What Does MPI Mean? Heart Imaging and Computing

MPI most commonly stands for myocardial perfusion imaging in medicine and magnetic particle imaging in emerging biomedical research. In computing, especially computational cardiology, it stands for Message Passing Interface. All three abbreviations show up in the cardiac sciences, which can create genuine confusion when you’re reading a report, searching a medical database, or trying to understand a diagnosis. The meaning that matters most to the largest number of people is the first one, because myocardial perfusion imaging is one of the most widely performed cardiac tests in the world.

Myocardial Perfusion Imaging, the Most Common Meaning

When your cardiologist orders an “MPI study,” they are almost certainly referring to myocardial perfusion imaging. This is a nuclear medicine test that evaluates how well blood flows through your heart muscle. A small amount of a radioactive tracer is injected into a vein, and a specialized camera captures images showing where the tracer has been taken up by heart tissue. Areas that receive good blood flow light up; areas that don’t show up as defects. The test is typically done in two rounds, once while you’re at rest and once while your heart is under stress, either from exercise on a treadmill or from a drug that mimics the effect of exercise.

The logic is straightforward. If a region of your heart looks normal at rest but shows reduced tracer uptake during stress, that region has a blood supply problem that only manifests when the heart is working hard. That pattern strongly suggests a narrowed coronary artery. If a region looks diminished in both the rest and stress images, it may represent scar tissue from a previous heart attack. The greater the extent and severity of these stress-induced perfusion defects, the higher the risk of future cardiac events during follow-up.¹

SPECT and PET, the Two Cameras Behind MPI

Myocardial perfusion imaging can be performed with two different types of cameras: SPECT (single-photon emission computed tomography) and PET (positron emission tomography). SPECT is far more common because the equipment is less expensive and more widely available. PET produces higher-quality images but costs more and requires different tracers, some of which have very short half-lives and need an on-site cyclotron to produce.

Head-to-head comparisons show that PET has a modest edge in accuracy. A meta-analysis pooling data from multiple studies found that PET had a pooled sensitivity of about 93% for detecting coronary artery disease, compared with about 88% for SPECT. Specificity was similar between the two, around 75–81%, without a statistically significant difference.² In practical terms, both are good at catching disease, but PET misses fewer cases.

That advantage becomes especially important in patients with smaller hearts. Women, for instance, tend to have smaller left ventricles, and SPECT performance drops significantly in that group. In a substudy from a phase III trial, SPECT’s diagnostic accuracy fell from an area under the curve of 0.75 in larger hearts to 0.67 in smaller ones, while PET held steady at 0.77–0.79 regardless of heart size. The sensitivity gap was stark: in smaller hearts, PET correctly identified disease in about 67% of cases compared with 43% for SPECT.³ If you are a woman who has been told your MPI results are “equivocal,” it may be worth asking whether PET would give a clearer answer.

PET also enables more precise measurement of blood flow in absolute terms. Rather than just comparing one region of the heart to another, PET can quantify the actual volume of blood reaching the heart muscle per minute. Newer SPECT cameras are beginning to offer this capability as well, and early data suggest that newer solid-state SPECT systems can identify reduced blood flow with accuracy comparable to PET.⁴ But PET remains the reference standard when quantitative flow measurement is the goal.

Newer SPECT Technology and Lower Radiation Doses

One of the biggest practical concerns patients have about MPI is radiation exposure. The tracers used in nuclear imaging deliver a measurable dose, and earlier-generation cameras required relatively large injections to get usable images. That has changed considerably with the arrival of cadmium-zinc-telluride (CZT) detectors, a type of solid-state technology that is far more sensitive than the older sodium-iodide crystals used in conventional SPECT cameras. CZT detectors can be up to seven times more sensitive, with spatial resolution more than doubled compared with older systems.⁵

What this means for you as a patient is tangible. In a direct comparison, CZT cameras required about a fifth less injected radiotracer per patient than conventional cameras and cut imaging time from roughly 20 minutes down to about 6 minutes. Perhaps more meaningfully, the improved stress-image quality meant that fewer patients needed the additional rest-phase scan at all, which itself carries an extra dose of tracer. Only about 35% of patients scanned on the CZT camera needed rest imaging, compared with 56% on the conventional camera.⁶ Fewer injections, less radiation, and a shorter time lying still under a camera — that is a real improvement in patient experience.

