1 INTRODUCTION

With an aging population and the rising prevalence of cardiac disease, cardiovascular implantable electronic devices (CIEDs) are used more frequently. Today, there are more than 3 million Americans with CIEDs,1 and the number grows by 80 000 annually.2 It is estimated that 50%–75% of these patients may need MRI for cardiac or noncardiac indications,3 with many patients requiring repeated examination.4

Several studies have assessed the safety of MRI in patients with CIED leads connected to working devices. Some patients, however, require procedures that lead to “lead extraction”, leaving fractions of the device in situ, in which case the original safety studies of the intact device are not applicable. For example, patients undergoing heart transplant usually have a CIED in place, often with multiple leads in the vascular space.5 At the time of surgery, the physician attempts to remove the device by cutting the leads and extracting the pulse generator (leaving behind an exposed metal contact). Complete lead removal, however, is not always possible, especially when there is adhesion of leads to the vessel wall or significant calcification.5 This leaves a sizeable cohort of patients with fragmented retained leads (FRLs).6, 7 Magnetic resonance imaging is needed in most of these patients either for cardiac indications (e.g., rejection monitoring, perfusion assessment) or neurological and orthopedic exams. Today, however, there is no consensus on the safety of MRI in patients with retained cardiac leads. Although MR-conditional CIEDs have been available since 2011, the conditional labeling of these devices applies only to the intact device and leads in their originally intended configuration. As such, studies assessing the MRI safety of pacemaker or implantable cardioverter defibrillator generators mostly exclude patients with FRLs.8, 9

Due to lack of guidelines, the decision on whether to perform MRI on a patient with a FRL falls upon the physician’s judgment, and is often based on his or her assessment of risk-to-benefit ratio. The goal of this work is to provide a comprehensive assessment of MRI hazards in patients with FRLs, to provide actionable information to clinicians encountering such cases. We expanded our previous work, which assessed specific absorption rate (SAR) amplification around patient-derived FRL models in a single homogenous body model10 by performing a thorough search for the worst-case heating scenario in body models with 30 unique combination of permittivity and conductivity covering a range of low to high values reported for biological tissues. Additionally, we included new FRL models with folded trajectories and double leads in close proximity, as these topologies are shown to alter RF heating.11 To provide action points for clinicians, a conservative estimation of CEM43 for different exposure times and B1+ values was provided along with thresholds of tissue thermal damage. We also performed experiments measuring RF heating around fragments of commercial CIED leads implanted in gel phantoms to gain additional confidence as to typical level of RF heating around FRLs in vitro. Finally, we assessed other sources of MRI hazard including the potential for unintended tissue stimulation.

2 METHODS 2.1 Sources of MRI hazard in patients with FRLs

The technical specification ISO-TS 10974 describes known sources of potential hazardous interaction between MRI fields and an active implantable medical device. These include risk of gradient-induced and RF-induced device heating, potential harm due to gradient-induced vibration, gradient-induced extrinsic potentials, and gradient-induced device malfunction, as well as risks associated with B0-inserted force and torque.12

Gradient-induced device heating is the result of time-varying imaging gradient dB/dt, inducing eddy currents on active implantable medical device conductive surfaces such as enclosures, battery components, and circuit traces. For active implantable medical devices with extended leads that do not contain larger conductive surfaces, there is no known mechanism for MRI-induced eddy current heating to occur in the lead (ISO-TS 10974:9). Therefore, this is not a source of concern in patients with FRLs.

Gradient-induced eddy currents also generate a time-varying magnetic moment that interacts with the static magnetic field (B0), causing vibration of conductive surfaces (and subsequently the device), which could lead to device malfunction (ISO-TS 10974:10). For the same reasons explained previously, this is not a source of concern in patients with FRLs. Finally, the risks associated with lead traction and dislodgment due to B0-inserted force and torque are negligible, as CIED leads are composed of nonmagnetic material. Radiofrequency-induced heating and gradient-induced extrinsic voltages are therefore the only potential sources of MRI hazard in patients with FRLs.

