Final design and initial results of the first MINDView brain PET insert prototype

May 2, 2016, 5:20 PM


A. Gonzalez Martinez, I3M Valencia


The first prototype of the MINDView project, a brain PET insert MR compatible, is currently being assembled. The scanner is composed of 3 rings of 20 detector blocks each. The detector block includes a monolithic LYSO crystal with 50x50x20 mm3 and a custom 12x12 SiPM array (TSV-type). The system defines an axial and transaxial field of view (FOV) of about 150 mm and 240 mm, respectively. Detector blocks are kept at a stable temperature in the range of 20-25ºC using controlled temperature air cooling. The X and Y light projections of each detector are measured and from them the planar and depth of interaction (DOI) positions deduced. Here, several methods have been studied namely traditional Center of Gravity (CoG), Rise to Power (RTP), Fitting profiles to the light distribution but also Neuronal Networks. Using the standard approaches (CoG and RTP) it has been possible to characterize the DOI with a resolution of about 5 mm. This makes it possible to reach an average detector spatial resolution (without source finite size corrections) of about 2.6 mm for the whole crystal volume, improving to 1.7 mm, at DOI’s values closer to the photosensor. Average energy resolution ranging from 17% at the crystal entrance down to 16% near the photosensor, is obtained. Parts of the detector ring have been successfully tested for RF shielding and eddy currents. This was carried out by using a Radio frequency (RF) screen structure based on carbon fiber composites with specific thickness and orientations.



To show the latest performance of the detector blocks building up the first prototype of the brain PET insert under the MINDView project. The mechanical design, including final detector geometry, cooling system and the RF shielding method is also presented. Currently, the first ring is being built with the aim to run experiments in Spring 2016 at the TUM-MED in Munich inside the Siemens mMR. These results will additionally be reported.

Material and methods

One of the objectives of the MINDView project is to develop a brain PET insert compatible with most of the already installed clinical MR based systems (3 Tesla). The main components of the first prototype have already being defined and the system is currently being commissioned. The PET is composed of 3 rings of 20 detector blocks each. The system has an aperture of about 330 mm to allocate a brain dedicated birdcage RF coil. This defines a geometry with an axial FOV of roughly 150 mm and a transaxial FOV of about 240 mm, see Figure 1.

The detector block uses a rectangular monolithic 50x50x20 mm3 LYSO crystal [1]. Using 20 mm thick crystals without trapezoidal shape [2][3] challenges an accurate 3D determination of the 511 keV photon impact. To solve this, arrays of 12x12 SiPMs have been custom designed at i3M. In contrast to the pitch distance of standard arrays built by the partner organization SensL of 4.2 mm, the MINDView arrays have a slightly larger pitch of 4.36 mm, with an active area of 51x51 mm2 reducing border effects in the crystal, see Figure 2. The SiPM package is TSV (through silicon vias) and they were already successfully tested in magnetic field environments [1].

SiPM arrays are readout through 40 cm long flexible PCBs avoiding connectors (typically containing Nickel) in the useable PET-MR FOV region. Each row and column of the SiPM array is digitized. Characterizing the X and Y projections of the light distribution allow us to accurately determine the planar and depth of interaction (DOI) photon impact position [4]. The DOI is obtained through fits to the measured light distributions [5] or estimated by calculating the ratio of the energy to the maximum SiPM row or column [1][4]. The preferred crystal surface treatment is all faces black painted except the one in contact to the photosensors array.

A new data acquisition system has been developed, including an ADC board allocating up to 66 channels with 12 bit precision each. This permits to fed 2 detector modules (12+12 channels each) to every ADC board. The photon impact coordinates XY and DOI are calculated in the ADCs using FPGA processing and then are transfered to the workstation using 10 GB Ethernet [5]. Currently, planar coordinates are obtained using Rise To the Power (RTP) calculation with powers 1 (Center of Gravity) or 2 [4].

