1 Introduction

Recently, various types of PET–MRI systems have been developed by a number of research groups, and several PET–MRI scanners are now commercially available [1–15]. Almost all of the PET detectors used in these PET–MRI systems have no depth-of-interaction (DOI) capability, even though the systems have small-ring diameters to be used for rodent and human brain scanning [1–6, 9, 10, 14], except in three reports [7, 8, 13].

The small-ring-diameter PET scanner with the DOI–PET detector can realize high spatial resolution and high sensitivity. The closely placed detector positions of the small-ring diameter which reduce the influence of collinearity of the annihilation photons can achieve higher spatial resolution compared with large-ring PET scanners. In addition, the large solid angles of the small-diameter rings achieve a higher scanner sensitivity, compared with the large-diameter rings, especially in the case of the same axial length of the scanner. On the other hand, closely placed detector positions near the measured object increase the annihilation photons entering the scintillation crystals obliquely in the periphery of the field of view (FOV). The oblique entry of the annihilation photons generally degrades the spatial resolution of the PET scanners without DOI capability (this is known as parallax error). On the other hand, the DOI–PET detector can suppress the parallax error and then realize a high sensitivity without degradation of the spatial resolution.

Based on these considerations, we have been developing an integrated PET–MRI system in which PET detectors with four-layer DOI capability are positioned close to the measured object so that high spatial resolution and high sensitivity are achieved. The four-layer DOI encoding method which we reported previously [16] can identify all four-layer crystals with photo-detectors attached to only one side of the crystal block. With this method, one can decrease the number of photo-detectors and their readout circuits which may influence on the MRI images in the PET-MRI, compared with other DOI detectors, such as dual-end readout [17, 18] and stack-type [19–21] DOI detectors.

As a first step, we investigated the reciprocal influences of the four-layer DOI–PET detector and the MRI system [22]. In this previous experiment, we used LGSO scintillation crystals whose size was 2.9 mm × 2.9 mm × 5.0 mm, and we are able to achieve a sufficient detector performance in simultaneous measurements with an MRI. To achieve higher spatial resolution, such as lower 1-mm imaging resolution, we changed the dimensions of the scintillation crystals to 1.45 mm × 1.45 mm × 4.5 mm. In addition, the scintillation material was changed from an LGSO crystal to a lutetium yttrium orthosilicate (LYSO) crystal, because the Gd included in the LGSO was magnetic material and the static magnetic field was distorted by the LGSO in the previous prototype detector.

We evaluated the performance of a prototype four-layer DOI detector using the 1.45 mm × 1.45 mm × 4.5 mm crystals attached to a birdcage-type RF head coil in a 3.0 T MRI. In addition, we evaluated the influence of the four-layer DOI detector on the MRI image.

2 Materials and methods

A photograph of the prototype four-layer DOI–PET detector is shown in Fig. 1. The prototype four-layer DOI–PET detector consisted of an LYSO crystal block and a 4 × 4 multi-pixel photon counter (MPPC) array (S11064-050 series, Hamamatsu Photonics K.K). The size of the crystal element was 1.45 mm × 1.45 mm × 4.5 mm. The crystals were arranged as 6 × 6 elements × 4 layers with reflectors. The reflector arrangements in each layer are shown in Fig. 2a. RTV rubber (KE420, Shin-Etsu Chemical Co., Ltd.) was applied between the crystals, except for the spaces where the reflectors were inserted. Using this arrangement, we can represent crystal elements in all four layers on one position histogram, as shown in Fig. 2b. The crystal block had a minimum array size to demonstrate the practicality of the crystal identification performance of the four-layer DOI–PET detector. The details of the four-layer DOI encoding method are described in Ref. [16]. The readout pixels of the MPPC were 3 × 3 mm2, and they consisted of 50-µm cells. The crystal block was attached on 3 × 3 pixels, as shown in Fig. 3. The 50 Ω terminations were applied for each MPPC. The detector and some electric components were packaged in a shielding box made of aluminum plates which was easier to make than the other materials. Because the thickness of the aluminum plates ranged from 1 to 3 mm, as shown Fig. 1, it was possible to achieve a higher shielding property than with shielding boxes made of thin shielding materials, such as copper foil or mesh. The shielding box was covered with aluminum foil (~100 μm thickness) for shielding of light entering through gasps between the aluminum plates. Almost materials in the DOI detectors, such as the scintillator block, MPPC array and aluminum shielding box, were non-magnetic to avoid any influence on the static magnetic field of MRI. However, the contact pins of the MPPC array, the sockets for them, a connector of signal cables, and some circuit elements may include magnetic materials.

