A Review of Magnetic Field Emissions from the Human Body: Sources, Sensors, and Uses

It has long been common practice to capture the electric fields emanated by the human body as a means of detecting and/or monitoring diverse health conditions. However, these electric fields are strongly impacted by the complex permittivity of biological tissues which deteriorates their waveforms and limits their diagnostic capabilities. As an alternative, recent progress has been made in the measurement of bio-magnetic fields occur from the natural currents flowing through the body. The advantage in this case is, since tissues are non-magnetic, magnetic fields propagate in an uninterrupted manner towards the skin surface where they are eventually collected. This unveils game-changing opportunities for future medical diagnostics. Nevertheless, a major challenge associated with sensing these naturally emanated magnetic fields is that they are extremely weak, and in fact orders of magnitude smaller than those generated by the Earth. To this end, extensive efforts have been pursued to realize sensing technology that is sensitive enough to collect bio-magnetic fields. Example fields of use include magnetomyography (MMG), magnetocardiography (MCG), magnetoencephalography (MEG), and Magnetoneurography (MNG) (including magnetospinography (MSG)). This review will provide an overview of technologies used to sense bio-magnetic fields, list their merits and limits in a critical manner, and discuss clinical applications.


I. INTRODUCTION
The human body consists of more than 200 types of cells and four different types of tissues where the action potential impulses propagate through.These electrically mediated signals travel along the axon as the ions flux in and out of the membrane forming various electrophysiological activities in the human body that gives rise to electric and magnetic fields [1].These currents are responsible for the electrophysiological activity in the human body that gives rise to electric fields.Examples of sensing these electric fields include electrocardiography (ECG) for the heart and electroencephalography (EEG) for the brain [2].However, However, electric signals can be greatly impacted by the complex permittivity and conductivity of biological tissues, leading to inaccurate diagnostic results [3].In addition, devices that capture bio-electric fields rely on electrodes that are in direct contact with the skin or scalp, making the procedure intrusive [4].Most importantly, the electric signal can only provide limited information.Complimentary features of the magnetic signals, such as field localization and vortex current identification, can help accelerate diagnostic procedures and provide more accurate diagnosis.
An alternative way is to measure the magnetic fields produced by the aforementioned ionic currents.Examples include magnetic fields radiated by the muscles (magnetomyography, MMG), heart (magnetocardiography, MCG), brain (magnetoencephalography, MEG), and nerves (magnetoneurography, MNG, such as from the spine, magnetospinography, MSG).Since biological tissues have magnetic permeability similar to vacuum, bio-magnetic fields can propagate to the surface of the skin without any distortion [5].In addition, bio-magnetic fields can be captured in a noncontact manner, in turn eliminating obtrusiveness, skin irritation and possible allergic reactions [6].However, the main challenge associated with capturing these magnetic waves is their extremely weak field strength and wide bandwidth.Referring to Table I and the example waveforms in Fig. 1 (original figures available in [7]- [11]), their frequency ranges from couple Hz to thousands of Hz and their magnitude is in the range of 10 -10 T to ~10 -15 T at the recording site [9] [10] 1 fT -1 pT 1-300 Hz 300 Hz [12]- [20].That is orders of magnitude lower than the earth's magnetic field (10 -4 T) [21] [22].Given the difficulty in capturing such weak signals, original focus had been on solely capturing bio-electric fields [23].However, this changed when extremely high sensitivity magnetometers were introduced, namely superconducting quantum interference devices (SQUIDs) [24], alongside magnetic shielding techniques (e.g., high permeability materials, such as Mu metal and Metglas) [25].However, SQUIDs operated in shielded rooms are bulky and extremely expensive [26]- [28].They also require cooling structures or liquid helium to maintain subzero temperatures (i.e., below 80K [29]), which further increases complexity and cost [26].These can only be partially mitigated via novel approaches that employ different cryogen (such as nitrogen) and cryocoolers [26] [30].More recently, other types of magnetometers or gradiometers have been developed for biomagnetic field sensing, aiming to overcome limitations of SQUIDs [31]- [33] [8].These sensors can be mainly categorized as: Atomic Magnetometers (AM), including Optically Pumped Magnetometers (OPM) and Spin Exchange Relaxation Free Atomic Magnetometers (SERF AM), and induction coil gradiometers.Here, to achieve the desired sensitivity, most of the sensors require some sort of shielding.This can be either partial shielding upon a targeted region [34] or complete shielding in the form of a shielded room where measurements are taken in [35]- [37].Even more recently, promising results have been reported for recording biomagnetic fields in non-shielded environments [8] [38]- [40].These approaches take advantage of advanced Digital Signal Processing (DSP), including, but not limited to: bandpass filtering, window averaging, moving averaging filtering [40] [8], and integration of a feedback mechanism to lower magnetic noise in the final processed data [41]- [43].
In this paper, we review the state-of-the-art technology in bio-magnetic field sensing, including both sensors and shielding techniques.Merits and limits of these approaches are presented in a critical manner.We also discuss the sources and clinical value of MMG, MCG, MEG and MNG (MSG), as well as provide example technologies used to capture these signals.To our knowledge, this is the first paper that reviews bio-magnetic field sensing and related technologies.Our ultimate goal is to familiarize readers with the state-of-the-art and inspire new technology development and clinical uses in this area.

