Wrapped Around the Heart: Cylindrical Quantum Sensor Arrays Peer Deeper Into Cardiac Electricity
Every heartbeat broadcasts a whisper of magnetism, a signal measuring tens of picotesla at the chest wall — around a million times weaker than Earth’s magnetic field and detectable only by ultrasensitive quantum instruments. For decades, researchers eavesdropping on that whisper have worked from a single vantage point: a flat array of sensors hovering in front of the chest. A new study from the University of Tokyo argues that the geometry itself has been quietly limiting what medicine can see. Writing in Annals of Biomedical Engineering, Wenyu Shang, Motofumi Fushimi, Shinichi Chikaki, and Masaki Sekino report that a cylindrical sensor array wrapped around the torso localizes the heart’s electrical sources more accurately than a conventional planar array, with the advantage growing the deeper and more hidden the source. Combining computer simulations, phantom measurements, and experiments in living rats using cryogenics-free quantum sensors, the team delivered some of the first experimental evidence that the shape of a magnetocardiography sensor array — long treated as a fixed constraint of the hardware — is a design variable worth optimizing.
Magnetocardiography, or MCG, is the magnetic sibling of the electrocardiogram. Where ECG electrodes record the voltages cardiac cells impress on the skin, MCG magnetometers record the magnetic fields generated by the same ionic currents. Because magnetic fields traverse biological tissue largely indifferent to the wildly different conductivities of muscle, fat, lung, and blood, MCG can carry information complementary to the ECG about how activation waves sweep through the heart. The catch is magnitude. Superconducting quantum interference devices — SQUIDs — the workhorses of MCG for half a century, achieve sensitivities of roughly one femtotesla per root hertz, but only at the price of cryogenic cooling. The insulated Dewar vessels holding their cryogens must stand between sensors and body, pushing the detectors centimeters from the heart and forcing engineers to arrange them in a flat plane facing the chest. That planar geometry was never really chosen; it was inherited from the physics of keeping superconductors cold. And it came at a cost: sources far from that plane — above all the heart’s posterior wall — are the ones a flat array struggles to localize.
The mathematical obstacle is the inverse problem: reconstructing the three-dimensional distribution of current sources inside the heart from magnetic measurements made outside the body. The problem is famously ill-posed, admitting no unique solution and amplifying noise and modeling errors without mercy. Recent advances in quantum sensing have loosened the hardware’s grip on geometry. Optically pumped magnetometers, which read magnetic fields through the spin states of atoms in a vapor cell, and nitrogen-vacancy diamond magnetometers both operate without cryogenic cooling, permitting sensors to be placed closer to the body and at positions and angles a SQUID Dewar forbids. Earlier theoretical studies had hinted that posterior or multiplane measurements might improve reconstruction, but experimental validation remained scarce, because array geometry was hard to vary in a real system. The Tokyo team hypothesized that a cylindrical array enclosing the torso would sample the cardiac magnetic field more informatively than a planar one, particularly for deep or posterior activity, and set out to test the idea at three levels of realism: simulation, a tissue-mimicking phantom, and living animals.
The foundation was a numerical model built from magnetic resonance imaging of a ten-week-old male rat. Using 3D Slicer, the researchers segmented the heart, lungs, and torso into boundary surfaces and assigned conductivities of 0.239, 0.067, and 0.033 siemens per meter respectively. They then solved the MCG forward problem with the boundary element method, using the Helsinki BEM Framework to compute the external field via the Geselowitz quasistatic formulation, in which the measured field is a superposition of the primary current generated by cellular electrical activity and secondary return currents that accumulate on tissue boundaries of differing conductivity. Against this model they pitted two geometries with matched projected areas: a planar array of 49 points on a 30-by-60-millimeter rectangle, and a cylindrical array of 48 points on a cylinder 30 millimeters long with a 30-millimeter radius. Five families of test dipoles — vertical, tangential, horizontal, radial, and randomly oriented — each comprising 100 sources of 1000 nanoampere-meters distributed through the heart region, were split into front-side and back-side groups relative to the cardiac midline, and Gaussian noise of 3 and 10 picotesla was injected to probe two signal-to-noise regimes.
