Should patients with brain implants undergo MRI?

An MR unit is a machine that maintains three qualitatively different electromagnetic fields, and their precise control allows an MR image to be acquired. The basic functional principle is as follows: (1) A strong, homogeneous static magnetic field, called the B₀-field, is created within the bore of the MR scanner. Its purpose is the polarisation of half-integer nuclear spins along the z-axis. The net magnetisation exposed to an external magnetic field will precess at the Larmor frequency ω which is related to B₀ and the gyromagnetic ratio γ  =  42.25 MHz T⁻¹ such that $\omega = \gamma \cdot B_0$. (2) Powerful radio frequency (RF) pulses are driven at the Larmor frequency, thereby flipping the magnetisation out of its alignment direction. In turn, the flipped and precessing magnetisation now induces a signal in the receiver coils. The relaxation characteristics of the excited nuclei differ among tissue types and thus enable tissue contrast after signal acquisition. (3) Another 'layer' of weak magnetic fields arranged as a spatial gradient in three orthogonal directions is selectively superimposed onto the B₀-field. This alters the precession frequency of the nuclei as a function of position and hence allows the spatial separation of the measured signals into volume elements (voxels) through spatial encoding of the obtained signal. The gradient magnetic field is rapidly switched on and off during MR acquisition at a frequency of some kHz. After recording, the data is computationally assembled into an MR image.

The greatest strength of MRI lies in its potential to measure a large number of various physical parameters, such as relaxation times, proton density, flow and diffusion, temperature, abundance of molecules and oxygenation of blood, all with the same hardware. This flexibility is achieved by the software-controlled adjustment of timing and magnitude parameters for the RF and gradient fields. The imaging recipe, which is a chronological ordering of RF and gradient field switching, is called an imaging sequence, and is usually displayed in a timing diagram. All sequences are based on fundamental blocks such as the gradient-echo (GE) or spin-echo (SE) sequence; hence, fMRI or other advanced recording techniques utilise modified standard imaging blocks.

What is the effect of different sequences on the hazards that implant recipients are facing inside the MRI? Indeed, because the hardware does not change, the type of hazard is equal for all possible imaging sequences, i.e. it is the magnitude of the interactions which depends strongly on the sequence used, because the timing and the resulting magnitude of the generated magnetic fields varies with the applied sequences. Hazard levels with respect to RF interactions depend on the power and number of RF pulses per unit time, defining the RF power deposition (SAR level). SE sequences, for example, make use of many RF refocusing pulses. This leads to a larger number of RF pulses per unit time compared to the GE sequences [52], and as a consequence, to a higher RF power deposition. Similarly, the gradient slew rate and the number of gradient switching operations per unit time determine the strength of gradient-dependent interactions. fMRI is particularly vulnerable to influence from implants. The most widespread application is the detection of the blood-oxygen-level-dependent (BOLD) signal in the brain as an indicator for nerve cell activity [53]. It is the small difference in the magnetic susceptibility of venous blood to its vicinity that enables the BOLD contrast. Consequently, interference with the magnetic susceptibility of implants can be detrimental as the acquired echo-planar imaging (EPI) data is very sensitive to image distortion from susceptibility gradients due to the low bandwidth in the phase encoding direction [53].

An electric voltage is induced in an electrically conducting material whenever the magnetic flux ${\Phi\, = \, A \vec{B}\cdot\vec{n}}$ passing through the area A of the material changes over time (figure 5(a)). The induced voltage is given by Faraday's law of induction,

$$\begin{align} \newcommand{\e}{{\rm e}} \displaystyle \label{u_ind} U_{\rm ind} \, = \, - A\, \frac{\partial\, \Phi}{\partial\, t} \ \ . \nonumber \end{align} \tag{ 1 }$$

