PET/MRI insert using digital SiPMs: Investigation of MR-compatibility

Abstract

In this work, we present an initial MR-compatibility study performed with the world's first preclinical PET/MR insert based on fully digital silicon photo multipliers (dSiPM). The PET insert allows simultaneous data acquisition of both imaging modalities and thus enables the true potential of hybrid PET/MRI. Since the PET insert has the potential to interfere with all of the MRI's subsystems (strong magnet, gradients system, radio frequency (RF) system) and vice versa, interference studies on both imaging systems are of great importance to ensure an undisturbed operation. As a starting point to understand the interference, we performed signal-to-noise ratio (SNR) measurements as well as dedicated noise scans on the MRI side to characterize the influence of the PET electronics on the MR receive chain. Furthermore, improvements of sub-components' shielding of the PET system are implemented and tested inside the MRI. To study the influence of the MRI on the PET performance, we conducted highly demanding stress tests with gradient and RF dominated MR sequences. These stress tests unveil a sensitivity of the PET's electronics to gradient switching.

Keywords: Digital SiPM; MR-compatibility; PET/MRI.

Fig. 1

Hyperion-II^D PET insert with 10 PET modules (2 stacks each) mounted on the patient table top of a 3T clinical MRI.

Fig. 2

One Singles Detection Module (SDM) hosts up to six detector stacks. Plastic optical fibers are used for synchronization and communication. The module is housed inside a carbon fiber shield.

Fig. 3

PET data acquisition scheme: PET data is acquired over a longer time period (green) while for certain smaller time windows (30–120 s) RF and gradient stress tests are performed (red). (For interpretation of the references to color in this figure caption, the reader is referred to the web version of this paper.)

Fig. 4

Noise scan results without (black) and with PET detector (green: power off, blue: power on, red: power on + source (2.8 MBq)). The noise floor is strongly increased and shows broad peaks which are 250 kHz apart, corresponding to the switching frequency of the switched mode power supply. (For interpretation of the references to color in this figure caption, the reader is referred to the web version of this paper.)

Fig. 5

Comparison of the switched mode power supplies (right: unmodified PSU, left: improved version): the shielding of the PSU is improved by replacing all shielding PCBs (yellow) by thick copper plates. Additional fan grilles are installed on the outside to reduce the leaking EM fields. (For interpretation of the references to color in this figure caption, the reader is referred to the web version of this paper.)

Fig. 6

Noise scans without PET (reference, black) and with PET (blue: unmodified PSU, red: modified PSU). Measurements were performed with one PET module (equipped with one stack) and the dedicated mouse coil. (For interpretation of the references to color in this figure caption, the reader is referred to the web version of this paper.)

Fig. 7

Noise scans without PET (reference, black) and with PET (green, without carbon RF shielding). Measurements were performed with one PET module and the dedicated mouse coil. (For interpretation of the references to color in this figure caption, the reader is referred to the web version of this paper.)

Fig. 8

Singles rate (in counts per second (cps), averaged over 5 s intervals, energy cut: (511±30) keV, all pixel around main pixel present) during highly demanding gradient sequences with different switching directions (as indicated with color shaded areas) for two different over voltages (OV) (top: OV=2.9 V, bottom: OV=2.5 V). Only the measurement with the higher OV shows singles rate drops by up to 5.6% during active z gradient sequences. (For interpretation of the references to color in this figure caption, the reader is referred to the web version of this paper.)

Fig. 9

Energy histogram (left: overall, right: photo peak range) for time windows without gradient switching (black, shaded) and with switching z gradients (red) for two different over voltages (OV) (top row: OV=2.9 V, bottom row: OV=2.5 V). (For interpretation of the references to color in this figure caption, the reader is referred to the web version of this paper.)

Fig. 10

Measurement of the bias voltage V_B (top) and bias current I_B (bottom) for two different over voltages (OV) (black: OV=2.5 V, red: OV=2.9 V) as a function of time. In time regions with active z gradients (indicated by green boundaries) a strong ripple on V_B and I_B occurs. (For interpretation of the references to color in this figure caption, the reader is referred to the web version of this paper.)