(a) Laser cut catheter tip with the positions of the additional tip markers labeled. (b) Rendering of the handle showing the dual-dials, mechanical locking mechanism for shape locking the catheter's distal end and a proximal luer lock for contrast agent injection and/or guidewire insertion through the central lumen of the catheter. (c) Catheter tip manipulation in 4-directions using the dials demonstrated in the handle design. (d) Handle functional prototype.

Sequence of steps showing the right renal artery probing (a-c) and subsequent contrast injection (d)

Figure 5: a) Automatic slice steering based on reconstructed wire geometry (top) and vascular information (bottom) with a non-moving wire. A slight jittering of the real-time slices can be seen in the top row where orientation of the slices changes between timepoints (see differences in blue circle). b) and c) Time series of real-time images, acquired with slice steering and while performing guidewire motion. The experiments were performed at different MRI sites using two different MRI phantoms.

Figure 3: An additional option allows to use a-priori information of pre-processed vessel masks and the envisioned path for intervention (a). A smoothed path representation is synchronized to the reconstructed wire geometry (b) and sample points are calculated based on the interventional path (c) with which the imaging plane information can be calculated (d). This allows to better align the real-time MRI slices to the vascular characteristics.

Figure 4

Results of active shimming, 1 x 1 x 1 mm³ 3D GRE images and fieldmaps. Excellent recovery of lost signal is achieved in all orientations using pre-estimated shim currents. The width of the signal void approaches the needle width in all cases. Fieldmaps show correction of the underlying ΔB₀. Note that the regions closest to the rod with field information in the ‘With Shim’ case have no corresponding field information in the ‘No Shim’ case due to signal loss. The percentage of signal recovered (not accounting for the needle itself) is indicated for all orientations.

Figure 2

Active shimming hardware (a) Shim insert design with bevel and orthogonal slots for shim wires (b) Shim insert inside the titanium needle (c) Needle placed in holder with compartment for electronics (d) Phantoms with guide holes at different angles used for experiments.

Figure 1.(a) Diagram of the modified 3D SPACE sequence. Compared to the conventional SPACE sequence, each readout in the modified SPACE sequence is shifted by a short time Tshift. The nature of the SPACE sequence helps to reduce the rapid signal dephasing and phase wrapping due to the high susceptibility of the metallic devices. By processing two sets of imaging data (with and without Tshift), the local field map (phase change) induced by the susceptibility of interventional devices can be calculated. (b) Variable-flip-angle of the 3D SPACE with refocusing pulse.

Figure 5.Representative results of the in vivo experiments of a human patient with one tumour of the scapula. (a) Magnitude images generated by the proposed method; (b~c) The positive contrast images generated by the proposed 3D SPACE-based method, 2D FSE based sequence; (d) Magnitude images generated by the SUMO, susceptibility gradient mapping using the original resolution; (e~f) The positive contrast images generated by the 3D SUMO, 2D GRASP (gradient echo acquisition for super-paramagnetic particles).

Figure 3 In vivo transmit RF field (B₁⁺) maps in coronal, sagittal and axial planes. Top: without passive wires. Bottom: with 11-cm-long passive wires. The full-FOV maps including the water bath are shown in the first row of each case, and maps masked to the brain are shown in the second row of each case. By using 11-cm wires, the B₁⁺ uniformity was considerably improved. However, it is also noted that the B₁⁺ is still not so uniform as that of in simulation. This could be attributed to the fact that wires deviated from the central position in the axial plane.

Figure 2 Optimization setup to evaluate the best length for the passive wires (left) and the simulated RF transmit magnitude (|B₁⁺|) maps. The optimal wire length was 11 cm (highest B₁⁺ efficiency near the middle of the brain) which was used for the in vivo experiments.

Figure 1: (a) 3D model of the organ structures. (b) Real organ structures. (c) Final phantom: Organ structures embedded in ballistic gelatin.

Figure 2: (a) Body casting mold. (b) Prostate casting mold including prostate lesions and urethra. The lesions and urethra were placed inside the mold before molding. (c) Bladder and Bladder casting mold.