A. SUPERCONDUCTING QUANTUM INTERFERENCE DEVICES (SQUIDS)
SQUIDs are the most commonly used devices for sensing biomagnetic fields [44].They operate based on the principles of Josephson Junctions, electron quantum tunneling, and the idea that magnetic flux through a superconducting loop is quantized [45].A typical SQUID consists of two parallel Josephson junctions forming a circular loop from two superconductors separated by two thin insulating layers, per Fig. 2 (a).A constant biasing current is applied and maintained in the circular loop making each side of the loop having half of the total current.As the bio-magnetic flux goes through the loop, the current caused by the magnetic flux is added on the original half of current on one side of the loop, whereas on the other side, the flux current is removed from the original half.This leads to phase difference between the junctions, making one side of the junction reach the critical current (i.e., maximum current that can pass through the junction) before the other one.Based on the quantization of magnetic flux, the voltage is oscillating between the maximum and minimal value of the quantum where the maximum values are happening at integer values of the flux quantum and the minimal values are happening at half integer values of the flux quantum.Counting the oscillations/changes in voltage, magnetic flux changes can be evaluated.
For biomagnetism, two different types of SQUID sensors have been typically used, viz.High-Tc and Low-Tc, as based upon their operating temperature.Earlier studies and commercially available devices use low-Tc SQUIDs that operate on high-cost liquid helium cooling systems at ~4K [46].More recent works introduced high-Tc SQUIDs [29] [26] that operate at ~77K [26] using a liquid nitrogen cooling system to reduce operational cost.A trade-off here is that high-Tc SQUIDs tend to have worse noise performance than the low-Tc SQUID counterparts, roughly a noise figure of one  This work is licensed under a Creative Commons Attribution 4.0 License.For more information, see https://creativecommons.org/licenses/by/4.0/order of magnitude lower [47].Nevertheless, noise performance may ultimately not impact the lowest biomagnetic signal that can be captured [47]: because of their higher operation temperature, high-Tc SQUIDs have less thermal insulation, leading to smaller distance between the signal source/subject and the cooled sensor, thus enhancing the signal strength that is fed into the SQUID.As a result, the final signal to noise ratio (SNR) can be comparable between high-Tc and low-Tc SQUIDs [48].Table II summarizes SQUID designs and their performance, noting that extensive work has been pursued towards improving upon the original concept and enabling higher sensitivity, lower cost, more compact design, and larger bandwidth [19] [35] [36] [38] [39] [44] [49]- [58].As seen, SQUIDs are extremely sensitive and able to detect magnetic fields as low as pT to fT.As expected, low-Tc tend to have lower noise floor: their noise performance typically lies in the range of 1-10 fT/√Hz at ~10 Hz.For some low-noise low-Tc SQUIDs, noise can go down to <1 fT/√Hz at ~100 Hz.On the other hand, for high-Tc SQUIDs, noise lies in the range of 10-100 fT/√Hz at ~10 Hz.
Key limitations associated with SQUID sensors entail: a) High cost and bulkiness.Besides the cost of the device itself, operation becomes further expensive due to the use of cryogenic cooling [30].Cost is also exacerbated by the 20-ton shielding room needed to accompany the device, made of multiple layers of expensive high-permeability alloys [44], [57].A detailed description of shielding technologies is provided in Section II.D. Fig. 2(b) shows a typical SQUID system in a superconducting magnetic shield (original figure available in [59]).As seen, the size of the sensor itself is already very bulky.On top, the shielding room introduces substantial additional volume.
b) Sophisticated fabrication, especially for high-Tc SQUIDs.Per Table II, commercially available SQUIDs all rely on the principle of low-Tc.One important factor that limits the commercialization of high-Tc SQUIDs is the poor fabrication yield and low reliability [60] [48].Typical low-Tc SQUIDs use Nb-based nano-SQUID technology, where the fabrication process is already mature for large scale manufacturing [61].For high-Tc SQUIDs, the most promising fabrication method utilizes grain boundary junctions based on YBa2Cu3O7-x (YBCO) thin films [62] deposited on SrTiO3 (STO) or LaAlO3 (LAO) [63] [64].More recent works report the use of ion beam milling to fabricate these nanojunctions [65] [62].Both fabrication processes are very challenging and only a small number of labs have the ability to produce them with high enough quality [66] [26].As a result, though high-Tc SQUID sensors seem to be a better alternative to low-Tc in terms of cost, progress needs to be made to empower their wide adoption.