Zoom In Zoom Out Reset image size

This results in an induced electric current—the 'eddy current'

$$\begin{align} \newcommand{\e}{{\rm e}} \displaystyle \label{i_ind} I_{\rm ind} \, = \, - \frac{A}{R}\, \frac{\partial\, \Phi}{\partial\, t}, \nonumber \end{align} \tag{ 2 }$$

if the electrical resistance of the loop is finite, i.e. for a closed conductive loop, or an open loop with an attached circuit with finite resistance. When the law of induction is applied to the situation of implant recipients in an MR environment, it becomes clear that implant geometry, the movement of the patient relative to the scanner, and the switching gradient field are relevant. Three typical situations during MRI examination may lead to induced voltages: (1) When the patient is moved towards and into the MR scanner. In this case the magnetic flux density increases from near zero outside the scanner room to the maximum value of B₀. The magnitude of induced voltage depends on the speed at which the patient is moved. (2) Inside and in the direct surroundings of the MR scanner (i.e. in the spatially varying fringe field of the scanner), any movement of the patient that results in rotation of the implant with respect to the magnetic field may induce a relevant voltage. (3) During MR image acquisition, the rapid switching of the gradient fields induce an alternating voltage at the corresponding frequency, usually in the kHz range. To provide an example, the voltage induced in a $(10~{\rm cm}){}^2$ loop perpendicular to a magnetic field change of 1 T per second is 10 mV. If this loop is a short-circuited copper wire with a ${0.1~{\rm mm}^2}$ cross section, the eddy current will be ${0.7 \, {\rm A}}$. These numbers scale proportionally to the frequency and to the cross-sectional area, and are highly dependent on the geometry. The induction of voltages and eddy currents in a device may cause unintentional electrical stimulation of the (nervous) tissue. This has caused life-threatening situations for CP patients in the past [91–93]. Such situations are unlikely to happen with neural implants as they do not support vital functions. However, artefacts in the data recordings or damage to the implant may occur, which consequently can make device reimplantation necessary.

Magnetic forces result from the interaction of a magnetic moment with a magnetic field and may lead to movement relative to the surrounding tissue. The magnetic moment is essentially the strength of a permanent magnet. It is also caused by the magnetisation of any material when exposed to an external magnetic field such as the B₀ field of an MR environment, with a magnitude proportional to the magnetic susceptibility $\chi_v$ (see table 2) and its volume V: $\vec{\mu} \, = \, V \chi_v \vec{B}/\mu_0$, see figure 5(b).

Materials are distinguished into three different categories according to their value of magnetic susceptibility. The force exerted by diamagnetic ($\chi_v < 0$) and paramagnetic materials ($\chi_v > 0$) in MRI with $|\chi_v| \sim 10^{-5}$ is negligible. However, even small amounts of ferromagnetic material with $\chi_v \gg 1$ may exert a significant force.

A magnetic moment is also exerted by currents around a closed loop, such as the eddy currents described by equation (2). It is perpendicular and proportional to the surface area A of the loop, $\vec{\mu} \, = \, \vec{n}\, A \, I$ (see figure 5(c)), or, in the case of eddy currents, described by equation (2): $\vec{\mu} \, = \, - ({A^2}/{R})\, {\partial\, \Phi}/{\partial\, t}$. This may occur when the patient is moved in or out of the MR scanner, or moves during imaging. The interaction of a magnetic moment with a magnetic field gradient causes a linear force $\vec{F}$ (figure 5(d)), while a tilt of the magnetic moment relative to the magnetic field causes a torque $\vec{\tau}$ (rotational force, figure 5(e)):

$$\begin{align} \newcommand{\e}{{\rm e}} \displaystyle \vec{F}\ = \ \vec{\nabla}\left(\vec{\mu}\cdot \vec{B}\right) \ \ {\rm and} \ \ \ \vec{\tau} = \vec{\mu} \times \vec{B} \ \ . \nonumber \end{align} \tag{ 3 }$$