Figure 3: Typical motion reconstruction at R~18. The computed DVFs were computed using NUFFTreconstructed respiratory-resolved images with 70 respiratory phases (top row, R~18), while the bottom row shows optical flow computed on compressed sense reconstructions. Compared to these DVFs, our model achieves a mean EPE of \(2.48\pm0.42\) mm compared to compressed sense optical flow. However, some motion artifacts remain at the anterior side due to undersampling artifacts.

Figure 1: Schematic of 4D reconstruction and multi-level motion estimation. Long radial acquisitions (7.3 min.) with motion mixing are respiratory-binned and reconstructed to three spatial resolutions by cropping k-space around k₀. These images serve as training data for our multi-resolution motion model -- together with optical flow to inhale, exhale, and midvent positions -- to learn a three-dimensional deformation vector field. The lowest resolution CNN has 3x3x3 3D convolutional filters, the middle CNN has 5x5x5 filters, and the final CNN again uses 3x3x3 filters.

Figure 5. Animation. Proof of concept interactive acquisition in a healthy subject. Changes of orientations are performed interactively both abruptly and through continuous motion between RVOT, PA and SAX. Short transition periods can be observed before convergence to good image quality.

Figure 4. Animation. Pulmonary artery (top) and right ventricular outflow tract (bottom) views in two different catheterized patients where the balloon can be seen. The videos are shown for the conventional scan, the gridded images (input to the network), retrospective compressed sensing reconstruction (CS with temporal TV regularization) and proposed real time images.

Figure 4: (a) The needle feature at the entry point and after insertion were displayed for initial plane. The incomplete needle feature after insertion was caused by misalignment between the initial plane with the needle trajectory. Orientation difference (dθ) and Hausdorff distance (HD) are 19.2° and 13.9 mm. The reference needle was extracted by segmenting needle feature on a high-resolution 3D confirmation scan. (b) After executing the proposed automated workflow, the final plane was aligned with the needle and the complete needle feature is visible. dθ and HD were 1.8° and 1.8 mm.

Figure 2: The workflow for automatic MRI scan plane alignment with the needle using Mask R-CNN and the scan plane control (SPC) module. (I) An initial plane was manually selected based on the needle feature at the entry point on 3-plane localizer images. (II-IV) Scans 1-3 automatically started using the plane selection from the previous step. Mask R-CNN needle localization results were used to automatically select a new plane that aligned with the needle.

Figure 2: (a) Visual comparison between (1) original 4D MR, (2) synthetic 4D T1w, (3) synthetic 4D T2w, (4) synthetic 4D DWI (b=50), and (5) synthetic 4D DWI (b=800) images. Tumors were indicated by the yellow arrows. (b) Tumor motion during 2 respiratory cycles in superior-inferior (SI), anterior-posterior (AP), and mid-lateral (ML) direction measured from different 4D-MRIs. Synthetic 4D-MRIs overall showed a well-matched tumor motion with the original T1w 4D-MRI.

Figure 1: (a) Training: reference DVFs were calculated between the starting frame and following frames and predicted DVFs were generated from the DL model with the two frames as inputs. The training loss was defined as the difference between the two sets of DVFs. (b) Application: appropriate starting frames were registered to other frames using the DL model, and the predicted DVFs warped the 3D MRIs to generate MRI frames at corresponding respiratory phases.

Figure 5, Local pituitary coil and Siemens head coil combined SNR, versus simulation SNR. Siemens commercial head coil SNR is not affected by the local coil angle.

Figure 1, local pituitary coil placement against pituitary gland, and the illustration coil angle with respect to the scanner bed (B₀ field).

Figure 1: Schematic view of the EP workflow under MRI guidance. A 3D bSSFP acquisition is used as a reference for the roadmap & EP mapping. The cavities are segmented and converted to meshes. Then catheter navigation is performed using active catheter tracking followed by EP mapping. RFA ablation is performed under MRI temperature monitoring followed by post-ablation imaging without contrast agent. Evaluation of the lesion size and transmurality is done on the post ablation images after epi/endo segmentation. The segmentation steps could benefit from machine learning.