B. ATOMIC MAGNETOMETERS (AM)
Two types of AMs, specifically Optically Pumped Magnetometer (OPM) and Spin Exchange Relaxation-Free (SERF), rival the SQUID sensors in terms of performance under certain conditions.Both AMs operate on a similar principle, i.e., heating atoms to a high temperature and detecting the magnetic attenuated atomic spins via pumping and probing of the optical system.Most AMs require alkali atoms such as K, Rb, and Cs to be vaporized by heating them up to a certain temperature, typically 100-190 ⁰C [67].These vaporized alkali atoms interact with the magnetic field such that, by detecting the atomic spins, the corresponding magnetic field can be identified [68] [69].More specifically, vaporized alkali atoms are first circularly polarized so that they spin alone in the same direction.The target magnetic fields interact with the polarized atoms, causing precession of their magnetic moment, which can be used to infer the magnetic field [68].SERFs, as the name suggests, differ from OPMs as they are free from decoherence due to spin exchange collision.This is typically achieved by elevating the temperature of the cell to increase the density of alkali metal atoms and eventually the rate of spin exchange collision so that it exceeds the Larmor Precession Frequency [69].We remark here that, though this principle describes traditional AMs, recent works have improved upon this technology in many ways (i.e., in terms of the atoms used) [37].Comparing the two, SERF AMs surpass OPM by exhibiting much higher sensitivity.[37] [70]- [82].As expected, SERF AMs have better sensitivity when compared to OPMs.Notably, some of the most recent SERFs can achieve noise performance of around 10 fT/√Hz at ~10 Hz which is comparable to most of the high-Tc SQUID sensors.In fact, theoretical studies have shown that SERF AMs could replace SQUIDs in the future [68] [83] [84], with some even claiming that SQUIDs have already been surpassed [68].Comparing AMs to SQUIDs, we note that: (1) AMs do not require any type of cooling structure; (2) AMs significantly reduce the operational cost as the device itself is less expensive, shielding is less complex, and no coolant is required [44] [85]; (3) AMs are very small in size.For example, Fig. 3 (original figures available in [86] [87]) shows the typical size of an OPM and SURF.Most of AMs are just couple centimeters in their largest dimension [75] [37] [67] [34].
Nevertheless, AMs still suffer from a number of limitations.One of the fundamental drawbacks is that they can only operate near zero ambient magnetic field [69].For example, to implement the SERF state, the external stray magnetic field needs to be <10 nT or ideally zero [69].To date, all types of AMs require magnetic shielding to maintain near zero ambient noise for ultra-high sensitivity [68] [78], with some options being more affordable than others [78] [34].As another drawback and due to theoretical limitations, AMs have extremely small bandwidth [84], restricting the bio-magnetic signals that can be captured.Another concern is related to safety as AMs are placed on top of/near the skin, where the sensor temperature can reach up to 65 ⁰C [82].Finally, the pump and probe laser that are used to polarize and detect the atomic spin make the procedure non-passive and may lead to safety concerns as well.