2.2 Radiofrequency heating

Radiofrequency heating is due to the antenna effect, in which the electric field of MRI transmit coil couples with conductive leads and amplifies the SAR of RF energy in the tissue, usually occurring around the tip of an elongated implant.10, 13-17 Temperature rises up to ΔT = 30°C have been reported at tips of abandoned cardiac leads in phantom experiments at 1.5 T.18, 19

Magnetic resonance imaging RF heating of an elongated lead is known to be a resonance phenomenon, depending on lead’s length and the distribution of MRI electric field’s phase along lead’s trajectory.13, 20, 21 We created clinically relevant FRL models based on medical images of 10 representative patients, including trajectories with folds or multiple leads in close proximity, as they may alter RF heating.19, 22-24

Another important factor influencing RF heating is the electrical property of the medium surrounding the lead.25-28 To determine the worst-case scenario, we performed numerical simulations with each FRL model registered to homogenous body models with 30 unique combinations of permittivity and conductivity covering a range of low to high values reported for biological tissues (σ[0.1, 1]S/m, εr[40, 80]). The rationale for using a homogenous body model with varying electrical properties was based on a recent study that showed such a model can predict the worst-case SAR generated in a heterogeneous body mode at a reduced computational cost.29 Finally, the position of the lead within the MRI RF coil affects its RF heating.30, 31 Thus, RF heating was assessed for each FRL+ body model positioned inside the MRI body coil at landmarks corresponding to head, chest, and abdomen imaging.

Temperature rise in the tissue was conservatively estimated by solving the simplified Penne’s bio-heat equation, excluding cooling effects of the perfusion: where T is the temperature; ρ is the density (1000 kg/m3); c is the specific heat capacity of the tissue (4150 J kg−1℃−1); and k is the isotropic thermal conductivity (0.42 W/m−1℃−1).32 For each FRL model, thermal simulations were performed in the human body model that generated the highest local SAR at the corresponding imaging landmark (abdomen, chest, and head), with the input power of coils adjusted to produce B1+ in a range of 1 to 5 µT. The selected B1+ values cover the whole range of routine clinical protocols including high-SAR sequences. To provide actionable information, we calculated the worst-case cumulative equivalent minutes at 43°C (CEM43) for a range of exposure times (1–10 min) and B1+ values (1–5 µT) at 1.5 T and 3 T. CEM43 is currently the accepted metric for thermal dose assessment that correlates well with thermal damage in variety of tissues.33, 34 The calculation of CEM43 requires knowledge of thermal history, as: where Δt indicates integration over the length of exposure; T is the average temperature during the exposure time; and R is a constant equal to 0.25 for TT > 43°C.33 To be conservative, CEM43 was calculated using temperature rise ΔT from the FRL model that generated the maximum heating. Finally, we performed experiments with fragments of a commercial cardiac lead implanted in a gel phantom during RF exposure at 1.5 T and 3 T to gain additional confidence as to level of RF heating observed in vitro. 2.3 Gradient-induced voltage

Extrinsic electric field potentials are gradient-induced voltages that develop between spatially separated electrodes within a single lead or between electrodes of a multilead system (ISO-TS 10974:13). If the FRL is in contact with the excitable tissue, these voltages could potentially cause unintended stimulation. We assessed the possibility of unintended tissue stimulation due to gradient-induced extrinsic potentials along FRLs based on a conservative estimation of electric potentials developed along the length of FRL models as described in ISO-TS 10974:13. These voltage values were then used to calculate a conservative estimation of electric field E induced in the tissue developed between the two ends of each FRL, and the potential for unintended tissue stimulation was assessed in the context of cardiac stimulation thresholds.35, 36

2.4 Patient-derived FRL models

Chest CT and X-ray images of 100 patients with a history of implanted cardiac devices who had been admitted to Northwestern Memorial Hospital between 2006 and 2018 were obtained through a search of the Northwestern Universities’ Enterprise Data Warehouse. Images were inspected by a radiologist (A.R.) for the presence of FRLs. From patients identified with FRL, 10 patients who had both chest X-ray and CT images that clearly delineated FRL trajectory and topology were included in the study. Patient characteristics are given in Table 1. Use of imaging data for the purpose of simulation and modeling was approved by Northwestern Memorial Hospital’s ethics review boards.