Figure 2: Specific features of the 3D images. Top: 3D bSSFP acquisition or 3D roadmap acquisition after catheter introduction into the cavities. The artefacts of the catheter (red arrows) are larger than the actual size of the catheter, in particular at the vicinity of the tip. The epicardial fat in hyper signal( blue arrows). Bottom: 3D long-TI acquisition performed after RFA. The created lesion can be visualized in hyper signal (yellow arrows) compared to the surrounding tissue attributed to edema (green arrows). The epicardial fat is highly visible in anterior and posterior view.

Figure 2: Frequency dependence of the labelling signal fitted with a Lorentzian (a) along with the power dependence (b). Image of the catheter (c) and map of the relative signal difference created by the labelling (d). The difference image indicates the transmit B1 profile of the labelling coil.

Figure 3: Image of the perfusion phantom setup (a) and maps of the relative signal difference between the labelled and unlabeled case for two different flow velocities (c and d). A plot of the normalized signal difference shows the exponential decay within the perfusion phantom.

Three displays, all individually acquired. Images are acquired in real time using a real-time Spiral GRE sequence. Each display can show a different view plane in real time by applying interactive tools. For example, a sample catheter trajectory is shown in red. The top-left axial display was rotated. The top-right view shows a perpendicular, but collinear display. The bottom image is perpendicular to the trajectory and shows the probe’s eye view. Parameters also adjustable.

Three reference planes and a main display. Images are acquired in real time using a real-time Spiral GRE sequence. The main display can be changed to show a different view plane in real time by manipulating a graphic slice Rx on any of the three reference planes. For example, a sample plane holding the catheter trajectory is shown in red on the sagittal display in red. By selecting a perpendicular plane, shown in green, the probe’s eye view can be observed in real time. Parameters adjustable.

Fig1. Example results in a 63-year-old female patient who had left stent angioplasty in the left middle-cerebral-artery M1 segment. Preoperative (top row) and post-operative (bottom row) images are shown for time-of-flight magnetic resonance angiography (TOF-MRA, left), pointwise encoding time reduction with a radial acquisition (middle, PETRA-MRA), and digital subtraction angiography (right). Preoperative images indicated a similar degrees of stenosis (blue arrow). Obvious signal loss was observed in post-operative scans, with PETRA-MRA outperforming TOF-MRA.

Table 1. Comparison of PETRA-MRA/TOF-MRA and DSA in measurements of stenosis of MCA

Figure 1: Needle predictions on clinical interventional images (red overlays), generated by U-net trained with synthetic data only.

Figure 5: (a) random selection of simulated images generated using the AustinMan model. All images were simulated in the transversal plane, tilted by up to ±45° to the coronal plane. The needle was either placed in liver tissue from a reasonable angle or completely random (11, 12, 14, 15). (b) Image including the automatically generated needle label (green).

Figure 1: In-phase ZTE without (top) and with 2xFOV extension plus DL image reconstruction (2^nd row) for head&neck, together with corresponding DL derived synthetic CT (3^rd row) and true CT (bottom row).

Figure 3: 2D T2 PROPELLER based automated OAR segmentation in the head & neck (i.e. brain, brainstem, eyes, lens, optic nerves, chiasm, pituitary gland, cochlea, parotid glands, mandible, oral cavity, submandibular glands, larynx, spinal cord, and body contour).

Figure 2: Denoising of undersampled Cartesian images of a liver tumor (arrow).

Figure 3: Denoising of undersampled radial images of a liver tumor (arrow).

Figure 2: Transverse images for four representative test datasets with their corresponding a) MRI, b) MVCT, c) sMVCT, and d) HU difference (sMVCT-MVCT) maps.