C. INDUCTION COIL GARDIOMETERS
Induction coil-based gradiometers rely on the simple principle of Faraday's law that changing magnetic field can induce voltage upon a coil's ends.This voltage can then be translated back to the magnetic flux being induced.Work reported in this area is very limited mainly due to high intrinsic noise.Nevertheless, recent works have shown the potential of capturing MCG activity (the strongest bio-magnetic field generated by the human body) using relatively lightweight and small coil sensors [40] [88] [8].Table IV compares the relevant parameters reported in [40] [88] [89] [8].In terms of the sensor's physical design, ferromagnetic cores [40] and optimal coil size/ratios [40] [88] [8] are used to boost sensitivity.For example, in [88] [89], the optimal coil size was chosen such that inductance (L) is maximum at a fixed winding length (similar to designing Brooks-Coils).However, this approach is not as effective when designing for a low noise bio-magnetic gradiometer.Based on the model developed in [88] [89], sensitivity (S) is defined as the induced voltage (Vout) vs. magnetic field strength (H).Note that S is always proportional to the coil winding (n) and coil mean radius (Ra), regardless of the operation frequency.However, fixed winding length doesn't mean fixed Ra and n.Also, L depends on the coil geometry, meaning that different Ra and n values lead to different L. In turn, different L also leads to different Ra and n values.Therefore, all variables are not independent of each other and the exact relationship between L and S is not clear without considering other parts of the coil parameters.The work in [40] and [8] uses a similar approach to determine the optimal coil geometry ratios.Ferromagnetic core is used in [40] to boost the coil sensitivity at the expense of increasing the sensor's weight.Comparing the two, (1) [40] is estimated to be at least 216 times heavier than [8], (2) [40] is significantly larger in size; but (3) [40] has better noise performance.
Table V compares the output voltages (Vout) and thermal noise (Vt) from coils with same outer diameters using the aforementioned three designs.Ideally, we want the gradiometer to be as small as possible with high output voltage and low thermal noise.Here, optimal coil geometry ratios from [40] [88] [8] are used to calculate the exact coil size.Vout and Vt are calculated using a tightly winded air-core coil model.Indeed, for a given 1 pT 40 Hz magnetic field, Brooks coil designs (i.e., [88] [89] ) have shown to have the smallest output voltage and smallest Vout:Vt ratio.Further, [40] and [8] have very similar performance with [8] while also exhibiting a better Vout:Vt ratio.
Compared to SQUIDs and AMs, induction coil gradiometers have the following advantages.(1) Operation cost is significantly reduced as no shielding or coolant are needed and the sensor itself is very inexpensive.(2) Size is comparable or smaller to that of a typical compact AM sensor.As an example, Fig. 4 shows one of the developed induction coil sensors (original figure available in [8]).(3) Portability is enhanced as there are no shielding requirements.(4) There is no theoretical constraint in terms of bandwidth.(5) Operation is fully passive.(6) There is no requirement for cooling or heating structures.
The only limitation of induction coil gradiometers relates to their noise performance.In unshielded environments, noise performance of an induction coil gradiometer is about 10 times worse than that of a typical OPM or high-Tc SQUID in a shielded environment.As a result, to date, even with adequate filtering and signal processing, induction coil gradiometers are only capable of capturing MCG signals when averaging the data for couple of minutes.However, not much work has been done using this technology yet, and the area is still developing and expanding.Future directions in this area may involve integrating machine learning, advanced signal processing, and partial shielding to optimize the system.