Dynamic imaging with CZT-SPECT, where the camera captures images continuously as the tracer first enters the heart, is also opening the door to absolute blood-flow measurement on SPECT systems. Early protocols use relatively low tracer doses for this purpose.⁷ This is an active area of research that could make the quantitative advantages of PET more accessible at SPECT-level cost.

How MPI Compares With CT Angiography

If you’ve been told you need a heart test, you may wonder why your doctor chose MPI over a CT angiogram, or vice versa. They answer slightly different questions. CT angiography directly visualizes the coronary arteries and can show plaque buildup and narrowing. MPI shows the downstream consequence of that narrowing: is the heart muscle actually receiving enough blood? A coronary artery can look moderately narrowed on CT but still deliver adequate flow under stress, or it can look only mildly narrowed but be functionally significant because of the way the plaque is shaped.

Cost-effectiveness analyses have found that CT angiography strategies tend to be less expensive per correct diagnosis than MPI-first strategies for patients with chest pain and no known coronary disease.⁸ That doesn’t mean CT is always better. MPI excels when the question is not “are there narrowed arteries” but “is the narrowing actually causing problems.” In patients with known disease, after stent placement, or when anatomy is complex, MPI often provides information that CT cannot.

MPI as Magnetic Particle Imaging

Entirely separate from myocardial perfusion imaging, a newer technology called magnetic particle imaging shares the same abbreviation. This form of MPI is a tracer-based imaging modality that works by detecting the magnetic response of superparamagnetic iron oxide nanoparticles (often called SPIONs). These tiny particles are injected into the body and then detected by applying alternating magnetic fields. Because the particles have a distinctive nonlinear magnetic response, the signal comes exclusively from the tracer itself, with essentially no background noise from surrounding tissue.⁹

That zero-background property is what makes magnetic particle imaging exciting. In MRI, the signal from the tracer has to compete with signals from all the water in your body. In CT, contrast agents change the density of blood slightly against an already-complex tissue background. In magnetic particle imaging, if there are no nanoparticles in a region, there is no signal at all. The result is extremely high contrast.¹⁰

The technology also avoids ionizing radiation entirely. Unlike nuclear perfusion imaging (which uses radioactive tracers) or CT (which uses X-rays), magnetic particle imaging relies only on magnetic fields. And because the iron oxide nanoparticles are the tracer, there is no need for iodinated contrast agents, which can be problematic for patients with kidney disease.

Cardiovascular Applications of Magnetic Particle Imaging

Researchers are particularly interested in magnetic particle imaging for the heart because it combines high spatial resolution with very fast image acquisition, making it suitable for capturing something as dynamic as a beating heart. Early work has explored using it to visualize heart muscle perfusion and to quantify blood flow, roles currently filled by nuclear MPI and cardiac MRI.¹¹

Beyond perfusion, magnetic particle imaging has shown promise for tracking cells and therapeutic agents in the body. In one study, researchers used it to follow the movement of transplanted stem cells in real time. Intravenously administered stem cells were immediately trapped in the lungs and then migrated to the liver within a day, a trafficking pattern the researchers could monitor and quantify without surgery or repeated biopsies.¹² That kind of capability could eventually prove useful for monitoring stem-cell therapies aimed at repairing damaged heart muscle after a heart attack.

Magnetic particle imaging is still in the preclinical stage. No commercial scanners are approved for routine clinical use on humans yet, though several groups are working toward that goal. The main hurdles are scaling the scanner bore size to fit a human torso, optimizing the nanoparticle tracers for regulatory approval, and establishing the clinical workflows that would make the technology practical in a hospital setting. Reviews of the field note that standardization of tracers and scanning protocols is needed before the transition from research tool to clinical device can happen in earnest.⁹

MPI as Message Passing Interface in Cardiac Computing

The third meaning of MPI in the cardiac sciences has nothing to do with imaging patients directly. Message Passing Interface is a computing standard that allows multiple processors to work together on a single problem by sending data back and forth. It is the backbone of most high-performance computing and is widely used in cardiac simulation research, where the computational demands are enormous.

