MRI spectroscopy, often abbreviated as MRS, turns a standard magnetic resonance scanner into a chemistry lab. Instead of producing the familiar grayscale anatomical images, mri spectroscopy measures the concentration of specific metabolites inside a chosen volume of tissue, such as N-acetylaspartate, choline, creatine, lactate, and lipids. The result is a graph of peaks plotted against chemical shift in parts per million, and each peak tells a clinician something different about cellular health, membrane turnover, energy metabolism, or hypoxia in the region of interest.

For radiologists, neurologists, and oncologists, MRS adds a biochemical layer to morphology. Two lesions can look almost identical on a T2-weighted image, yet a single voxel placed over each one can reveal dramatically different metabolite signatures. A high choline-to-NAA ratio raises concern for a high-grade glioma, while an elevated lactate doublet at 1.33 ppm suggests anaerobic glycolysis from necrosis, infarction, or mitochondrial disease. That extra information often changes the differential diagnosis before any biopsy is performed.

The technique exploits the same nuclear magnetic resonance physics that drives every MRI exam, but with a critical twist. Conventional imaging suppresses chemical shift to keep fat and water from misregistering, whereas spectroscopy preserves and exaggerates it so that protons attached to different molecules can be separated along a frequency axis. If you want a refresher on the underlying scanner technology, our guide to What Is an MRI Test? How Magnetic Resonance Imaging Scans Diagnose Disease in 2026 walks through the basics before you tackle MRS.

Most clinical MRS today is proton-based (1H-MRS) because hydrogen is abundant and the standard MRI body coil and head coil are already tuned for it. Phosphorus, carbon-13, sodium, and fluorine spectroscopy exist on research platforms, but they require multinuclear hardware and broadband transmitters that most community hospitals do not own. For this article we focus on 1H-MRS at 1.5 T and 3 T, which covers more than 95 percent of clinical practice in the United States.

Reading a spectrum is part pattern recognition, part quantitative analysis. Software packages like LCModel, jMRUI, Tarquin, and vendor tools such as Siemens syngo.via and GE READYView fit a basis set of known metabolite line shapes to the acquired data and report concentrations or ratios. Even with automation, the technologist must still place the voxel correctly, shim the magnetic field tightly, suppress the water signal, and choose an echo time that highlights the metabolites of interest.

This guide explains how MRI spectroscopy is acquired, what each peak means, the most common clinical applications, the pitfalls that can wreck a spectrum, and the steps registry-bound technologists should master before exam day. Whether you are studying for the ARRT MRI registry, preparing patients for a brain tumor workup, or simply curious about how a scanner can detect tumor cellularity without cutting tissue, the sections below will give you a working framework for spectroscopy in 2026.

Once the spectrum is on the screen, the next job is to read it. The horizontal axis is chemical shift, expressed in parts per million, with water at 4.7 ppm by convention and the scale running right to left so that higher ppm values sit on the left. The vertical axis is signal intensity, which is roughly proportional to metabolite concentration once corrections for relaxation and water content are applied. Most clinical reports anchor the eye on three landmark peaks before moving on to the more subtle ones.

N-acetylaspartate, or NAA, is the tallest peak in healthy adult brain tissue and resonates at 2.02 ppm. It is considered a neuronal marker because it is synthesized in mitochondria of mature neurons. Loss of NAA signals neuronal damage or loss, which is why it falls in stroke, multiple sclerosis plaques, high-grade tumors, and neurodegenerative disease. In infants the NAA peak is small at birth and rises through the first two years as myelination matures, an important normal variant.

Choline-containing compounds peak at 3.22 ppm and reflect membrane synthesis and breakdown. Choline rises in any process with rapid cell turnover — tumors, inflammation, demyelination, and gliosis. Creatine and phosphocreatine produce a peak at 3.03 ppm and serve as the internal reference because total creatine is relatively stable across the brain. When you read Cho/Cr or NAA/Cr ratios, creatine is the denominator on which the whole interpretation rests.

Lactate is the troublemaker that everyone learns to spot. It produces a doublet at 1.33 ppm, two closely spaced peaks caused by J-coupling between the methyl protons of lactic acid. At an intermediate TE of 144 ms the doublet inverts below the baseline, a behavior unique to J-coupled species that is exploited to distinguish lactate from overlapping lipid signals. Lactate is absent in normal adult brain, so its appearance is always pathological and points to anaerobic metabolism.

Lipids resonate between 0.9 and 1.3 ppm and appear in necrotic tumors, abscesses, demyelinating lesions, and any voxel contaminated by scalp fat. Myo-inositol at 3.56 ppm is a glial marker that rises in Alzheimer disease and in low-grade gliomas. Glutamate and glutamine, collectively called Glx, sit between 2.1 and 2.5 ppm and are difficult to separate at 1.5 T but become resolvable at 3 T with short echo times. For deeper context on what MRI abbreviations like NAA, TE, and PRESS actually mean, see our quick reference on MRI Medical Abbreviation: What MRI Stands For and Why It Matters.