The simulations delivered a consistent verdict. Signal-to-noise ratios fell with distance for both arrays, and radial dipoles, which generate little or no external magnetic field, were consistently the hardest to detect. For sources on the front of the heart, the planar array held a modest SNR edge; for back-side sources, the cylinder won. Crucially, the cylindrical array produced smaller average localization errors than the planar array across every tested distance and orientation, with the gap widening for back-side and deeper sources and for vertically oriented dipoles in particular, while radial dipoles grew markedly more error-prone under high noise. The most counterintuitive finding concerned goodness of fit, the standard measure of how well a reconstructed source explains the data. The planar array often posted slightly higher GOF values than the cylinder — even for back-side sources the cylinder localized far better. The authors’ explanation is subtle: a planar array by itself cannot fully distinguish the magnetic field patterns of dipoles at different locations, so a fit can look deceptively good while pointing to the wrong place. A high goodness of fit, in short, is not the same as a correct answer.
To confirm the effect in physical reality, the team built a rat-sized wet phantom: a cylindrical container filled with 0.9 percent saline of conductivity 0.21 siemens per meter, housing a physical single-dipole source — a small coil with exposed contacts roughly four millimeters apart, driven by a 20-hertz, five-volt sinusoidal voltage producing a dipole moment of approximately 860 nanoampere-meters, near the simulated value. Measurements were made inside a four-layer permalloy magnetically shielded room using four Quspin Gen-3 dual-axis optically pumped magnetometers. Because only four sensors existed, the researchers engineered a custom non-magnetic scanning apparatus — motors mounted outside the shielded room, motion transmitted mechanically — that rotated the subject for cylindrical sampling and translated it for planar mapping, reconstructing dense arrays from sequential positions. After averaging 500 cycles at each point, the cylindrical configuration localized the phantom’s source to within about 1.9 millimeters, versus roughly 13.0 millimeters for the planar array — even though the planar fit was nominally slightly better, with goodness-of-fit values of 0.94 against 0.91. Strikingly, the cylinder’s error was smaller than in simulation, while the planar array performed comparably to its simulated results.
The decisive test came in living animals. The team recorded MCG from five healthy, ten-week-old male rats, anesthetized with 1.5 to 2 percent isoflurane, hearts beating at six to seven hertz, positioned prone at the phantom’s distances: planar at 28 millimeters, cylinder at 32-millimeter radius. Each point was recorded for about two minutes, the electrocardiogram’s R-peak serving as a timing trigger to average 500 heartbeats per position — a strategy that sacrificed simultaneity for signal-to-noise. Magnetometers were recalibrated before every point, and anesthesia was managed to keep physiology stable. When the researchers applied single-dipole fitting to the R-wave of the averaged signals, both arrays placed the source broadly in the lower ventricular free wall and apex of the rat heart, consistent with known R-wave activation. But the cylinder’s simulated advantage did not clearly materialize: some cylindrical estimates landed outside the heart model altogether, and the cylindrical array showed lower goodness-of-fit values in every measurement, sometimes below 0.7 while the planar array stayed above 0.9. The reason, the authors argue, lies in the source model rather than the sensor geometry. Normal ventricular activation is spatially distributed — a coordinated depolarization wave sweeping across the myocardium — and cannot be honestly compressed into a single point.
The team therefore turned to a distributed model better matched to the biology. They placed 500 fixed source locations one millimeter beneath the heart model’s surface, each modeled as a freely oriented current dipole, and estimated their moments with minimum norm estimation — an L2-norm inverse solution stabilized by zero-order Tikhonov regularization, with the regularization parameter scaled to the lead-field matrix’s singular values. Reconstructed maps from both arrays concentrated around the lower ventricular and apical regions, matching previously reported rat activation patterns. Yet the two geometries told different stories. The cylindrical array consistently assigned more reconstructed amplitude to the posterior heart, pushing the source-weighted center — the amplitude-weighted centroid of the map — deeper along the anterior-posterior axis than the planar array did. Data fits were high for both, mostly above 0.95, and source entropy, a Shannon-entropy gauge of spatial concentration, differed only slightly between them. Repeatability separated the arrays most sharply: across five repeated measurements per animal, the cylindrical array’s source maps correlated with one another at values generally above 0.8, exceeding the planar array’s consistency in four of five rats. The cylinder, in short, produced a more stable picture of activity the flat array could barely resolve.