TABLE 1. Characteristics of patients with cardiovascular implantable electronic device fragmented retain leads Patient ID# Age Sex FRL location Device type and manufacturer Lead model(s) Date of implantation (month/year) Date of extraction (month/year) Apparent/True length of FRL (cm) FRL1 24 M Left subclavian vein Medtronic dual-chamber ICD Capsurefix Novus 5076–RA 9/2007 1/2013 13.4 & 10.5 SVC Sprint Fidelis 6931–RV 67.9 &10.5 FRL2 56 M SVC/RA junction Medtronic biventricular ICD Capsurefix Novus 5076–RA 5/2002 04/2007 14.2 Left subclavian vein

Sprint Quattro 6944–RV

4193 Attain OTW–LV

9/2005–LV 90.4 FRL3 52 F SVC Boston Scientific dual-chamber ICD

1688 TC–RA

Reliance G 0185–RV

1/2007 08/2011 8.9 & 4.4 Left subclavian vein Abbott St Jude RA lead 13.3 & 6.5 FRL4 64 F SVC/RA junction Medtronic biventricular ICD Capsurefix 5076–RA 4/2007 7/2008 18.0 Attain OTW–LV 41.2 Sprint Fidelis–RV FRL5 66 F SVC/RA Medtronic single-chamber ICD Sprint Quattro 6947-65–RV 4/2012 9/2012 17.2 Left subclavian vein 107.3 FRL6 57 M SVC Medtronic biventricular ICD Capsurefix Novus 5076–RA 4/2003 12/2009 15.5 Left subclavian vein Sprint Quattro 6947–RV 9/2008–LV 54.3 Attain Starfix 4195–LV FRL7 50 M SVC/RA junction Medtronic biventricular ICD Capsurefix Novus 5076–RA 1/2006 6/29/2015 2.0 Attain OTW 4194–LV 3/2011–LV 26.4 Sprint Quattro 6947–RV 11/2011–RV FRL8 57 M SVC/RA Medtronic biventricular ICD 5071–LV 8/2012 05/2013 4.5 Capsurefix Novus 5076–RA 46.9 6947–RV FRL9 64 M SVC Boston Scientific dual-chamber ICD Boston Scientific 4052–RA 6/2003 11/2005 12.7 & 12.7 Left subclavian vein Guidant Endotak Reliance 0185–RV 21.8 & 18.1 FRL10 29 M SVC Medtronic biventricular ICD Capsurefix Novus 5076–RA 2/2005 08/2015 8.5 Left subclavian vein Sprint Fidelis 6949–RV 10/2009–LV 96.0 Attain Ability 4196–LV Abbreviations: ICD, implantable cardioverter defibrillator; LV, left ventricle; RA, right atrium; RV, right ventricle; SVC, super vena cava.

Lead trajectories were semi-automatically segmented from CT images using Amira (Amira 5.3.3; FEI, Valley City, ND) by applying a thresholding algorithm based on an intensity histogram analysis, which extracted a preliminary mask of the hyperdense lead from the CT image (Figure 1A). Lead centerlines were manually extracted and exported to a CAD tool (Rhino 3D; Robert McNeal and Associates, Seattle, WA), where 3D models of FRLs were constructed around them. The CIED leads usually include individual coiled wires extending from the proximal to the distal end of the lead. The topology of these coiled wires changes during the extraction process as the physician applies force to extract the lead due to surrounding adhesions. We used X-ray images to reconstruct the details of the FRL structure, including the number of microloops and the variation in their pitch (Figure 1B,C). To correctly position the FRLs in the human body model, we created a triangulated surface of each patient’s ribcage from CT images, which was then used to align and coregister the FRL of that patient to the ANSYS multi-compartment human body model37 (Figure 1D). Once the FRL was in its correct position, we merged different body tissues to create a homogenous body model that was then assigned to a range of different electrical conductivity (σ = 0.1, 0.2, 0.4, 0.6, 0.8, and 1 S/m) and permittivity (εr = 40, 50, 60, 70, and 80) to cover all relevant biological values.38 This is reasonable, as a recent study has shown a homogeneous body model with conductivities varied in a range of low to high values (0.01 S/m → 1 S/m) can predict the full range of SAR variations predicted by a heterogeneous body model.29

Steps of image segmentation and fragmented retained lead (FRL) model construction. (A–C) Computed tomography images were used to extract the 3D trajectory of FRL, whereas X-ray images were used to reconstruct the FRL’s structure (e.g., number and pitch of loops). White arrows show the FRL on each image. (D) A triangulated surface of patient’s ribcage (green) was created and aligned with the ANSYS multicompartment body model (blue) to position the patient-derived FRL model inside the ANSYS human body model. (E) Body model positioned inside the MRI coil