Clinical interpretation almost always uses ratios rather than absolute concentrations because absolute quantification requires water referencing, coil loading correction, and relaxation modeling that many sites do not perform. A Cho/NAA ratio above 2 in adult brain is highly suspicious for malignancy, while a Cho/Cr above 2 plus loss of NAA in a ring-enhancing lesion favors high-grade glioma over abscess. These numbers are not absolute rules, but they form the working vocabulary that radiologists use every day at the workstation.

Finally, remember that every spectrum is a sample of a specific volume of tissue. A voxel placed two millimeters off the lesion can give a completely normal-looking spectrum from adjacent healthy brain, and a voxel that catches a sliver of CSF will be diluted and unreadable. Reading the peaks is therefore inseparable from reading the voxel placement on the corresponding anatomical images.

Even the most beautifully designed MRS protocol can be ruined by a handful of common pitfalls. Recognizing them early — ideally at the console, before the patient leaves the table — separates a useful study from an unreadable one. The most frequent problem is voxel contamination. A voxel that overlaps the skull, scalp, or paranasal sinuses will pick up enormous lipid signals between 0.9 and 1.3 ppm that swamp the metabolite peaks and create a tilted baseline that no post-processing can fully repair.

Poor shimming is the second great enemy of spectroscopy. The shim coils have to flatten the static magnetic field across the chosen voxel so that all the spins resonate at nearly the same frequency. If the linewidth of the water reference peak exceeds about 10 Hz at 1.5 T, individual metabolite peaks begin to overlap and quantification falls apart. Areas near the skull base, paranasal sinuses, and dental hardware are particularly hard to shim because of susceptibility differences between air, bone, and tissue.

Inadequate water suppression leaves a giant residual peak at 4.7 ppm that distorts the baseline of the entire spectrum. Most modern scanners apply three or more frequency-selective saturation pulses, but if the center frequency drifts because of patient motion or scanner heating, water suppression efficiency drops and the surrounding metabolite peaks ride on a sloping baseline. A quick frequency adjustment between scans usually fixes the problem.

Motion is especially destructive in spectroscopy because each acquisition is averaged with the others. A single cough or swallow during a 128-average single-voxel scan can introduce phase errors that broaden every peak. Restless patients should be coached carefully, padded comfortably, and if necessary scanned with a shorter NEX and accepting somewhat noisier spectra rather than catastrophic motion artifact.

Chemical shift displacement is a subtle but important effect. Because metabolites resonate at slightly different frequencies, the slice-selective pulses used to define the voxel actually excite slightly different volumes for each metabolite. The displacement is more severe at higher field strength and with narrower-bandwidth RF pulses, which is why 3 T MRS protocols use high-bandwidth pulses and why technologists should be aware that the voxel they draw on the screen is not exactly the volume sampled for lactate or lipid.

Hemorrhage and metallic artifacts can ruin a spectrum because the local susceptibility gradients destroy the shim and produce massive frequency offsets. Voxels placed in or near blood products, surgical clips, or hemosiderin staining will return broad, distorted spectra. The same applies to lesions immediately adjacent to the skull base or sphenoid sinus. When patients ask why the scanner is so loud during these long sequences, point them to our explainer on the Noise of MRI Machine: Why MRI Scanners Are So Loud and What to Expect so they know what to expect before the exam begins.

Finally, beware of pediatric normal variants. Infant brain has low NAA, high choline, and visible myo-inositol simply because myelination is incomplete. Reading a six-month-old's spectrum with adult reference ranges will produce a false-positive impression of an aggressive process. Age-appropriate references and consultation with a pediatric neuroradiologist are essential whenever spectroscopy is performed on children.

For technologists studying for the ARRT MRI registry or the ARMRIT exam, MRI spectroscopy shows up in roughly five to ten percent of questions, scattered across the physics, sequence, and pathology sections. The questions tend to focus on a handful of high-yield facts rather than deep biochemistry, so a focused study session can cover most of what the registry will ask. Start by memorizing the chemical shift positions of NAA, choline, creatine, lactate, lipid, and myo-inositol, along with the clinical implication of each.

Understand the difference between single-voxel spectroscopy and chemical shift imaging. Single-voxel exams are faster, easier to shim, and best when the lesion is well defined and you only need one spectrum. Chemical shift imaging, also called magnetic resonance spectroscopic imaging or MRSI, samples a grid of voxels across a slice or volume and is preferred when you need to compare lesion metabolism to surrounding tissue or when the lesion is heterogeneous.