D. SHIELDING TECHNOLOGY
Magnetic shielding rooms (MSRs) are often used in addition to sensing systems to reduce the background noise that is present when recording bio-magnetic fields.Two types of MSRs are commonly used that rely on active and passive shielding, respectively.As the name suggests, passive shielding utilizes only passive components, typically multiple layers of high permeability materials.Active shielding adds components, such as multiple demagnetization coils, to the traditional passive shielding to further calibrate the remaining noise [90] [74] [44].
For direct-current (DC), MSRs rely on ferromagnetic materials with high permeability to create a preferential path for the magnetic field lines, shunting the magnetic flux and drawing the field into themselves.For alternating current (AC), MSRs utilize eddy currents induced on a conductive plate to cancel out the external magnetic noise [91].For most bio-magnetic applications, MSRs that utilize ferromagnetic shunt (i.e., DC operation principle) alone, can provide sufficient shielding [92].This is because, though the environmental noise contains both DC and AC components, the AC noise of interest (overlapping with the bio-magnetic signal) is slowly varying.With frequencies lower than a few hundred kHz, the eddy current induced on the conductive plate cannot effectively cancel out the external noise due to loss.Therefore, for human emitted bio-magnetic fields that focus on low-frequency noise, ferromagnetic material MSRs are typically used.
Nevertheless, active shielding can be used in conjunction with passive MSRs.In this case, coils are typically added to the outside of the passive MSRs to compensate the noise [93] [94].Active shielding systems work by sensing the noise using multiple gradiometers on the outer layer of the passive MSR.The recorded noise is then compensated by multiple coils within the compensation system.Currently, traditional MSR together with active shielding is the primary configuration used in the field of bio-magnetism.To numerically measure shielding ability, the concept of shielding factor (SF) is introduced.This is calculated as the ratio of the magnetic field induced when no shielding is present (B0) over the magnetic field induced at the same location when shielding is present (B) [95].Higher SF indicates higher level of shielding.To design rooms with high SF for bio-magnetic applications, two factors are critical: shape/dimensions and material.For a typical cylindrical or spherical shaped shield, several studies have discussed the theoretical optimal design, and for each case, SF has been calculated [96] [97].For materials, a general rule is to use materials with high permeability [92].One of the materials that is widely known for good shielding performance is called "Mu" material and is an alloy of nickel and iron.Different compositions can produce variations of Mu materials, all of which have high permeability and can be used for biomagnetic shielding.Example compositions include: 80% nickel, 4.2% molybdenum, 0.5% manganese, 0.35% silicon,  This work is licensed under a Creative Commons Attribution 4.0 License.For more information, see https://creativecommons.org/licenses/by/4.0/0.02% carbon and 15.03% iron (i.e.HyMu 80) [92], 80% nickel, 4.6% molybdenum and 15% iron [98], 80.5% nickel, 4.9% molybdenum and 13.7% iron [98], 77.9% nickel, 4.5% molybdenum, 3.5% copper and 13.2% iron [98], and 79.7% nickel, 15% iron, and 4.6% molybdenum [99].Nevertheless, more recent studies show that several layers of high magnetic permeability metals can attenuate the external fields, limiting the shielding ability [100].New materials such as MnZn ferrites are studied as a promising alternative [100].

III. APPLICATIONS
In this Section, we report clinical applications that rely on biomagnetic field sensing, and specifically MCG, MEG, MNG/MSG, and MMG.These applications are, of course, not limiting and the field is ever expanding.