The important points are, on the one hand, that linear force only arises in regions with a magnetic gradient, i.e. at the entrance of the scanner and during the imaging procedure. For example a ${1\, {\rm mm}^3}$ piece of 'non-magnetic' 316 stainless steel (see table 2) is magnetised by a ${1.5 \, {\rm~T}}$ magnetic field to exert a magnetic moment of $11\cdot 10^{-6} \, {\rm A}~{\rm m}^2$. A ${1 \, {\rm~T~ m^{-1}}}$ gradient then creates a force of ${11\, \mu{\rm N}}$—14% of its weight of ${78\, \mu{\rm N}}$. A ${1\, {\rm mm}^3}$ neodymium magnet with a magnetic moment of $0.8 \cdot 10^{-3} ~{\rm A}~{\rm m}^2$ would, in the same situation, be subject to a force of ${0.8\, {\rm mN}}$—11 times its weight.

The rotational force, however, arises only when the magnetic moment of the material is not parallel to the magnetic field, and hence it does not occur in paramagnetic materials that are magnetised by the magnetic field, but only in permanent magnets, ferromagnetic materials with hysteresis, or in conductors with induced eddy currents. The ${1\, {\rm mm}^3}$ neodymium magnet, for example, perpendicular to a ${1.5 \, {\rm~T}}$ magnetic field is subject to a torque of $1.2 \cdot 10^{-3}~ {\rm Nm}$. This corresponds to its weight acting on a $17\, {\rm m}$ long lever.

The ${(10 \, {\rm cm}){}^2}$ conductor loop in a time-varying magnetic field of ${1\, {\rm~T ~s^{-1}}}$ described in section 2.3.1 exerts a magnetic moment of $0.7 \cdot 10^{-3} {\rm A}~{\rm m}^2$, similar to the ${1\, {\rm mm}^3}$ neodymium magnet. The exertion of force onto an implanted device may result in mechanical stress on the tissue and thus severe pain. The dislocation of an implanted magnet can compromise the device function and make surgical intervention necessary.

The specific absorption rate (SAR) measured in units of power per mass: (W kg⁻¹) quantifies the absorption of electromagnetic fields in a material and results in heating. The interaction between the RF field and water molecules causes the heating of tissue and, even without an implant, poses a risk to the patient during MRI procedures [98, 99]. Conductive materials, however, interact differently with the RF field and may be subject to more intense heating [100]. An RF field that is incident to a metal surface will be partially reflected and partially transmitted into the material (figure 5(f)). The transmitted fraction is absorbed and converted into heat. While this is a microscopic process that is independent of the shape of the object, it may be greatly enhanced whenever a system is in resonance with the RF field due to its macroscopic structure. This occurs on a small scale when the resonant frequency $1/(2\pi\sqrt{LC})$ of a circuit with inductance L and capacitance C matches the RF frequency. On a larger scale, a similar effect occurs when the length of a conductor is a multiple [101], the same, or an integer fraction (half or quarter) [102] of the RF wavelength. Note that the cables for active implants are often helical wound wires to allow the elongation caused by body movement. Thus, the effective wire length may be a multiple of the cable length. The RF wavelength scales inversely proportional to the B₀ field strength and is further reduced by the relative permittivity $\newcommand{\e}{{\rm e}} \epsilon_r$ of the surrounding medium (tissue), $\newcommand{\e}{{\rm e}} \lambda \simeq {7}/{((B_0/{\rm~T}) \sqrt{\epsilon_r})}\, {\rm m}$. With $\newcommand{\e}{{\rm e}} \epsilon_r\approx 100$ for the tissue and the RF at a frequency of 100 MHz [103], the RF wavelength is ${\simeq 50\, {\rm cm}}$ for a 1.5 T scanner or ${\simeq 25\, {\rm cm}}$ at 3 T, so the half or quarter wavelength matches the cable lengths of the DBS implants, for example.

