Researchers at the University of North Carolina School of Medicine have introduced SORDINO, a functional magnetic resonance imaging acquisition sequence that dramatically quiets the notoriously loud scanning environment while producing clearer, less distorted brain images. As reported by Medical Xpress, the technique was developed in the lab of Yen-Yu Ian Shih, Ph.D., professor of neurology and associate director of the UNC Biomedical Research Imaging Center, and described in a paper published in Nature Neuroscience. For those of us who build pipelines around fMRI data, this is not a marginal refinement; it is a methodological shift that changes which behavioral paradigms we can confidently capture at all.
The acoustic problem and what it costs the signal
Functional MRI is built on the blood-oxygen-level-dependent signal, a downstream proxy for neural activity whose amplitude is vanishingly small and easily contaminated. Conventional sequences drive this contrast through rapid gradient switching, and that hardware choreography is acoustically brutal: peak sound levels inside the bore routinely reach 120 to 138 decibels, comparable to a jackhammer running at close range. Consider the implications for a mouse. Loud acoustic noise elevates stress hormones, induces movement, and introduces electromagnetic artifacts that bleed into concurrent electrophysiological recordings. The brain you are trying to measure is, in a very literal sense, reacting to the instrument you are using to observe it. That is the diagnostic dilemma at the heart of years of small-animal fMRI: the very behaviors researchers wish to study are distorted by the act of scanning.
SORDINO as an acquisition-layer solution
SORDINO, which stands for Steady-state On-the-Ramp Detection of INduction-decay with Oversampling, is not a postprocessing patch or a denoising algorithm layered onto existing data. It is an acquisition sequence, the set of instructions the scanner uses to collect and translate blood flow and oxygenation measurements into images. By reshaping how gradients are played out, the method substantially reduces acoustic noise, electromagnetic interference, and stress-related hormone release while producing images with less geometric distortion. The group was granted a U.S. patent for the technique in 2024, signaling that it is treated as a genuine engineering contribution rather than a one-off demonstration. In mouse models, the improvements have been substantial enough to enable studies of voluntary skilled movements and even social interactions between two animals scanned at the same time, paradigms that would be nearly impossible to interpret under conventional noise. Shih, the senior author, noted that SORDINO has become routine in his laboratory and expressed hope that it will help other groups overcome long-standing technical barriers and eventually translate to human MRI. The paper, by first author Martin J. MacKinnon and colleagues, frames the sequence as silent, sensitive, specific, and artifact-resistant in awake, behaving animals.
What this opens up, and what to track
When the acoustic environment stops fighting the experiment, the experimental menu expands, but the trajectory from a small-animal method to a clinical scanner is rarely a straight line. Performance metrics reported in rodent systems, including sensitivity gains and artifact profiles, will need independent replication, and the degree to which SORDINO's noise reduction generalizes to human gradient coils and field strengths remains an open engineering question. For developers building preprocessing pipelines and software around fMRI, the practical invitation is to track this work closely: acquisition-layer innovations tend to flow upstream into reconstruction algorithms, denoising toolchains, and the assumptions baked into statistical models. A quieter scanner does not merely make existing analyses easier; it repositions what counts as a feasible experimental question in the first place, and that is a quiet shift worth paying attention to.