2.5 Magnetic resonance imaging coils

Numerical models of two high-pass birdcage body coils (620 mm length, 607 mm diameter) were implemented in ANSYS Electronic Desktop and tuned to their respective Larmor frequencies: 64 MHz (1.5 T) and 127 MHz (3 T). Each RF coil consisted of 16 rungs connected at each end to two end rings and shielded by an open cylinder (1220 mm length, 660 mm diameter). The coils and shields were made of copper, and their dimensions were chosen based on a typical clinical body coil. A quadrature excitation was implemented by feeding the coils at two ports on the bottom end ring that were 90° apart in position and phase. Coils were tuned by lumped capacitors distributed at the end-ring gaps (86.5 pF for 64 MHz and 17 pF for 127 MHz) and matched to 50 Ω using a single capacitor at each port (100 pF at 64 MHz and 45.6 pF at 127 MHz) in series with an ideal voltage source with a 50-Ω internal resistance.

2.6 Specific absorption rate and thermal simulations

A total of 1800 numerical simulations were performed, with 10 realistic FRL models incorporated in body models with 30 unique combinations of electrical conductivity and permittivity, positioned in two MRI body coils (1.5 T and 3 T) at three different imaging landmarks (abdomen, chest, and head). The FRL wire diameters ranged from 0.35 to 0.6 mm and were assigned as titanium for the simulation. The maximum value of 1g-averaged SAR (referred to as MaxSAR1g) was calculated inside a conformal cylindrical volume (diameter = 2 or 3 cm depending on the shape of the FRL) that surrounded the FRL as illustrated in Figure 2.

Top row: Trajectories and detailed structures of FRL models extracted from 10 patients (namely FRL1-FRL10). FRL 4 and 7 represent folded trajectories. FRL 9 was a case of a patient with two closely situated retained leads. Middle row: Front and side views of reconstructed FRLs and their relative locations in the human body model. Bottom row: Position of the body model inside the MRI coil at different imaging landmarks (i.e., abdomen, chest, and head). The high-resolution mesh area around the lead in which 1g SAR was calculated is also shown

To ensure good numerical convergence, the initial mesh was set such that the maximum element size was <2 mm on the FRL, <2 mm in the surrounding conformal region in which MaxSAR1g was calculated, and <50 mm in the body. The ANSYS HFSS was set to follow an adaptive mesh scheme with successive refinement of the initial mesh between iterative passes until the difference in magnitude of the S-parameters fell below a set threshold of 0.02. Total time taken for each simulation was from 30 min to two and half hours (depending on the length and the shape of FRL) on a Dell server with 1.5 TB memory and 2_Xenon Gold 6140 CPUs each having 18 processing cores. Table 2 gives mesh statistics for a typical simulation.

TABLE 2. Details of mesh statistics for the simulation of FRL4 at 3 T for the abdomen imaging landmark Parts Num tets Min edge length (mm) Max edge length (mm) RMS edge length (mm) Min tet. vol. (mm3) Max tet. vol. (mm3) Mean tet. vol. (mm3) SD vol. (mm3) Lead 22 597 0.0538 0.7218 0.4354 7.9218-08 9.5171-03 1.8853-03 8.6679-04 SAR region 50 6346 0.0566 2.3516 1.5475 7.7631-08 7.6573-01 1.7217-01 1.0858-01 Body 89 169 1.0457 56.6824 24.927 8.3179-03 9.712303 8.747902 1.295603

Thermal simulations were performed in the human body model that produced the highest local SAR at each imaging landmark (abdomen, chest, and head), with input power of coils adjusted to produce a spatial mean of B1+ amplitude (i.e., complex magnitude of B1+ averaged over an axial plane passing through the isocenter of the coil; Figure 2) in a range of 1 to 5 µT.

2.7 Radiofrequency heating measurements

To gain additional confidence as to typical values of RF heating occurring in the tissue surrounding FRLs, we performed phantom experiments with four FRL models reconstructed from commercially available pacemaker leads (Medtronic 5076; diameter of internal wires 0.1 mm), with trajectories mimicking those that generated highest temperature rise in simulations. Specifically, we reconstructed FRL #6, which generated highest temperature rise at 1.5 T for all three landmarks, and FRL models #4, #3 and #9, which generated highest temperature rise at 3 T at abdomen, chest and head landmarks, respectively (see Figure 3).