Towards simultaneous PET-MR imaging both PET and MR systems should not affect or be affected by the other imaging modality. To shield the PET electronics from the Radiofrequency (RF) field without generating eddy currents, we surrounded the PET by a Carbon Fiber (CF) structure, see Figure 1. The CF screens are made by three unidirectional CF layers of 200 um each. The CF layers are always at 90º one to each other. Two sets of CF orientations were tests. The first had the exterior CF layers aligned to the axial axis of the PET and MR. The second had the exterior CF layers at 45ºC, middle layer at -45ºC.


Using the black painted LYSO block, we obtained a DOI resolution ranging from 3.5 to 6.5 mm FWHM, at the crystal center and at the crystal border, respectively. This allows us to efficiently separate photon impacts in at least four DOI regions. The DOI has been characterized for 3 main XY planar regions (see Figure 4) as a consequence of the light truncation, namely center, corners and laterals. As depicted in Figure 3 for a centered ROI, the distribution of light widths follows the exponential attenuation and the impacts can be accurately assigned to the proper crystal DOI (left). On the right side, we depicted the estimated DOI resolution FWHM for the three XY planar regions of interest in the crystal volume.

We irradiated the crystal block with 9x9 Tungsten collimated 22Na sources. The detector spatial resolution was evaluated at four DOI layers. Region 1 and 2, 20-15.5 mm (photosensor=0 mm) and 15.4-11 mm, respectively, showed a slight image compression. However, regions 3 and 4, 10.9-6.4 mm and 6.3-2.1 mm, respectively, showed an almost linear dependence of the measured and real coordinates. A detector spatial resolution FWHM at the crystal center, without correction for the 1 mm source size (collimator diameter 1.2 mm), of 2.6 mm, 2.4 mm, 2.0 mm and 1.7 mm (± 0.3 mm) was determined for DOI layers 1, 2, 3 and 4, respectively. An energy resolution for a centered ROI (15x15 mm2), was determined to be 17%, 16.7%, 16.7% and 16.1%, for DOI regions 1, 2, 3 and 4, respectively. Figure 4 depicts the flood maps for the 9x9 sources as a function of the DOI region.

Concerning the RF field shielding, tests were carried out using a network analyzer. The RF shielding based of CF tubes with orientations of the exterior layers aligned to the axial MR axis showed to be sufficient for the use in PET/MR systems. When EPIs (Echo Planar Imaging) have been acquired, only small ghosting effects were observed. The appearance of eddy currents has also been inspected by PRESS (Position Resolve Spectroscopy) tests.

Discussion and conclusion

This work summarizes the final design of the first prototype of the MINDView project composed of 3 rings of 20 detector blocks each (5x5 cm2) with rectangular monolithic LYSO crystals of 20 mm thickness.
Average DOI resolution nearing 5 mm is reached when the crystal is black painted. In this configuration, an average detector spatial resolution FWHM well below 2.6 mm (σ=0.3 mm) for the whole crystal volume (50x50x20 mm3) is obtained. This improves to 1.7 mm (σ=0.2 mm) close to the photosensor detector, without correction for the 1 mm source size.

Currently, the first ring is being assembled and initial tests inside the mMR PET-MR system at the TUM-MED are planned for Spring 2016. To avoid noise to the PET electronics, a novel RF shielding based on carbon fiber structures has been developed and already successfully tested, also mitigating eddy currents.


[1] A.J. Gonzalez, et al., IEEE Medical Imaging Conference M11-81-1374, Seattle, USA, 2014.

[2] L. Moliner, et al., Medical Physics 39, 5393, 2012.

[3] A.J. Gonzalez, et al., Transaction on Nuclear Science, in press, DOI: 10.1109/TNS.2016.2522179, 2016.

[4] R. Pani, et al., Journal of Instrumentation 10, C06006, 2015.

[5] P. Conde, et al., IEEE Medical Imaging Conference M11-9-1422, Seattle, USA, 2014.

Primary author

Presentation materials

There are no materials yet.