A. MAGNETOCARDIOGRAPY (MCG)
Fig. 1(b) (original figures available in [8]) shows the MCG signal superimposed with an ECG signal.Key features of the wave consist of the P, Q, R, S, T, U spikes, where the QRS complex is the most prominent feature (main spike) lead by the P spikes and followed by the T and U spikes.MCG empowers advanced diagnostics of various cardiac-related health conditions in clinical settings (e.g., arrhythmia [101] [102], cardiac ischemia [101] [103], right atrial hypertrophy [104], coronary artery disease (CAD) [105]- [107], and Brugada syndrome [104]).Referring to CAD, one of the most common heart conditions, limitations of traditional ECG that can be overcome by MCG entail: (a) Difficulties in diagnosing symptomatic patients without persistent ECG features (i.e., ST-segment elevation) [107].By contrast, MCG is highly sensitive towards tangential and vortex currents that cannot be detected in ECG [105], enabling better detection of CAD with the DC injury current (note that injury current is slowly decaying (near-DC) current) [108].
(b) Low sensitivity/accuracy in detecting ECG property changes, particularly for rest-ECG, implying that several patients need to undergo expensive and invasive diagnostic procedures [102].By contrast, MCG provides a well-defined change, especially on the ST segment and T wave.Given these well-recognizable markers on MCG, research shows that MCG can achieve twice as high success rate for Sensitivity (Sens) and Negative Predictive Value (NPV) (i.e., 95.1% vs. 33.9%and 84.8% vs. 27.4%)while having slightly higher Specificity (Spec) and Positive Predictive Value (PPV) as compared to its ECG counterpart (i.e., 92.8% vs.91.1% and 97.8% vs.93.3%) [109].
(c) Lack of field localization which, in turn, may prohibit localization of coronary stenosis.By contrast, MCG provides much more localized fields and detailed 3D imaging over the heart so that the exact location of coronary stenosis can be detected.Research also shows that other MCG features, such as magnetic pole characteristics, are also associated with location identification [110].
In other cases, MCG has been explored for prognostic and monitoring tests [102] [111]- [116]: (a) Arrhythmogenic risk assessment: Arrhythmia is the leading cause of sudden cardiac death (SCD); therefore, prevention is of crucial importance [117].MCG features that can be extracted for arrhythmogenic risk evaluation include the late field, intra-QRS fragmentation, and QT dispersion (QTd) [113] [118]- [120].The late field refers to small abnormal deflections at the end of the QRS wave.Studies have found late field parameters extracted from MCG to discriminate post-myocardial infarction (MI) patients who had arrhythmia (ventricular tachycardia (VT)) from those who hadn't [113] and to distinguish patients with early repolarization patterns (ERP, known as a risk factor for ventricular fibrillation (VF)) [121].Compared to late fields extracted from signal averaged ECG (SAECG), MCG late fields provide more accurate prognosis of VT in post-MI patients [113].Intra-QRS fragmentation refers to the high frequency components in the QRS region [122].It can be quantified through the number of peaks (M) within the QRS and the intro-QRS fragmentation score (FRA) which is calculated as the product of peaks and the sum of the peak amplitudes [123].Studies have found both M and FRA extracted from MCG to predict arrhythmic events, VT, and even all-cause mortality in post-MI patients [118] [123] [124].Compared to SAECG intra-QRS parameters, MCG intra-QRS has higher specificity rate (100% vs. 91%) while having the same sensitivity (95%) [123].Finally, QTd measures the difference between the maximum and minimum QT duration [119].Studies have shown evidence that increased QTd can identify post-MI patients at risk for malignant arrhythmia, predict long-term prognosis in acute myocardial infarction (AMI) and even serve as a predictor of SCD [120] [125] [119].
(b) Rejection monitor: MCG shows promising results in monitoring post heart transplantation patients for rejection reaction.Traditionally, transplant rejection is monitored via serial cardiac endomyocardial biopsy (EMB) where a small myocardial tissue is obtained with an invasive procedure [126].Studies have shown MCG mapping to non-invasively test/monitor for rejection reaction using the intensity of the equivalent current dipole (ECD) [126] [127].In fact, it can detect acute graft rejection reaction as early as EMB with possibly higher detection rate.However, up to now, studies have very limited sample size (i.e., up to 15 patients) and further explorations are needed [128].
(c) Fetal MCG (fMCG): fMCG monitors the fetus's cardiac activities outside the maternal abdomen, which can serve as a diagnostic and prognostic tool for the baby.It can be used for several applications, such as: detection and classification of fetal arrhythmia, detection of congenital heart diseases, first-degree atrioventricular block, fetal long QT syndrome, fetal Wolff-Parkinson-White syndrome, and more [30] [129] [130].This is made possible by the fMCG's ability to clearly show all parts of the P, QRS, and T waves.As biological tissues have magnetic permeability similar to that of free space, fMCG can propagate relatively undisturbed through the body.By contrast, other commonly used fetal surveillance tests only provide mechanical assessment of heart rhythm which is not sufficient for diagnosis (i.e., echocardiography), are non-passive and may cause safety concerns (i.e., fetal MRI), and are unreliable (i.e., fetal ECG which, due to tissue attenuation, can only record cardiac signals with adequate information in about 50% of the cases).fMCG surpasses these approaches with the ability to provide good quality P, QRS and T waves information in normal pregnancies from the 20 th onward with a success rate close to 100% [131].In some studies, it is shown that fMCG can be recorded as early as the 13 th week of gestation [132].