Experiments were performed in a Siemens Aera 1.5T scanner and a Siemens Prisma 3T scanner (Siemens Healthineers, Erlangen, Germany). The FRL models were positioned inside a human-shaped container filled with a gel (12-cm thick) prepared by mixing 10 g/L of Polyacrylic Acid (Sigma-Aldrich, St. Louis, MO) and 1.32 g/L sodium chloride with distilled water. The anthropomorphic phantom was designed using segmentation of patient images as described in our previous work.27 Bulk conductivity and relative permittivity of the gel was measured to be σ = 0.46 S/m and εr = 87 using a dielectric probe kit (85070E; Agilent Technologies, Santa Clara, CA) and a network analyzer.

RF heating was measured for two scenarios. First, FRL models were positioned in a location analogous to the middle of the chest, similar to what we observed in patient images, namely position P1 as illustrated in Figure 6A. Second, and to be more conservative, we also measured temperature rise for FRL models in a high field exposure location, namely position P2 in Figure 6A, which was near the phantom wall. To find the location of maximum field exposure, we performed experiments with a simple insulated copper wire (Figure S1), which was orientated parallel to the phantom’s long walls at three different depths from the gel’s surface. The temperature profile showed a right–left asymmetry in local E-field distribution, which was in agreement with the earlier findings.39 The highest temperature rise occurred when the wire was located along the phantom’s left wall and 2 cm from the gel’s surface. This location was chosen as P2 to position the FRL at the maximum E-field exposure. Two fluoroptic temperature probes (OSENSA, Burnaby, BC, Canada) were secured at both tips for all four FRL models. Temperature rise ∆T in the gel was recorded during 10 minutes of RF exposure using a high-SAR T1 turbo spin-echo sequence (TE = 7.3 ms, TR = 814 ms, flip angle = 150° for 1.5T scans and TE = 7.5 ms, TR = 1450 ms, flip angle = 150° for 3T scans). An additional temperature probe was also positioned inside the phantom far from the implant position, to observe the background heating (Figure 6A). The sequence parameters were adjusted to generate the RMS B1+ of 2 µT in the phantom. Imaging was performed using the scanner’s built-in body coil. For each FRL model, the phantom was positioned alternately such that the abdomen/chest/head of the phantom was at the coil’s iso-center. For all experiments, precise positioning of the FRLs inside the gel was ensured by using the support pillars with adjustable height, which were positioned on a gird fitted to the bottom surface of the anthropomorphic phantom. Figure 6A shows the details of the experimental setup.

2.8 Estimation of gradient-induced extrinsic voltages

To obtain a conservative estimation of electric field induced in the tissue, we first calculated the injected Vemf along each FRL using the Tier 1 approach described in ISO-TS10974:13.12 This approach is based on simulated values of electric fields induced on the surface of a cylinder of 20-cm radius by a dB/dt = 100 T/s, and gives the most conservative estimation of extrinsic voltages for leads with length <63 cm. We then estimated the electric-field E induced in the tissue between the two ends of each FRL by dividing the Vemf by the distance between the end tips (Figure S2). These values were compared against cardiac-stimulation thresholds provided in the literature.40-42

3 RESULTS 3.1 Specific absorption rate distribution and maximum temperature rise in the tissue

Figure 3 gives MaxSAR1g values for each FRL model during RF exposure at 64 MHz (1.5 T) and 127 MH (3 T) and for the body model positioned at head, chest, and abdomen landmarks. At each imaging landmark, the input power of the coil was adjusted such that the spatial mean of B1+ amplitude was 2 µT at the coil’s iso-center. Boxcars represent the variation of MaxSAR1g as a function of electrical properties (σ and εr) of the body model. As it can be observed, the specific combination of tissue permittivity and conductivity that produced maximum SAR was different for each FRL model, emphasizing the interdependency of different factors affecting RF heating.24 Plots of variation of MaxSAR1g as a function of conductivity and permittivity are found in Supporting Information Figures S3–S7. In general, lower values of conductivity generated higher SAR in most models, with the trend being substantially more pronounced at 3 T.

