What Is MRS Imaging and Its Medical Applications?

Magnetic resonance spectroscopy, usually called MRS, is a scanning technique that uses the same machine as a standard MRI but reads the body’s chemistry instead of its anatomy. Where a conventional MRI produces detailed pictures of organs and tissues, MRS produces a chemical fingerprint of a specific region, revealing which molecules are present and in what concentrations. Clinicians use it to investigate brain tumors, neurological diseases, liver fat, heart metabolism, and a growing list of other conditions, often without needing a biopsy.

How MRS Works and How It Differs From a Standard MRI

An MRI scanner detects signals from hydrogen atoms (protons) in the body after exciting them with radiofrequency pulses inside a strong magnetic field. A standard MRI uses those signals to build an image of tissue structure. MRS exploits a physical phenomenon: the tiny electronic environment surrounding each atom shifts its resonant frequency by a small but measurable amount. Because hydrogen atoms sitting inside a fat molecule resonate at a slightly different frequency than hydrogen atoms inside a sugar or an amino acid, the scanner can distinguish one chemical compound from another.¹ The output is not a picture but a spectrum, a graph of peaks at different frequencies. Each peak corresponds to a particular molecule, and the height of the peak reflects how much of that molecule is present.²

From the patient’s perspective, the experience is identical to a regular MRI: you lie in the same tube, hear the same banging noises, and hold still for a similar length of time. The difference is entirely in the software instructions the scanner receives. In many hospitals, an MRS sequence is simply tacked onto the end of a standard MRI exam, adding a few extra minutes to the scan.

The Metabolites MRS Measures and Why They Matter

The real power of MRS lies in the specific chemicals it can detect. In the brain, the workhorses are a handful of metabolites that serve as biological markers for what is happening inside tissue at a cellular level.

N-acetylaspartate (NAA): found almost exclusively in healthy neurons. When NAA drops, it signals that neurons are damaged or dying. This makes it one of the most clinically useful signals in brain MRS.
Choline: involved in cell membrane turnover. Elevated choline often points to rapid cell division, which is why it rises in tumors and certain inflammatory conditions.
Creatine: related to energy metabolism. Because creatine levels tend to be relatively stable across healthy brain tissue, it is frequently used as an internal reference to express other metabolites as ratios.
Myo-inositol: a marker of glial cells, the brain’s support cells. It tends to rise in certain neurodegenerative conditions and infections.
Lactate: normally almost undetectable in the brain. When it shows up, it indicates that tissue is not getting enough oxygen or is metabolizing abnormally, as happens in stroke, some tumors, and certain inherited metabolic diseases.
GABA and glutamate: the brain’s principal inhibitory and excitatory signaling molecules. Measuring them requires specialized editing sequences because their signals overlap with other, stronger peaks at standard field strengths.³

Outside the brain, different metabolites take center stage. In the prostate, MRS tracks citrate and choline. In the liver, it measures fat content. In the heart and skeletal muscle, phosphorus-based MRS reads energy molecules like phosphocreatine and ATP. Each organ has its own metabolic signature, and disease disrupts that signature in characteristic ways.

Brain Tumors and the Question of Recurrence

One of the most established clinical uses of MRS is in evaluating brain tumors. When a mass shows up on a standard MRI, the image alone sometimes cannot tell you with confidence whether it is a high-grade glioma, a metastasis, an infection, or even an aggressive form of demyelination. MRS helps narrow the diagnosis by reading the chemical profile of the lesion.⁴ A classic tumor pattern shows high choline (rapid membrane turnover from dividing cells), low NAA (neurons being destroyed), and sometimes elevated lactate (oxygen-starved tissue).