Further heating of implant components may arise from eddy currents based on induction and joule heating [104]. From the induced voltage and current equations (1) and (2), the dissipated power is

$$\begin{align} \newcommand{\e}{{\rm e}} \displaystyle \label{w_ind} W_{\rm ind} \, = \, U\, I\, = \, \frac{A^2}{R} \left(\frac{\partial\,\Phi}{\partial\, t}\right)^2 \ . \nonumber \end{align} \tag{ 4 }$$

For the ${(10 \, {\rm cm}){}^2}$ conductor loop with ${0.1\, {\rm mm}^2}$ wires and a time-varying magnetic field of ${1\, {\rm~T ~s^{-1}}}$, as it is typical for a patient moving into the magnet, the dissipated power is just ${7\, {\rm mW}}$. Because of the quadratic scaling in equation (4), however, the power induced by gradient fields switching in the kHz-frequency is on the order of several Watts. The heating of implant components may result in tissue burns, which can cause tissue necrosis [105, 106] or permanent neurological dysfunctions when occurring in the brain [31].

Imaging artefacts are image disturbances that have no equivalent in the real object. There are several kinds of imaging artefacts which occur in commonly practised MRI and may cause dark patches of signal voids, or artificial appearances due to the superposition of multiple voxels. With respect to implants, susceptibility artefacts are the most relevant. Susceptibility artefacts are caused by the magnetisation of the materials in the B₀ field as described in section 2.3.2. The magnetisation creates its own locally varying magnetic field that superposes onto the homogeneous B₀ field. The resulting field distortions depend highly on the object geometry and are proportional to the local susceptibility mismatch (see table 2) through which a spatial shift in the Larmor frequency is caused, introducing an error into the imaging assumptions. Large field distortions simply shift the Larmor frequency outside the RF excitation band and due to the resonance frequency mismatch, no signal for any affected voxel is received. The inhomogeneity of the distortions also causes a variation of the Larmor frequency inside a voxel such that the resonance signal received from this voxel is suppressed by destructive interference (figure 5(g)), causing dark artefacts near the object. This effect is particularly relevant near sharp edges and small objects. Such frequency shifts further result in erroneous spatial encoding, i.e. geometric distortion of the image.

While these artefacts cannot be described by a simple formula, an example might provide an indication. Thus an object 1 cm in size with a susceptibility mismatch of a few ppm may cause artefacts on the scale of a millimetre in its immediate vicinity, depending on the MR sequence. An example of where such imaging artefacts naturally occur is at the interface between air and tissue, such as near the mouth, nose and sinuses. However, much larger susceptibility imaging artefacts occur in the vicinity of metals (or even worse—permanent magnets), as the difference in magnetic susceptibility compared to the tissue is enormous. The artefact of a ferromagnetic object, such as a permanent magnet, is so drastic that it essentially has a corrupting effect on the entire image [107]. Imaging artefacts do not harm patients directly, but become hazardous through misinterpretation [108], or when regions of diagnostic importance are obscured. Possibilities of tackling imaging artefacts include the utilisation of susceptibility matched materials [95], such as carbon nanotube yarn for electrodes with susceptibility close to water [109], or the development of artefact-resistant MRI sequences.

The complication of moving magnets in MRI was foreseen. One group proposed simply applying a tight head wrap to grant extra stability to the internal magnet and thereby prevent magnet displacement during MRI. An accompanying cadaver study demonstrated 14 magnet displacements in 16 MRI scans when no external fixation was applied, and no magnet displacements when a head wrap was applied. The authors suggested distinguishing magnet displacement and canting, where canting refers to the magnet not traversing the lip of the silicone housing [5]. In the following, both are referred to as magnet dislocation. The procedure of applying a head wrap, as an educated guess, was adopted as a mandatory requirement in all CI manuals in the MRI guidelines of the major CI companies MED-EL^®, Cochlear^® and Advanced Bionics^® [145–148].

It took until 2008 for the very first adverse event related to CI in MRI to be reported, where the authors carefully state the 'very rare event of magnet displacement during MRI scanning' even though the patient was wearing a head wrap [49]. Once the first report had been published, the attribute 'very rare' was challenged simply by the number of reports on MR-related magnet dislocations, which followed readily thereafter [3, 6, 19, 23, 27, 50, 51]. One elaborated example reports 12 MRI-related magnet dislocations [6]. Therein, magnet dislocation is described as a serious complication, which can lead to transdermal magnet extrusion or meningitis caused by the infection spreading along the implant. Regarding canting, one group proposed the application of gentle firm pressure over the scalp to bring the magnet back into place and thereby avert the need for surgical revision [3].