B. MAGNETOENCEPHALOGRAPHY (MEG)
MEG has shown to achieve similar accuracy in localizing epilepsy as compared EEG [133]- [135].Fig. 1(a) (original figures available in [7]) shows an example auditory MEG where the spikes indicate activation of the brain.Notably, for patients with temporal lobe epilepsy, MCG offers higher sensitivity as much smaller brain activation areas are required for epileptic spikes detection as compared to scalp EEG (6-8cm 2 vs. 20-30cm 2 ) [136] [137].In addition, MEG is particularly useful for patients with frontal lobe epilepsy and neocortical epilepsy, as studies show that MEG offers higher spatial resolution and better signal-to-noise ratio, as well as yields significantly higher spike rate than EEG, hence facilitating advanced source localization [138] [139] In other cases, fetal MEG (fMEG) has been reported as a non-invasive means of monitoring brain electrophysiology.It can assess the maturation of different parts of the fetal brain (i.e., auditory evoked field, visual evoked field) which can serve as prenatal assessment, including diagnosis of developmental delays [140]- [142].This information can't otherwise be extracted with fetal EEG (fEEG) and invasive procedures are needed instead that place electrodes close to the head of the fetus [142].In turn, fEEG studies are primarily focused on primates (sheep) and postnatal newborns [143] [144].
MEG is also known for presurgical and preoperative evaluation, especially for epilepsy, focal cortical dysplasia (FCD), and brain tumor surgery [145]- [147].Studies show that MEG can provide more localized epileptogenic zoom and can detect locations that might otherwise be missed if only evaluated through EEG or MRI [148].Information provided with MEG, though not required for presurgical evaluation, has been demonstrated in several studies to provide favorable outcomes (i.e., seizure free, localized resection volumes) [149]- [151].For space occupying lesions, such as tumors, pre-operative functional mapping using MEG can help reduce intraoperative time, positively impact the extent of resection, and improve patient outcomes [152]- [154].
In other cases, studies show the potential of using different activation patterns within MEG to serve as biomarkers of differentiating between different emotions/feelings [155] [156] (e.g., neural activity from the primary somatosensory cortex in MEG at ~10 Hz is shown to link to pain stimuli [157], MEG theta band (4-8 Hz) signal patterns in the amygdala are shown to vary when exposed to angry, fearful and neutral faces [158]).Additionally, MEG-based technologies have been reported that quantify cognitive workload with the auditory steady-state response (ASSR) [159]; control brain computer interfaces (BCIs) with higher accuracy than EEG when it comes to multiple tasks with activities corresponding to distinct brain area [160]; and move a tetraplegic patient's index finger [161].

C. MAGNETONEUROGRAPHY (MNG) AND MAGNETOSPINOGRAPHY (MSG)
MNG and MSG are both developing technologies.Though not approved for clinical use yet, several studies have shown their potential for peripheral nerve and spinal cord applications in the future.Specifically, one potential application for MSG/MNG relates to visualizing spinal cord injury.Studies have found that evoked electrophysiological activities can be detected though MSG and MNG in various locations along the spinal cord.This indicates the potential of using MSG and MNG as a noninvasive tool for visualizing neural activity in the cauda equina [162], examining lumbar diseases [162], localizing lesion site in the lumbar canal [163], diagnosing conduction block even at the site of spinal stenosis in cervical myelopathy patients [164]- [166], detecting spinal root and dorsal horn dysfunction [166], visualizing ulnar nerve stimulation at spinal tracts at C5/6/7 [167], and measuring neural activity in the dorsal column and dorsal horn in the cervical cord [168] [10]) are examples of evoked MNG and MSG where each spike in the waveform is followed by the applied stimuli.The time difference between the applied stimuli and the appearance of the spike is called the peak latency.Here, the peak latency characterizes how fast the signal travels along each specific nerve.It can be used to check if there is any damage to that nerve pathway.Though only a very limited spinal nerve paths have been studied to date, MNG/MSG still show huge potential for numerous spinal related clinical applications.Applications of MSG/MNG beyond the spinal cord have also been demonstrated.Specifically, the technology has been used to diagnose and localize conduction block in brachial plexus neuropathy [169], detect lesions proximal to Erb's point in peripheral nerves [170], visualize neural activity in the brachial plexus [171], diagnose and localize focal neuropathies of cervical nerve roots [172], and diagnose functional electrophysiological somatosensory pathways [173].
Comparing MSG/MNG to traditional clinical practices, one key advantage is that MSG/MNG can detect electrophysical activities in the nerve non-invasively.By contrast, to detect electrical potential in the nerve, conventional approaches need to surgically identify the nerve and place the recording electrode which introduces numerous complications.