The scenario where MRS arguably proves most valuable is after treatment. When a patient with a glioma has had surgery and radiation, follow-up MRI scans often show a new area of enhancement that could be the tumor growing back or could be radiation damage to normal tissue, called radiation necrosis. This distinction is critical because the treatments for each are completely different. MRS can help sort them out: recurrent tumor tends to show high choline-to-NAA ratios and high choline-to-lipid ratios, while radiation necrosis typically does not. One study found the mean choline-to-NAA ratio was roughly 2.7 in recurrent tumors versus about 1.5 in radiation necrosis, a statistically significant gap.⁵ Other research has defined specific cutoff ratios that balance sensitivity and specificity, offering clinicians objective thresholds rather than relying on subjective visual pattern reading of the spectrum.⁶

A more recent development is the ability of MRS to detect a molecule called 2-hydroxyglutarate (2HG), which accumulates in gliomas carrying mutations in the IDH1 or IDH2 genes. These mutations define a biologically distinct and generally more favorable subtype of glioma, so identifying them matters for prognosis and treatment planning. Researchers have shown that optimized MRS sequences can detect 2HG noninvasively, and that the detection correlates with confirmed gene mutations in resected tissue.⁷ This is a striking capability: learning the molecular genetics of a tumor without ever cutting into it.⁸

Neurodegenerative and Neuroinflammatory Diseases

In Alzheimer’s disease, MRS consistently shows two changes in the brain’s parietal lobe gray matter: NAA goes down and myo-inositol goes up. The drop in NAA reflects neuron loss; the rise in myo-inositol may reflect the proliferation of glial cells responding to the disease process. Using both markers together improves the ability to distinguish Alzheimer’s patients from healthy older adults compared with using either marker alone.⁹

In multiple sclerosis, MRS has reshaped how researchers think about the disease. MS was long understood primarily as a disease of myelin, the insulating sheath around nerve fibers. MRS studies measuring NAA revealed that substantial axonal damage, not just demyelination, occurs throughout the brain in MS patients. The damage is present inside visible lesions and also in white matter that looks perfectly normal on conventional MRI.¹⁰ A longitudinal study found that in relapsing-remitting MS, most of the decline in the NAA-to-creatine ratio over time came from changes in normal-appearing white matter rather than from within lesions, and this decline correlated strongly with worsening disability.¹¹ This finding has practical implications: conventional MRI lesion counts alone may underestimate how much damage is accumulating.

Huntington’s disease is another area of active research. A longitudinal study using MRS in the putamen, a brain structure heavily affected in Huntington’s, found that total creatine and NAA were significantly reduced in people with manifest disease compared to premanifest gene carriers at two-year follow-up, and that certain metabolites tracked with shrinkage of the caudate nucleus over time.¹² The hope is that such metabolic markers could eventually serve as objective endpoints in clinical trials of drugs aimed at slowing the disease.

Epilepsy

For patients with epilepsy who are being evaluated for possible surgery, MRS provides another layer of information. The technique can lateralize seizure onset by showing a drop in the NAA-to-choline-plus-creatine ratio in the affected hippocampus. In one study of patients with complex partial seizures, the abnormal hippocampus showed a reduction in this ratio of roughly a quarter compared to healthy controls, and the technique detected abnormalities in every patient examined.¹³ When imaging and EEG findings are ambiguous, MRS can contribute additional evidence to help pinpoint which side of the brain is responsible.

Inherited Metabolic Diseases in Children

Some of the most dramatic MRS findings come from pediatric neurometabolic disorders, rare inherited conditions where enzyme defects cause abnormal chemicals to build up in the brain. In Canavan disease, for example, MRS shows a strikingly elevated NAA peak, an unusual pattern since most brain diseases reduce NAA. The biochemical reason is that the enzyme responsible for breaking down NAA is deficient in Canavan disease, causing it to accumulate.¹⁴ In maple syrup urine disease, MRS can identify a characteristic peak from branched-chain amino acids, and in galactosemia, a galactitol peak appears at a specific frequency.¹⁵ In several of these diseases, MRS can contribute to a noninvasive diagnosis that might otherwise require invasive testing.