In 2015, the MED-EL Mi1200 Synchrony CI emerged with the distinctive feature of a freely rotating and laterally polarised internal magnet. It was claimed that therewith the problem of magnet dislocation and demagnetisation had been solved [26, 146]. As a result, the device was approved for MRI at 3 T with an internal magnet in situ.

Another aspect of CI recipients in MRI examinations which is gaining immensely in importance is pain and discomfort, which has been explicitly mentioned as a concern only recently [3, 6, 19–26].

For instance, it was reported that 5 out of 18 CI recipients entering MRI were unable to complete the indicated MRI scans due to pain, although only one of them experienced magnet dislocation. In addition, yet another patient in this study had already experienced discomfort merely by approaching the MRI scanner [27]. Sedation was mentioned as a self-evident counter measure to discomfort and pain in some reports [8, 27], and additionally an anxiolytic was provided at request [3]. By contrast, Medtronic^® recommend not to sedate in their manual for physicians concerning MRI examinations in DBS recipients, so that they can provide feedback about any problems occurring during the examination [149].

Besides the magnet displacement risk, internal magnets also cause major imaging artefacts (as explained in section 2.3.4) [11, 14, 21, 26, 69, 73, 75]. This compromises the diagnostic value of the imaging procedure, where the extent of the image disturbance depends on the MRI sequence used. Commonly used sequences can show circular signal voids with a radius of up to 8 cm from the centre of the implant [26]. Assuming the signal void resembles a sphere, and a quarter of the artefact is inside the head, the obscured volume would equal approximately half a litre, which equals one third of the average human brain volume and is illustrated in figure 4(H). However, it is worth noting that the MR image still features imaging artefacts after magnet removal due to the implant housing and its electronics. Thus limited diagnostic value in the vicinity of the implant is still to be expected [68].

It has been claimed that the risk involved when surgically removing and replacing the magnet for scanning is higher than magnet dislocation [8]. This may be true if the probability of magnet dislocation related to MRI in a CI recipient is 0.59 % as assessed here: [6]. This value is derived from 12 dislocations, which occurred in a collective of 2027 CI recipients over the time span of more than a decade. However, CI patients were only allowed to undergo MRI examinations in the last four years of the mentioned time span, and the number needed for a meaningful probability assessment of CI patients who underwent MRI examination was not reported. Therefore, the derived probability of 0.59 % for magnet dislocations among all implanted CIs should not be taken as the correct value for the probability of adverse events in MRI procedures. In the same study, a total of three CI patients underwent MRI between January and May of 2013. If one extrapolates this number to the above mentioned period of four years, one can assume that around 29 of the 2027 CI patients underwent MRI. This rough estimate puts the probability of MRI-related magnet dislocation in CI recipients up to more than 40 % (12 out of 29). Others state a risk of up to 15 % even with head wrap application [3]. So how can the likelihood of magnet dislocation be diminished?