Know the two main localization techniques. PRESS uses one 90° pulse followed by two 180° refocusing pulses and gives a spin echo with strong SNR but a minimum TE around 30 ms. STEAM uses three 90° pulses and gives a stimulated echo at half the signal, but allows much shorter TE values useful for detecting J-coupled species like glutamate. Many registry questions ask which sequence allows shorter TE or which has higher signal — STEAM gets the short TE, PRESS gets the higher signal.

For TE selection, remember the rule of thumb: short TE shows everything, intermediate TE inverts lactate, long TE gives a clean baseline with NAA, choline, and creatine. The 144 ms lactate inversion is a favorite registry question and a clinically vital teaching point. If you can explain why lactate inverts and lipid does not, you will likely get a related anatomy or pathology question correct as well.

Practice reading sample spectra from textbook plates and online atlases. Identify each peak by position before looking at the answer, then check your reasoning. Many candidates struggle because they study MRS only as a list of metabolites without ever looking at an actual graph. After ten or fifteen worked examples, peak recognition becomes nearly automatic. For broader background on how the field developed, see our piece on the The History of MRI: From Discovery to Modern Medicine, which traces the path from NMR spectroscopy in the 1940s to today's clinical scanners.

Finally, know the contraindications and safety considerations that overlap with conventional MRI. MRS itself adds no contrast, no radiation, and no additional patient risk beyond the time in the magnet, but the standard MRI screening for pacemakers, cochlear implants, ferromagnetic foreign bodies, and pregnancy still applies. Registry questions sometimes embed an MRS scenario inside a safety question to test whether you remember the basics.

Build a one-page cheat sheet with the chemical shifts, the TE behaviors, the PRESS-versus-STEAM differences, and the three clinical scenarios (tumor, stroke, metabolic). Review it daily for two weeks before the exam. Combined with timed practice questions, this approach is enough to convert MRS from a feared topic into one of the easier point earners on the registry.

Putting MRI spectroscopy to work in the clinic comes down to three practical habits: protocol discipline, communication with the radiologist, and ongoing self-review of the spectra you acquire. Protocol discipline means saving and labeling your MRS sequences clearly on the scanner — short-TE brain, intermediate-TE brain, long-TE brain, prostate single-voxel, breast single-voxel — so that the right defaults load every time. A few minutes spent organizing the protocol tree saves hours of troubleshooting later.

Communication matters because the radiologist often does not see the patient or the scanner console. If the voxel had to be shifted because of a hemorrhage, if shimming was poor because of dental work, or if the patient moved during the acquisition, the reader needs to know. A two-sentence note in the technologist's section of the report prevents misinterpretation and protects both the patient and the practice. Most modern PACS systems allow free-text annotations on the spectrum image itself, which is the most direct way to flag concerns.

Self-review is the single fastest way to improve. Save a copy of every interesting spectrum you acquire, and once a week sit down with the radiologist's report and compare your impression of the peaks to the final interpretation. Within a few months you will develop a personal library of brain tumors, strokes, metabolic disorders, and normal variants that surpasses anything a textbook can provide. This kind of deliberate practice is what separates competent MRS technologists from exceptional ones.

Patient preparation is straightforward but easy to neglect. Patients should be screened the same way as for any MRI, with the standard ferromagnetic checklist and IV access if contrast will be given. Spectroscopy itself does not require contrast, but it is often added to a tumor protocol that already includes gadolinium. The MRS sequence is usually run before contrast administration so that the gadolinium does not slightly alter relaxation-based quantification, although in practice the effect on ratios is small.

For coil selection, use the highest-channel head coil available for brain MRS. A 32-channel or 64-channel head coil dramatically improves SNR over an older 8-channel coil and can cut acquisition time in half for the same noise level. For prostate MRS, an endorectal coil was once standard but most centers now use the body coil at 3 T with acceptable results. For breast MRS, dedicated breast coils with good fat suppression are essential.

Keep an eye on emerging applications. Hyperpolarized 13C MRS is moving from research into early clinical trials for prostate and cardiac imaging, allowing real-time tracking of pyruvate-to-lactate conversion as a marker of cancer metabolism. Deep learning reconstruction is shortening MRS scan times and improving peak fitting accuracy. 7 T systems, now FDA-cleared for clinical use, dramatically improve spectral resolution for difficult metabolites like GABA and glutamate. The field is more dynamic than it has been in twenty years.

Finally, remember why this technique exists. MRI spectroscopy lets you peer inside a living human being and measure the chemistry of disease without cutting, biopsying, or exposing the patient to ionizing radiation. It is one of the most elegant applications of nuclear magnetic resonance physics ever brought to the bedside, and a technologist who masters it brings real value to every neuro, oncology, and pediatric service. Treat each spectrum as a small experiment, learn from every one, and the technique will reward you for the rest of your career.