Applications Outside the Brain

Though neurological uses get the most attention, MRS has found a foothold in several other organs.

In the prostate, MRS measures the ratio of choline and creatine to citrate. Healthy prostate tissue produces and stores large amounts of citrate, but prostate cancer cells consume it. A rising choline-plus-creatine-to-citrate ratio can flag suspicious regions within the gland.¹⁶ This is sometimes used to guide biopsy toward the most metabolically active area rather than sampling blindly.

In the breast, MRS looks for elevated choline as a marker of malignancy. In locally advanced breast cancer, one study detected a choline signal in about 80% of malignant tumors before treatment, with a sensitivity of 78% and specificity of 86% for identifying cancer. After chemotherapy, the choline signal disappeared or shrank in roughly 89% of patients who responded, making it a potential early indicator of whether treatment is working before the tumor physically shrinks on imaging.¹⁷

In the liver, MRS has become a reference standard for measuring fat content noninvasively. Quantifying liver fat matters in the evaluation of fatty liver disease, and MRS measurements correlate closely with what liver biopsy shows. In one comparison study, the correlation between MRS-determined fat content and biopsy-determined steatosis was strong, and MRS could accurately distinguish between mild and more advanced levels of fat accumulation.¹⁸ This is especially useful in clinical trials testing drugs for fatty liver disease, where repeat biopsies are undesirable.

Heart and Muscle Energy Metabolism

The heart runs on a tightly regulated energy supply, and phosphorus MRS (which detects phosphorus-31 rather than hydrogen) can measure it. The key readout is the ratio of phosphocreatine to ATP, which reflects how much energy reserve the heart has. In heart failure, this ratio drops, and the drop has been linked to worse outcomes.¹⁹ A study of patients with chronic heart failure found that those who also had iron deficiency had a significantly lower phosphocreatine-to-ATP ratio than those without iron deficiency, suggesting that iron deficiency may worsen heart failure partly by impairing the heart’s energy-producing machinery.²⁰

The same phosphorus technique applies to skeletal muscle, where it can assess mitochondrial function by tracking how quickly phosphocreatine recovers after exercise. A slower recovery rate indicates that the muscle’s mitochondria are not performing well. This has been used in research on conditions ranging from mitochondrial diseases to Down syndrome, where adults with Down syndrome showed significantly slower phosphocreatine recovery compared to matched controls, pointing to impaired mitochondrial function in their muscles.²¹

Single-Voxel Versus Multi-Voxel Approaches

When a clinic orders an MRS study, one of the first decisions is whether to use single-voxel spectroscopy or multi-voxel spectroscopy (also called chemical shift imaging). Single-voxel spectroscopy measures one small volume of tissue at a time, typically a cube about 1.5 centimeters on a side. It is fast, technically straightforward, and produces clean spectra. Multi-voxel methods cover a larger region by dividing it into a grid of smaller volumes, producing a map of metabolite concentrations across an area. The tradeoff is longer scan time and more potential for artifacts.

In practice, the two approaches tend to produce comparable metabolite ratios in healthy tissue. A pediatric comparison study found that an accelerated three-dimensional spiral chemical shift imaging sequence could cover a much larger brain region in the same three-minute acquisition time as a single-voxel scan, with a fixed measurable offset in ratios that could be corrected for, making the multi-voxel technique a feasible substitute.²² For evaluating a focal lesion like a single brain tumor, single-voxel is often sufficient. For mapping metabolites across a larger tumor, tracking infiltration into surrounding tissue, or surveying both hippocampi in epilepsy, multi-voxel approaches provide spatial information that a single-voxel scan cannot.