Since Gubbels et al proposed the head wrap technique in 2006, it has become a strict requirement during the MRI examinations of CI recipients at field strengths higher than 0.3 T. This is in spite of several reports on adverse events occurring with the application of head wraps. Furthermore, Gubbels and co-workers suggested characterising the amount of necessary pressure applied by a head wrap to prevent magnet dislocation. To our knowledge, no such study has ever been published. The data from the initial cadaver study did not seem to have been verified nor systematically investigated against the hypothesis that a correlation between magnet dislocation prevention and the application of a head wrap exists. The magnetic field strength, however, has increased in clinical practice from 0.3 T over 1.5 T to 3.0 T, and exerts a force that is proportional to it, as explained in section 2.3.2. Let us estimate the forces exerted by a head wrap necessary to prevent magnet dislocation. In a study on the MRI-induced torque of a typical internal CI magnet, a magnet 12 mm in diameter was investigated, exhibiting a torque of 0.2 Nm in a 1.5 T MRI scanner [18]. This means that a force of 33 N applied to the rim of the magnet is necessary to counteract the torque and therefore prevent magnet movement. In other words, an equivalent of 3.3 kg weight has to be applied to the edge of the magnet to prevent it from moving when exposed to a 1.5 T MRI scanner. Without a head wrap, this task is left to the silicone pocket in which the magnet is contained, together with the surrounding tissue. Assuming that it is sufficient to apply this weight, not on the edge of the magnet but on an area of one cm² over the scalp, the necessary pressure to counter the magnet torque is equivalent to 3.3 bar or 3.3 kg cm⁻². When choosing to apply this pressure with a head wrap, every cm² covered with the head wrap must experience the same amount of pressure. For example, a 10 cm wide head wrap, as proposed in a Cochlear^® CI manual [145], bound around a head with a 60 cm circumference and providing a load of 3.3 kg cm⁻² would result in a total load of about 2 tons. Instead, a stiff piece of wood with a plane surface and dimensions of e.g. $5\times5\times2$ cm³ placed between the head wrap and the scalp directly above the magnet would be better suited. The idea proposed herewith of using a splint is not new. However, folded paper, plastic cards [19] or moulds [8] would either not be stiff enough, or still apply pressure to the whole scalp surface, instead of only where the magnet is situated underneath, therefore not serving the purpose that a splint should. Nevertheless, the principle of these methods was adopted in some manufacturer's manuals in 2012 [6, 145].

If a CI recipient is scheduled for an MRI exam, it is of paramount importance to be aware of the manufacturer's model-specific MRI recommendations. All companies instruct users to remove external components prior to MRI examination. How are CIs finally evaluated with respect to their risk-benefit-assessment in MRI? Unambiguous statements are rare. One review about the MRI safety of CIs concludes that MRI in 3 T scanners is generally considered unsafe, and that 1.5 T MRI is contraindicated for CI recipients unless the MRI indication outweighs the risks of MRI examination [22]. This statement raises two important points. First, some current CIs are certified according to regulations as MR-conditional with the magnet in place for 1.5 T, and for 3.0 T without the magnet. One is even certified MR-safe for 3.0 T with the magnet in place. Consequently, the statement of 'generally considered unsafe (3 T) or contraindicated (1.5 T)' is no longer valid. However, Cochlear^® advises magnet removal before every MRI procedure on CI recipients [145]. MED-EL^® propagates safe MR procedures in CI recipients with the magnet in situ, but a supportive head wrap must be placed over the implant. Furthermore, they point out that performing an MRI without a head wrap could result in pain in the implant area [145, 146]. These recommendations stand in contrast to the reports on the adverse events mentioned earlier. Secondly, most reported cases of MRI indication in CI recipients are due to severe, often life-threatening medical conditions such as NFII, brain tumours, potential breast cancer, etc [3, 8, 27, 49, 51, 66, 67, 143]. It is argued that the MRI indication outweighs the risk of MRI examination. To this effect, MRI in CI recipients can be considered safe enough to proceed with the imaging of severe indications. However, the chance of harm, discomfort, pain and the risk of inflammation either by magnet dislocation or surgery is so high that MRI in non-life-threatening indications of CI recipients does not appear to be performed. Should a CI recipient be able to undergo MRI to diagnose a rupture of the anterior cruciate ligament [27]? This could be a starting point for a debate on where the threshold for taking the risk of an MRI examination is, and how we can provide access to MRI for every neural implant recipient, so that the treatment of a disease does not restrict the diagnosis and treatment of another.