Comparison of maximum value of 1g-averaged specific absorption rate (MaxSAR1g) of all leads for each combination of landmark position and frequency. For each simulation, the input power of the coil was adjusted such that the spatial mean of the complex magnitude of B1+ (i.e., 1/2 ‖B1x + jB1y‖) was 2 μT on an axial plane passing through the iso-center of the coil

MaxSAR1g (mean ± SD) was 3.50 ± 4.23 W/kg for the head imaging landmark, 11.34 ± 14.20 W/kg for the chest imaging landmark, and 3.20 ± 6.56 W/kg for the abdomen imaging landmark for RF exposure at 1.5 T (averaged over all body compositions and all lead models; N = 300). MaxSAR1g was significantly higher at chest imaging landmark compared with both head landmark and abdomen landmark (one-tail t-test; p-value < 1−04). There was no significant difference between MaxSAR1g at head and abdomen landmarks (two-tail t-test; p-value = .5).

At 3 T, MaxSAR1g was 11.79 ± 14.52 W/kg for the head imaging landmark, 32.09 ± 34.65 W/kg for the chest imaging landmark, and 6.83 ± 17.92 W/kg for the abdomen imaging landmark (averaged over all body compositions and all lead models; N = 300). MaxSAR1g was significantly higher at chest imaging landmark compared with both head landmark and abdomen landmark (one-tail t-test; p-value < 1−04). MaxSAR1g at head landmark was slightly higher MaxSAR1g at abdomen landmark (one-tail t-test; p-value = 1−04). At all imaging landmarks, MaxSAR1g was significantly higher at 3 T compared with 1.5 T (one-tail paired t-test; p-values < 1−04).

Figure 4 shows histograms of MaxSAR1g distribution at each field strength and for each imaging landmark. An exponential probability density function (mean of µ) was fitted to each distribution using the MATLAB and Statistics Toolbox Release 2019a (The MathWorks, Natick, MA). For each case, the model that created the most extreme data point on the SAR distribution was used for subsequent thermal analyses. Figure 4 shows that the approach was conservative, as >99% of cases generated SAR values below this limit.

Normalized specific absorption rate (SAR) histograms pooling N = 300 body models and FRL leads for each field strength and at each imaging landmark. An exponential probability density function (mean = μ) was fitted to the SAR distribution for each case. Red markers show the location of extreme cases associated with the body + FRL model that created the maximum SAR. The value of the maximum SAR and the probability of an individual SAR being less than this value are also given

3.2 Thermal dose

The temperature rise in the tissue after 10 min of continuous RF exposure at 1.5 T and 3 T is given in Tables S1 and S2, respectively. To be conservative, thermal simulations were performed using the body model with electrical properties that generated the highest MaxSAR1g.

At each imaging landmark and for each B1+ value, we calculated CEM43 corresponding to the FRL model that generated the highest RF heating (Figure 5). A review of published data on tissue damage due to thermal exposure can be found in Yarmolenko et al.34 For muscle tissue, CEM43 > 80 min has been reported to cause chronic damage, whereas 41 < CEM43 < 80 min was associated with acute but minor damages. From Figure 5 it can be observed that for B1+ ≤ 2 µT, RF exposure at both 1.5 T and 3 T generated CEM43 < 40 at all imaging landmarks, indicating no thermal damage for acquisition times < 10 minutes.

Cumulative equivalent minute (CEM43) values calculated at different B1+ levels as a function of acquisition time. The CEM43 values are calculated for the FRL model that generated the maximum temperature rise at each imaging landmark and RF frequency. Reported B1+ values are spatial means of the complex magnitude of the B1+ calculated on an axial plane passing through the coil’s iso-center

It should be noted, however, that calculated CEM43 values for continuous RF exposure are highly conservative, as MRI sequences typically have duty cycles much below 100%. Table 3 gives examples of typical clinical sequences for neuroimaging, cardiac imaging, and body imaging with their corresponding B1+ and acquisition times. The sequence-specific CEM43 calculated for the FRL model that generated the maximum SAR is also given. As it can be observed that CEM43 remained well below 40 for all sequences, indicating no thermal damage.

TABLE 3. Sequence parameters and scanner-reported RMS B1+ values for different imaging protocols on 1.5 T and 3 T scanners 1.5 T imaging Protocol TE (ms) TR (ms) TA (min) FA (°) B1+ (µT) CEM43 Head landmark SE T1 SAG 10 450 2:09