Simulating the electrical activity of a human heart, for example, requires solving equations across millions of tiny elements representing cardiac tissue. No single processor can handle that in a reasonable time frame. Researchers break the heart model into domains and distribute those domains across hundreds or thousands of processor cores, using MPI to coordinate them. One cardiac electrophysiology solver demonstrated a roughly 11-fold speedup over a single-processor solution by leveraging GPU clusters with MPI coordination, and it could run 512 parallel simulations across 128 computing nodes.¹³

This kind of computing power enables “digital twins” of individual patient hearts, virtual replicas that can be used to test treatment strategies before applying them in the real world. The CRIMSON framework, an open-source platform for cardiovascular simulation, uses an MPI-parallelized flow solver that can scale to tens of thousands of processor cores. Researchers use it to simulate blood flow patterns based on a specific patient’s imaging data.¹⁴

MPI-based parallel computing also shows up in simulating mechanical heart valves. Modeling a prosthetic valve inside a pulsing flow of blood is a fluid-structure interaction problem that requires extremely fine meshes and small time steps. Parallel solvers using domain decomposition and MPI allow researchers to run three-dimensional simulations that would be impractical on a single machine.¹⁵ Similar approaches have been applied to simulating the natural aortic valve, capturing both the realistic bending of the leaflets and the swirling flow patterns in the aorta.¹⁶

Where These Meanings Intersect

The overlap between imaging and computing is not just a naming coincidence. The data generated by myocardial perfusion imaging increasingly feeds into computational pipelines that rely on parallel processing. Deep learning algorithms trained on MPI scan data, for instance, are being developed to automate the reading of perfusion studies. A meta-analysis of deep learning applications in myocardial perfusion imaging found that convolutional neural networks achieved an area under the curve of about 0.89 for classifying scans, outperforming simpler architectures.¹⁷ Training those networks on large datasets requires exactly the kind of massively parallel hardware that MPI (the computing standard) was designed to coordinate.

Meanwhile, the patient-specific hemodynamic simulations built using MPI-parallelized solvers often begin with imaging data. A cardiac MRI or CT scan provides the anatomical geometry, and a nuclear perfusion study may supply the physiological boundary conditions. The simulation then predicts things like wall shear stress or pressure gradients that are difficult to measure directly. The computing meaning of MPI and the imaging meaning of MPI are, in this workflow, literally dependent on each other.

Emerging Tracers for Nuclear MPI

On the nuclear perfusion imaging side, the tracer technology itself is evolving. Most SPECT MPI today uses technetium-99m-labeled agents like sestamibi or tetrofosmin, while PET commonly uses rubidium-82 or nitrogen-13 ammonia. Rubidium-82 PET has shown high sensitivity for detecting coronary disease, correctly identifying it in over 90% of patients with confirmed blockages in one study.¹⁸

A newer PET tracer, flurpiridaz F-18, has been in development for years and has attracted attention because it combines favorable imaging properties with a longer half-life than rubidium-82. Preclinical and clinical data show it has a high extraction fraction (meaning it gets taken up by heart muscle very efficiently), produces sharp images, and maintains stable contrast over time.¹⁹ Unlike rubidium-82, which needs a costly generator and decays in about 75 seconds, flurpiridaz could theoretically be produced at a regional cyclotron and shipped to facilities that lack on-site production. If it achieves broad clinical adoption, it could make high-quality PET perfusion imaging available in community hospitals rather than only academic centers.

How Automated Reading Is Changing MPI Interpretation

Historically, perfusion scans were read by trained nuclear cardiologists who visually assessed the images for defects. That process is subjective; two readers looking at the same scan sometimes disagree. Automated quantification software now provides a computer-generated score of perfusion abnormality, and studies have compared its performance against expert visual reads. In one study using a high-efficiency SPECT camera, automated reads had similar overall diagnostic accuracy to visual interpretation when both were compared against invasive angiography. The automated approach tended to be more specific (better at correctly identifying normal scans) while the visual approach was more sensitive (better at catching disease).²⁰ In practice, most labs now use both: the software provides a quantitative baseline, and the physician overlays clinical judgment.

The deeper integration of machine learning into this process is where things get interesting. Rather than simply scoring perfusion maps against a database of normals, neural networks can learn patterns from thousands of prior scans and outcomes. The potential upside is more consistent reads, less dependence on individual reader expertise, and the ability to flag high-risk scans for urgent attention. The practical limitation remains that these algorithms need diverse, well-labeled training data, and most current models have been validated on relatively small datasets from single institutions.