The researchers also confronted a blind spot in their own hardware. An optically pumped magnetometer’s vapor cell is physically large compared with a rat’s heart, so treating each sensor as an ideal point measurement could distort the comparison. Modeling each channel instead as eight integration points spread across the sensor’s actual volume increased absolute errors for both arrays, but the ratio of cylindrical to planar median error shifted only from about 0.46 to 0.48. The geometry effect was no artifact of idealized point sensors.
The implications extend beyond rats. For decades, SQUID-based MCG systems have sampled the chest from the front and little else, and few studies have asked what posterior coverage might add. By binding simulation, phantom validation, and in vivo measurement into one matched framework, the Tokyo group has created an experimental platform for vetting array designs before committing them to hardware. The authors caution that their cylindrical configuration cannot simply be scaled to humans, whose torso geometry, feasible array radii, and sensor-source distances differ substantially, and they are candid about its simplifications: a heart model without chamber-level segmentation, a torso truncated above and below the cardiac region, a single-layer phantom, sequential rather than simultaneous sampling, biaxial rather than triaxial sensor data, and healthy animals in normal rhythm — the least favorable condition for focal source localization. Localized activation patterns such as premature ventricular contractions, or controlled disease models, would offer cleaner tests. Yet the central message stands: when it comes to reading the heart’s magnetic secrets, wrapping around the problem sees more than staring at it from one side.
Subject of Research: The impact of sensor array geometry — a cylindrical torso-enclosing configuration versus a conventional planar array — on current-source estimation accuracy in magnetocardiography, evaluated through boundary element method simulations, phantom measurements, and in vivo rat experiments using optically pumped magnetometers.
Subject of Research: Medicine
Article Title: Enhancing Current-Source Imaging in Magnetocardiography: The Impact of Sensor Array Configuration
Article References: Shang, W., Fushimi, M., Chikaki, S., & Sekino, M. (2026). Enhancing Current-Source Imaging in Magnetocardiography: The Impact of Sensor Array Configuration. Annals of Biomedical Engineering. https://doi.org/10.1007/s10439-026-04343-y
Image Credits: AI Generated
DOI: 10.1007/s10439-026-04343-y
Keywords: Magnetocardiography, Sensor array configuration, Boundary element method, Inverse problem, Magnetic source imaging, Optically pumped magnetometers, Cardiac electrophysiology, Current-source localization
Cite Scienmag News
APA MLA Chicago
Arden W. (August 29, 2026). Sensor Array Layout Holds Key to Sharper Heart Magnetic Imaging. Scienmag. https://scienmag.com/sensor-array-layout-holds-key-to-sharper-heart-magnetic-imaging/
Arden W. “Sensor Array Layout Holds Key to Sharper Heart Magnetic Imaging.” Scienmag, 29 August 2026, https://scienmag.com/sensor-array-layout-holds-key-to-sharper-heart-magnetic-imaging/. Accessed 29 August 2026.
Arden W. “Sensor Array Layout Holds Key to Sharper Heart Magnetic Imaging.” Scienmag. August 29, 2026. https://scienmag.com/sensor-array-layout-holds-key-to-sharper-heart-magnetic-imaging/
Copy citation Download RIS
Tags: advanced cardiac imaging technologiesbiomagnetic signal detectionbiomedical engineering in cardiac diagnosticscardiac magnetic field imagingcryogenics-free quantum sensorscryogenics-free quantum sensors for medical applicationscylindrical sensor array designcylindrical sensor array design for improved heart imagingheart electrical source localizationinnovative approaches in magnetocardiography hardwaremagnetic field measurement at picotesla levelsmagnetocardiographymagnetocardiography signal measurement techniquesnon-invasive cardiac electrical source localizationnon-invasive cardiac imagingoptimization of sensor array geometry in biomedical imagingquantum sensor arraysquantum sensor arrays for cardiac magnetic field detectionsensor array geometry optimizationsensor array layout impact on heart magnetic imaging resolutionsimulation and experimental validation of sensor configurationsultra-sensitive quantum magnetometerswearable quantum sensors for cardiac diagnostics