Ultra-High-Field Scanners and What They Unlock

Most clinical MRS is performed at 1.5 or 3 Tesla field strengths, the standard for hospital MRI machines. Ultra-high-field scanners operating at 7 Tesla and above are increasingly available at research centers and are beginning to enter clinical use. For MRS, the benefit of a stronger magnet is twofold: the signal gets louder and the chemical peaks spread further apart, making it easier to tell overlapping molecules from each other.²³

This matters most for metabolites whose signals are buried under stronger neighbors at lower field strengths. Glutamate, glutamine, glutathione, and GABA can all be quantified with greater confidence at 7 Tesla, which is particularly relevant for psychiatric and neurological research where these neurotransmitters are central to the disease process.²⁴ Conditions like schizophrenia, major depression, and anxiety disorders all involve hypotheses about abnormal levels of these chemicals, and sharper spectral resolution helps test those hypotheses more reliably.

Hyperpolarized Carbon-13 MRS

Standard MRS is limited by sensitivity: the signals from metabolites are thousands of times weaker than the water signal used for conventional MRI, which is why scan times are long and voxel sizes are large. Hyperpolarized carbon-13 MRS sidesteps this problem using a technique called dynamic nuclear polarization, which temporarily boosts the signal from a carbon-13-labeled molecule by a factor of tens of thousands. The most common probe is pyruvate, a molecule that sits at a metabolic crossroads. Once injected, the hyperpolarized pyruvate is rapidly converted into lactate, alanine, or bicarbonate depending on the metabolic state of the tissue, and these conversions can be imaged in real time.²⁵

This is genuinely new territory. Traditional MRS gives a static snapshot of metabolite concentrations. Hyperpolarized MRS shows metabolic flux, the speed at which one molecule is being converted into another, which tells you about enzyme activity in living tissue in real time.²⁶ Because many cancers show dramatically altered pyruvate-to-lactate conversion, this technique has the potential to identify aggressive tumors and monitor treatment response much earlier than anatomical imaging can.²⁷ Human trials are underway in prostate, brain, and cardiac applications.

How Machine Learning Is Changing the Workflow

One longstanding barrier to wider clinical adoption of MRS has been the complexity of data processing. A raw MRS spectrum is messy: the peaks sit on a wavy baseline, noise can obscure smaller signals, and fitting each peak to determine its height requires specialized software and expertise. This means that in many hospitals, even when MRS data are collected, they may not be fully analyzed in routine practice.

Deep learning is starting to change that. Researchers have developed neural networks that can take a noisy, imperfect brain spectrum and output a cleaned, baseline-corrected metabolite spectrum, essentially automating the expert step of spectral fitting. One approach trained a convolutional neural network to map raw brain spectra onto noise-free, line-narrowed outputs, producing robust metabolite estimates even when the input quality was poor.²⁸ Another model designed for multi-voxel spectroscopic imaging in glioblastoma patients could fit a whole brain’s worth of spectra in about one minute on a standard computer, handling common artifacts automatically.²⁹ If these tools become validated and integrated into scanner software, MRS could become far more accessible to radiologists who do not have years of spectroscopy training.

Why MRS Is Not Used More Often

Given all these capabilities, you might wonder why MRS is not a standard part of every MRI exam. Several practical limitations keep it in the supporting-role category for now. The signals are inherently weak, so voxels need to be relatively large compared to the fine anatomical resolution of MRI, and scan times are longer. Patient motion during the acquisition degrades spectral quality, which is especially challenging in pediatric or critically ill populations. Tissue outside the target voxel can contaminate the spectrum. And the results require careful interpretation: metabolite ratios can shift for more than one reason, so MRS findings almost always need to be read alongside the clinical picture and conventional imaging rather than standing alone.

There is also a familiarity gap. Many radiologists trained without significant MRS exposure, and hospital MRI protocols are already packed. Adding an extra sequence means more time on the scanner, which translates to fewer patients scanned per day. The push toward faster, more automated MRS processing and the growing availability of vendor-supported spectroscopy packages on clinical scanners are slowly lowering these barriers, but MRS remains a tool used primarily at academic medical centers and large referral hospitals rather than in community imaging practice.