D. MAGNETOMYOGRAPHY (MMG)
MMG can be used to detect magnetic fields from muscles and has been demonstrated to achieve superior performance than gold-standard electromyography (EMG) on several occasions.Example applications include remote detection of mechanical/metabolic injury-related slowly decaying leakage/injury currents [174]; detection of female pelvic floor function associated with the contraction of levator ani muscle [19]; prediction of labor with uterine activity [55] [175] [176]; analysis of the muscular activity of the arm and study of the innervation of the hand [177] [34].Fig. 1(e) (original figures available in [11]) is an example of recorded levator ani muscles MMG side by side with the EMG during pregnancy.Each spike in the EMG has a corresponding spike in the MMG indicating the muscle contraction.
Among the above, one of the most studied applications of MMG is the detection/prediction of labor using uterine activity.MMG can measure the electrophysical activities of the uterine, extract information from action potentials in groups (bursts) related to the characteristic of uterine muscle contraction (i.e., frequency, duration, and number of simultaneously active cells), and, hence, a predict labor [55].
Analyzing uterine contractions and predicting labor could be useful clinically, especially for early identification and prevention of patients for premature delivery [55] [176].Currently, the most common approaches for labor prediction are intrauterine pressure catheter (IUPC), tocography (TOCO), and EMG.Comparing those approaches to MMG, IUPC and EMG are invasive.IUPC requires a catheter inserted into the uterus and rapturing of the amniotic membranes to measure the pressure changes [178], while EMG uses both internal electrodes and abdominal surface electrodes in direct contact with the patient [178] [179].TOCO is non-invasive but suffers from low resolution as it is sensitive to maternal motion artifacts [176].EMG exhibits high temporal resolution, but, due to the conductivity differences of tissue layers, results in attenuation of the recorded signal.Also, since EMG is recording the potential, results depend on the reference point which can produce only 2D views of the electrophysical activity.By contrast, MMG surpasses all approaches as (1) its operation is independent of tissue conductivity attenuation, (2) detection is non-invasive and non-contact and (3) may produce 3D mapping of localized sources.

IV. CONCLUSION
In this paper, we presented an overview of state-of-the-art and research-in-progress technologies used to detect human emitted bio-magnetic fields as well as their possible applications.In summary, detection of bio-magnetic fields is extremely challenging due to the low field strength and wide frequency bandwidth.Some of the most promising technologies entail SQUIDs, AMs, induction coil gradiometers, and shielding.Concurrently, future efforts should focus on (1) lowering the sensors' detection level; (2) improving noise performance; (3) increasing the shielding factor; (4) lowering the sensor cost, size, weight; (5) improving manufacturability; (6) enhancing safety; and (7) expanding the range of applications.Overall, detection of human emitted bio-magnetic signals opens up new opportunities for non-contact monitoring, diagnostics and prognostics in clinical healthcare settings and beyond.

FIGURE 2 .
FIGURE 2. (a) Operation principle of a typical SQUID.(b) A SQUID system in superconducting magnetic shield [59].

TABLE 2 . Relevant parameters for SQUIDs.
This article has been accepted for publication in IEEE Open Journal of Antennas and Propagation.This is the author's version which has not been fully edited and content may change prior to final publication.Citation information: DOI 10.1109/OJAP.2022.3186643This work is licensed under a Creative Commons Attribution 4.0 License.For more information, see https://creativecommons.org/licenses/by/4.0/

Table III summarizes
AMs that have shown to successfully capture human bio-magnetic signals

TABLE 3 . Relevant parameters for AMs
This article has been accepted for publication in IEEE Open Journal of Antennas and Propagation.This is the author's version which has not been fully edited and content may change prior to final publication.Citation information: DOI 10.1109/OJAP.2022.3186643 This work is licensed under a Creative Commons Attribution 4.0 License.For more information, see https://creativecommons.org/licenses/by/4.0/

TABLE 5 . Induction coil gradiometer performance comparison
This article has been accepted for publication in IEEE Open Journal of Antennas and Propagation.This is the author's version which has not been fully edited and content may change prior to final publication.Citation information: DOI 10.1109/OJAP.2022.3186643 This work is licensed under a Creative Commons Attribution 4.0 License.For more information, see https://creativecommons.org/licenses/by/4.0/

TABLE 6 . MNG and MSG excitation and recording sites
This article has been accepted for publication in IEEE Open Journal of Antennas and Propagation.This is the author's version which has not been fully edited and content may change prior to final publication.Citation information: DOI 10.1109/OJAP.2022.3186643 This work is licensed under a Creative Commons Attribution 4.0 License.For more information, see https://creativecommons.org/licenses/by/4.0/ [9]ince nerves serve as a pathway for transmitting signals, MSG and MNG are essentially recording the evoked potential traveling through the nerve.The stimulation typically involves external electrical stimuli applied to one end of the nerve and recording on the other end.Example stimulation and recording sites that have proven to be able to record evoked MSG/MNG together with their corresponding signal strength and peak latency are summarized in TableVI.Fig.1(c) and Fig.1(d)(original figures available in[9]