The first adverse event with DBS hardware related to MRI was reported in 2003, when a patient underwent MRI prior to pulse generator implantation with two unconnected, implanted electrode leads fixed outside the RF head coil in a straightened manner. The patient experienced transient focal dystonia and ballism immediately after MRI [30]. Two years later, a different DBS patient underwent MRI seven months after pulse generator implantation. The employment of an RF body coil had caused hemiparesis, which was observed upon the patient exiting the MR scanner. After the imaging procedure, a haemorrhage surrounding one of the two implanted probes was found. According to the authors, all the body-averaged SAR values ranged from 0.57 to 1.26 W kg⁻¹ with local SAR values of up to 3.92 W kg⁻¹ [31]. Consequently, the reported exceeding of limits resulted in responses within the community reminding responsible staff to follow the manufacturers' recommendations [156]. No incidents were reported throughout the next six years except for the failure of two pulse generators during the MRI procedures [151, 157]. In 2011, a case report on dyskinesia and tremor during a postoperative MRI procedure was published [32]. The scanning had to be aborted, however, no abnormalities could be found in the MR image. A second, non-stereotactic MRI was performed on the second postoperative day before pulse generator implantation. This imaging procedure revealed hyperintensities along the whole length of both electrode leads, suggesting oedema, which, according to the publication, was associated with patient movement during the first MRI scan [32]. A different case also reported hyperintense images observed in an MRI scan on an electrode lead 12 months after implantation, where the lead was connected to its pulse generator. The reporting authors assumed that the observed hyperintense signal was an oedema caused by MR-related heating, as the electrode had not been manipulated manually [37]. Similar hyperintensities around DBS probes were reported in other retrospective studies, but their correlation to MRI has not been determined [36, 38, 158].

Precise electrode positioning is crucial for therapeutic success, and avoids the need for additional surgery for later repositioning [157, 159, 160]. The postoperational probe localisation procedure, therefore, is essential. One way of replacing MRI with electrode localisation is the image fusion of preoperative MRI with a postoperative CT image. It is beneficial that the imaging artefacts caused by the electrode were claimed to be smaller in CT than in MRI, resulting in a more precise image [161]. Consequently, image fusion was declared to be a safe and fast technique for the postoperative assessment of the DBS electrode position [153]. A disadvantage of the fusion technique is that morphological information about the targeted tissue is lost and postoperative complications, such as haemorrhage or infarction [161], are much less distinct. Besides, the patient is exposed to ionising radiation and the image fusion may be impaired by a brain shift related to DBS implantation, causing tissue shifts of up to 4 mm [162]. As explained in section 2.3.4, the DBS electrode is disguised in the MR image by an imaging artefact. This artefact is larger than the dimensions of the active sites [163] and has led to significant discrepancies compared to brain CT [159]. It is known that the electrode position is eccentric to the imaging artefact [96, 161]. Yet, despite the MR artefacts, several studies comparing CT and MRI electrode localisation procedures conclude that MRI is a precise modality [78, 161, 163, 164]. MRI holds the advantage of superior diagnostic value in the identification of pathological abnormalities related (e.g. haemorrhage) or unrelated to DBS, such as tumour and stroke [82]. Therefore, it is desirable to perform MRI if patients exhibit poor or worsening outcomes of the stimulation treatment. In particular, MRI is often used in patients who undergo the implantation of an additional electrode, or reimplantation for more effective treatment [81]. Furthermore, great interest exists in combining DBS with fMRI [17, 165] to acquire more data and a deeper understanding of the brain. For these reasons, many groups rely on MRI for the position verification of DBS probes [32, 157, 163], despite the risk of severe heating of the DBS hardware in MRI [79].

As illustrated in figure 7, the tip of the DBS electrode sits in a most delicate location. Therefore, the MRI-related heating of DBS probes requires particular attention, since the sensitivity of the brain to temperature increase is also paired with no opportunity to sense it. Although tissue in the central nervous system can endure 43 °C for 30 minutes [79], it is strictly recommended by regulatory bodies not to exceed a 1 °C temperature increase.