Figure 3: Transverse images of a) dose distribution calculated on the sMVCT, b) the identical plan calculated on the actual MVCT, c) the difference in dose distributions, and d) the dose volume histograms for the target and several organs-at-risk. In panel (d), DVHs are shown as dotted and solid lines for the sMVCT plan and MVCT plan, respectively.

Fig.1 A) The pulse sequence diagram of the MR technique, with the red arrows being the navigator line locations. B) Illustration of the Cartesian spiral-in trajectory with Cartesian readout in Kx and spirals in Ky-Kz (PE-Partition) plane. C) Illustration of three-way tensor factorization to generate a multi-dimensional image 𝒜.

Figure 3[GIF]. In-vivo MT multi-dimensional image from two volunteers. A user can view any contrasts, any motion states and any slice available in this image.

Figure 1. a) A healthy volunteer in a radiotherapy immobilization mask fixed on a Qfix Portrait board placed on a modified Qfix Insight board. Two UFL18 coils attached by Velcro straps over and under the Insight board encompassed the volunteer’s head. b) The same volunteer and immobilization with the head surrounded by two attached FL4 coils placed in a Qfix Insight holder. c) The same volunteer in standard head/neck diagnostic position using a dedicated HN20 coil, where treatment masks do not fit.

Figure 3. Graphs (top) and boxplots (bottom) of ratios of image quality parameters obtained using the flexible coil arrangements relative to the parameters obtained using a diagnostic HN20 coil, over all subjects and imaging sequences. From left to right: ratio of SNR, artifact size (AS) and CNR from the UFL18 (blue) and FL4 (red) coil images over SNR, AS and CNR from the HN20 coil images. ***: statistically significant difference (p < 0.001) between SNR_UFL18/SNR_HN20 and SNR_FL4/SNR_HN20.

Figure 2: Resonant frequency offset during gantry rotation starting at 30° and rotating around counterclockwise back to 33°. Offsets were calculated using a navigator inserted into a bSSFP sequence (red) and by measuring the peak offset in a FID sequence (blue).

Figure 4: Illustration of the ViewRay MRI-Linac gantry showing the six mu-metal buckets and rotating passive shim tray. The 60° radial spacing between the buckets results in the periodic B₀ offsets observed at different gantry angles.

Figure 3. Images reconstructed using non-MoCo, MCNUFFT, CS and P2P from a healthy subject. The top row shows phase 1, corresponding to the end-of-expiration, and the bottom row shows phase 5, corresponding to the end-of-inspiration. Blue arrows point to a region of comparison for image quality among the three methods.

Figure 5. Images reconstructed using P2P at 0.35T and at 3T for different subjects for a rough qualitative assessment.

Figure 2: T2* maps obtained from fully sampled numerical phantom with the noise level of 0.1 for gradient delays [1 -1] (top) and [1 2] (bottom).

Figure 4: T2* maps (Top) and T2*-weighted images (TE=5 ms) reconstructed from the in-vivo dataset without correction and with the sequential, respectively joint approach. Arrows highlight a banding artifact, which disappeared on one side after trajectory correction but partially remained on the other side.

Figure 3 Selected parameter maps with heterogeneous low ADC region identified as tumour (red arrow). The ADC and IVIM D_t maps have similar features and become less hypointense at later time points. The VERDICT intracellular fraction f_I is higher in this region and decreases over time. The data are not always sufficient to fit the R parameter, particularly outside of the tumour ROI, resulting in dark sections in these maps.

Figure 4 (a) Summary of parameter values in the tumour ROI over the course of treatment for one patient. The ADC and D­_t are similar and increase at later treatment time points (* indicates statistically significant increase over Scan 1). The f_p from IVIM and VERDICT parameters show more variability and do not demonstrate consistent or statistically significant changes, although the f_I trend is inversely related to the ADC trend. (b) Changes in the ADC throughout treatment for all eight patients.

A,B). Coronal and sagittal view of the tumor are used to plan the scan to locate the clips.

C) 2DT1 scan trough the base clearly show the clips and is used to plan the subsequent scans perpendicular to the tumor base and through centre of the tumour and centre of clip C/D.

