activity
• Human connectomics projects mapping the functional architecture of large-scale brain networks