There is a high correlation between the SAR and heating at DBS electrode tips [82]. Therefore, the field strength of scanners and the SAR in conjunction with DBS hardware is restricted to 1.5 T and 0.1 W kg⁻¹, respectively. According to clinicians, this upper SAR limit is extremely low and limits the range of sequences which can be applied to a point of crucial impracticality [33, 166]. In contrast, studies at 3 T show the electrode lead tip temperature exceeding the limits of regulations [85, 87] and thus confirming the purpose of the SAR limit. Recent manufacturer adaptations allow less conservative settings [47].

The level of SAR applied during MRI is controlled by the MR scanner operator via the machine console and software. However, it has been revealed that large discrepancies exist between the displayed SAR and the actual SAR deposited by the scanner between manufacturers, scanners of the same manufacturer and even between software versions run on a single machine [81, 167]. Mostly, the deposited SAR is overestimated—in some cases by as much as 2.2-fold—which represents a conservative setting. Nonetheless, incorrect SAR monitoring can inappropriately restrict MRI scans even further and may hinder clinical scanning and pulse sequence development. In particular, these restrictions can entail trade-offs in the image acquisition time and spatial resolution, and therewith limit diagnostic utility [168]. Finally, it is crucially important to precisely manage the total power deposition for the safety evaluation of conducting MRI with implantable components.

Heating and the implant-correlated heating of all kinds of implants are problems that have been tackled in many studies [82, 169–174]. The major present-day challenge is the handling of the wave characteristics of the RF pulses, which makes it difficult to provide general solutions to the problem. Each case depends strongly on the geometry of the MR system, the patient, the implant and the scanning parameters. Furthermore, the position and orientation of the implant and the patient with respect to the electromagnetic fields has a great influence as well. Thus, it is challenging to derive general solutions for MR safety. An interlaboratory comparison study has pointed out widely varying results among groups evaluating the heating of implants, despite following current guidelines [84], in agreement with the aforementioned challenges. Consequently, implants have to be tested for individual scanning parameters and MR units. This, as well as the numerous influencing parameters, have made experimental work tedious, and a useful outcome remains rare. Therefore, simulations are currently used to compute the distribution of the electromagnetic fields and SAR hot-spots (see ISO/TS clause 10 [57]). As a result, the evaluation of implant heating in MRI using simulation software has become an indispensable tool [84, 175, 176].

Device manuals usually provide clear recommendations about whether and how MRI can be conducted safely. It is important to be aware of the influence the condition of a device can have on its heating behaviour. Some devices have special MRI modes, others should be turned off during examination. The condition of leads—e.g. whether they are externalised—has a tremendous influence on their heating behaviour, and the manufacturers' recommendations should be strictly met. Specifically in the case of DBS, it has been shown that the way probes are implanted strongly influences the heating characteristics during MRI. This mainly affects the excess lead management, where concentric loops in the axial plane around the burr hole may reduce heating [86]. The excess extension lead should be coiled around the perimeter of the pulse generator [177]. Furthermore, the use of a head transmit/receive coil dramatically reduces electrode tip heating when compared to body coils. This is because the pulse generator and a portion of the extension lead is outside RF exposure, which results in less induced current and consequently in less heating [80]. Sequence selection and development can make a contribution to the reduction of heating. Sarkar et al developed an alternative to spin-echo sequences which has been useful for MR-guided DBS procedures [83]. The presented short tau inversion recovery sequence features ultra low SAR while still providing adequate tissue contrast in the brain tissue of DBS recipients. However, the sequence exceeded approved limits once patients had been fully implanted [83].

Corporate research has led to technical innovations that address the multiparametric effects of heating from several technical aspects. For one, it has been shown that coiling wires can reduce the occurrence of heating significantly compared to straight wires [140]. Medtronic^® advanced this by introducing a braided body that acts as an RF shield and a dissipation surface for leads to reduce heating. In vitro experiments show how the temperature increase is distributed over the whole lead length rather than concentrated at the electrode tips [178]. Another approach to reducing implant heating by making adjustments on the scanner side involves the design of implant-friendly RF coils [179].