D,E). T2 spin echo sequence and T1 gradient echo showing the tumour (dagger), retinal detachment (double dagger) and clip (arrow). The clip location is better visualized on the gradient echo sequence, while a more accurate measurement can be made on the spin echo sequence.

F) T2 image of clip D.

Summary of the differences in clip-tumor distances between the optical peroperative measurement and the MR-measurement. For 51% of the clips no clinically significant difference was found.

Figure 4: 3D cinematic rendering of the reference CT skull and synthetic CT skull from a representative subject.

Figure 5: Transfer learning for 0.8mm³ isotropic synthetic CT skull: results of applying the trained model above to dual echo spiral UTE data (0.8mm³ isotropic) acquired on different 3T scanner. ab) dual echo spiral UTE; c) reference CT skull; d) results of directly applying the previous model trained on radial UTE without any retraining of the spiral UTE data; e) results of applying the model after retraining with single spiral UTE data. Bottom are zoom-in windows showing the details in posterior skulls. Note that all CTs here are 0.8mm³ isotropic, as CT were coregistered to UTE space.

Figure 2: Mean statistical difference maps of pre-HIFU minus post-HIFU, using a finger tapping fMRI paradigm. Arrows show regions with decreased activation after HIFU: green, motor cortex in precentral gyrus; red, motor regions of thalamus; blue, subthalamic nucleus; magenta, globus pallidus; orange, putamen. Threshold for top panel is p=0.07; for bottom panel is p=0.02.

Figure 1: Single subject sample of finger tapping time series data from a voxel in the right motor cortex. The maximum and minimum values as indicated by red dots are 1396 and 1252 respectively.

Figure 2. Comparison of typical Freesurfer and THOMAS thalamic segmentations. TopFreesurfer thalamic segmentation VLp (red) and VLa(green). Bottom THOMAS WMnMPRAGE thalamic segmentation, VLp (yellow) and VLa (blue). Ablation segmentation is overlayed in white.

Figure 1. Volume of ablation in the FreeSurfer VLa, VLp and whole Thalamus, as a percentage of the total ablation volume.

Figure 2. (a) 3D MR-ARFI sequence overview; (b) The proposed k-space undersampling scheme.

Figure 4. Representative reconstructed magnitude images (FUS-ON, positive MEGs) (a) and corresponding ARF displacement maps (b).

Fig. 3: The sensor system is controlled through a GUI. Its initial state is shown in (a), and a snapshot of a real-time acquisition in shown in (b). Instead of raw data (b), one can select velocity (c) or displacement (d) for the GUI’s display window, and switch between display types during the real-time acquisition if so desired. Raw data are saved to file and can be retrospectively re-reconstructed and displayed, which happens to be how (c) and (d) were generated; note the gold color of the selected ‘retro’ button in (c) in (d), as opposed to its pale-blue non-selected color of (a) and (b).

Fig 1: a) A compact, clonable sensor system was developed for ultrasound-based sensors, to monitor physiological motion during diagnostic imaging. Up to four sensors can be used at a time (red arrow), and an RF antenna can be used for synchronization with MRI (green arrow). While this system is still one-of-a-kind, it was built to be clonable, to enable possible collaborations across institutions and geographical locations.

Figure 3. a) Normalized radial acoustic intensity magnitude for 1MHz (left); 650kHz (center); 220kHz (right) along r-coordinate passing the focal point. b) 2D map of the intensity magnitude showing the effects of the RF coil on the magnitude and the focal point (1MHz (left); 650kHz (center); 220kHz (right). For each frequency, a magnified view of the focal point is displayed in the dotted rectangle. For the frequency of 220kHz, a magnified color bar is displayed. First row : With FUS-Flex coil present at 20 mm ; Second row : at 80 mm ; Third row : reference without coil .

Figure 1. a) Illustration of a patient positioned inside a focused-ultrasound transducer. b) Photograph of the FUS-Flex coil.

Figure 2. the 3-ch coil diagrams and photo

Figure 3. The matching/feeding schematic

Fig.5. 0.6 mm isotropic modulated flip angle technique in refocused imaging with extended echo train (MATRIX) images were acquired using the quadrature birdcage/47Rx head coil array at R = 7-fold acceleration. The acceleration direction was on phase encoding and slice phase encoding. H = head; P = posterior; R = right; A = anterior.

Fig.1. The (a.) application scenario and (b.) coil wiring of the quadrature birdcage/47Rx head coil array. The birdcage's size was displayed in red, the coverage of the surface coil was displayed in blue, and the size of a single surface loop was displayed in yellow.

Figure 3: (A) Boxplot of the simulation results showing the performance of the algorithm with and without measurement noise added. There is a high correlation (ρ=0.45) between the final cancellation and the initial feedthrough. (B) Monitoring the feedthrough level over 20 seconds, while heating the coil. Green dashed boxes mark the time intervals where the algorithm was running. The threshold for running the algorithm was -50 dBV. Different initial isolations are set up by adjusting the gradiometer knob.

Figure 1: (A) Diagram of the AWR (bottom) and vector modulator (up). (B) AWR coil (bottom) and the small pickup coil positioned on top. (C) Vector modulator. An Arduino Uno Rev 3 was used to establish the communication between the rheostats and the computer where the experiments were orchestrated. (D) Photo of the overall system.

Current setup of the scanner device. Additionally a receive coil is inserted with a diameter of 136 mm.

Scanner design with only the drive coils shown. In z-direction an FFL can be seen, which can be guided through the scanner along the symmetric axis. Simulated with software for particle imaging^.19

Figure 1. Diagram of a standard permanent magnet DC motor in (a) and an MR-compatible electromagnetic motor in (b). The magnetic components in (a) are labeled in red. The corresponding functional elements for the MR-compatible motor are labeled as blue in (b).

Figure 2. Images of MR safe DC motor and assembly are shown. The major motor components are shown in (a). An exploded view of the rotor, commutators, and end cap assembly is shown in (b). The assembled motor is shown in (c) with the twisted and shielded power supply cables shown in (d).

Fig. 2: The magnitude [V/m] of E-field from VCM (the first column) and the proposed method (the middle column), and the absolute difference between VCM and the proposed method (the last column).

Fig. 3: Evaluation results on brain surface: (a) target overlapping coefficient, (b) E-field peak distance, (c) E-filed correlation coefficient, (d) maximum absolute error.

Fig2.

Fig5.

Figure 3: Spatial pattern of relative % differences between PET images acquired with and without HD-EEG net (blue-red scale) over-imposed on phantom CT image or human brain T1w MPRAGE image (grey scale).

Figure 1: Schematic representation of the two acquisition protocols.

Figure 2:Triggering pulse generation unit.

Figure 3: Waveforms of ECG (blue), the antenna signals before IIR filter (orange) and the IIR filtered antenna signals (green). The inflection points of the IIR waveform and the R-wave in ECG are shown in green and blue circles, respectively

Figure 1: Experimental setup. The laptop is connected via ethernet cable to the 5G router and displays its web-interface.

Table 1: Results of experiments. Transmission times are given as mean ± standard deviation. The reference measurement was perfomed via ethernet connection between the DICOM client and server, therefore quality quantities of cellular networks are not available.

Figure 1: Clock and Data Recovery board prototypes. a) Micrel SY87701 can perform clock & data recovery from 28Mbps to 1.3Gbps, but requires a reference clock. b) Analog Devices ADN2817 operates from 10Mbps to 2.7Gbps, integrates a front end limiting amplifier for small signals, and has a programmable pseudo-random binary sequence (PRBS) generator.

Figure 2: a) PRBS generator uses an Arduino module to configure the ADN2817 CDR device as a PRBS generator. The clock is provided by an SiLabs Si571 programmable clock generator eval board. b) Demonstration of PRBS generation from 400-800 Mbps to a 3.2GHz amplitude shifted keyed modulator and direct detection by an ADL6012 envelope detector. The detector output is yellow, and clock/data (pink/blue) is recovered by the SY87701. The Si571 clock reference is the green trace.

(a) A double frequency parametric resonator included a rectangular conductor with a center conductor. Its top and bottom gaps were filled by variable capacitors. (b) A single frequency resonator with the same dimension was tuned by two capacitors. (c) The two resonators were overlapped through a substrate. (d) The modulator was placed on top of a phantom containing 1% agarose, before being inserted into the center of a 7T magnet. During MR signal acquisition, pumping power was applied on the antenna to activate the modulator whose oscillation signal was received by the volume coil.

When the relative SNR of MR images was measured at different distance separations, the amplified resonator (gray) could retain constant sensitivity for transmission attenuation up to 21 dB (at 4.5 cm separation). In comparison, the frequency modulator (yellow) could retain constant sensitivity for transmission attenuation up to 34 dB (at 11 cm separation).

Figure 5. Comparing mobile and fixed workstations across six dimensions. The Raspberry Pi based mobile workstation (this work, first column) is the most affordable and most portable. *An example of a commercially available portable MRI workstation is from Hyperfine (https://hyperfine.io/).

Figure 2. Illustration of the three hardware setups discussed in this work. The Raspberry Pi was used in combination with keyboard/mouse and display monitor input/output peripherals. The three Internet connectivity modalities studied in this work are fixed broadband (WiFi), cellular broadband (3G/LTE) and satellite-link. Picture at the bottom is an example of the hardware setup highlighted in yellow.

Figure 1: Concept diagram of the (a) switched transmission line gyrator and (b) 3-port circulator.

Figure 5: Measured TX-to-RX leakage when the circulator is terminated with 3 of the 8 channels of a 1.5 T MR coil. The dotted red line depicts the isolation presented by the circulator when the MR coil is loaded with a human head.

Amplifier Schematic. The BFP840 acts as the primary amplification device and the BFP183 acts as a cascode device.

Magnitude and Difference Images (x10). Peak input powers are (left to right): -40dBm, -30dBm, -25dBm, -20dBm. Difference images taken with regard to the reference dataset. A significant change in image contrast is evident upon reaching input powers of –25dBm, further evident at –20dBm. .

Figure 1. a) Diagram of the power splitter model. The final ratio between V_out1 and V_out2 are adjusted by varying the microstrip line. b) Transmission matrices for the hybrid coupler (T_hc), the transmission line (T_tm), and the RAPS (T_RAPS). The scattering matrix of RAPS (S_RAPS) is derived from T_RAPS. We can see from the S-parameter matrix, for any ratio, HP-RAPS is a lossless circuit with all ports well matched and output ports isolated. The output ratio has a tangent dependence on the length difference between the two microstrip lines (S₁₃/S₁₄ = tan[(φ₁ – φ₂)/2]).

Figure 5. Bench tests of the boards with designed magnitude ratios of a) 1:2, b) 1:4, and c) 1:8. The matching and isolation are similar to that of the 1:1 ratio board. The tested magnitude ratios are 1:2.4, 1:4.8, and 1:6.9, respectively.

Figure 4: Mock bore tests using a pair of 5 GHz Biquad antennas at 200, 500 and 600 Mbps show progressive signal degradation of the demodulated bit pattern and eye diagrams, though signal quality was sufficient for clock recovery.

Figure 5: Mock Bore multi-channel deployment and spectrum. Four ASK modulators at 200 Mbps with 3.2, 3.8, 4.4 and 5.0 GHz carriers and Biquad antennas are deployed. A single Taoglas UWB is used for signal reception. The spectrum shows the SINC-like nulls at 200 MHz offsets, and how further signal lobes would over-run to neighboring bands as adjacent channel interference.

Figure 1: Left: The induction coil (yellow) is being connected to a modified drip coffee machine on the patient table of a 3T MRI system. (Right) Coffee was ready at the rear end of the magnet about 5 min after the start of a pulse sequence with an oscillating z-gradient field.