Fig. 1
figure 1

Photograph of the prototype four-layer DOI detector used in the experiments

Full size image
Fig. 2
figure 2

a Reflector arrangement of the crystal block in the prototype DOI detector and b the ideal 2D position histogram for the reflector arrangement

Full size image
Fig. 3
figure 3

Positional relationship of the 6 × 6 × 4 LYSO crystal array and the MPPC pixels of the prototype detector. Gray-shaded areas show sensitive areas of the MPPCs. The seven white MPPCs were not used in the analysis

Full size image

Experiments were carried out with the 3.0 T MRI (GE, Signa HDx) and the birdcage-type RF coil (Fig. 4). The diameter of the RF coil was 30 cm. The shielding box with the detector and electric components was attached on the outside of the RF coil elements and inserted into the static magnetic field of the MRI. We placed an electric power supply for the MPPC and a data acquisition system outside the MRI room to reduce the influence of electric noise on the MRI performance; a schematic of the arrangement is shown in Fig. 5.

Fig. 4
figure 4

RF head coil and the four-layer DOI detector

Full size image
Fig. 5
figure 5

Experimental setup for simultaneous measurements of the four-layer DOI–PET detector and the MRI

Full size image

The data acquisition system consisted of a trigger system, amplifiers, and analog-to-digital converters (ADCs). The acquisition system was positioned outside the MRI room and connected to the MPPC array with 3-m-long cables through a gap at the door of the MRI room. The output signals passing through the 3-m-long cable from the MPPC are shown in Fig. 6. The shield of the cable was connected to the shielding box and the ground. The output signal from each MPPC channel was divided into two signals outside the MRI room. The first signal was individually recorded by the ADCs after passing through the amplifiers, and the other signal was used as a trigger signal. The trigger signal was generated by summation of the divided signals from all 9 channels of the MPPC array with a leading edge discriminator (LED) and was finally fed into the ADCs as a gate signal generated with a gate generator.

Fig. 6
figure 6

Signal after passing through the 3-m-long cable, measured with a digital phosphor oscilloscope

Full size image

Influences on the four-layer DOI–PET detector from the MRI were evaluated by comparison of position histograms and energy spectra obtained under three experimental conditions. First, the DOI detector was positioned outside the MRI room to obtain standard data. Second, the DOI detector was moved on the RF coil and evaluated without the MRI measurement. This meant that the DOI detector was affected only by the static magnetic field of the MRI. Finally, the simultaneous measurements of the DOI detector and the MRI were carried out for evaluation of the influence of the MRI measurement on the PET detector. The MRI was operated by the gradient-echo method (TE = 5 and 30 ms) during the data acquisition of the DOI detector. The two MRI sequences of the different TE settings were operated sequentially in the simultaneous measurement. The MRI operations were used not only for evaluation of the PET detector performance in the simultaneous measurement but also for estimation of the uniformity of the static magnetic field at the same time. In all of the measurements, a 22Na point source was positioned in front of the DOI–PET detector. The acquisition time was 5 min under each measurement condition. Temperature control and correction of variance of the MPPC gains were not applied. The operating voltage of the MPPC arrays was 72.48 V for all of the pixels.

Next, the influence of the four-layer DOI–PET detector on the MRI images was evaluated. A 13-cm-diameter cylindrical phantom (CuSO4·5H2O) was measured by the gradient-echo method. First, we obtained a phantom image without the DOI detector. In this measurement, no DOI detector components were in the MRI room, and the door of the MRI room was closed. In the second measurement, the PET detector was attached to the RF coil, and the power supply of the MPPC array was turned off. Under this experimental condition, the MRI was affected by the presence of the PET detector material. Finally, the MPPC was turned on, and simultaneous measurements of the PET detector and the MRI were carried out.

3 Results

3.1 Influence of the MRI on the four-layer DOI–PET detector

Figure 7 shows 2D position histograms for uniform irradiation of the 22Na point source under the three experimental conditions: (a) outside the MRI room, (b) on the RF coil without MRI measurement, and (c) on the RF coil with MRI measurement. Only 511 keV photo-peak events are extracted and represented on these 2D position histograms. Each spot indicates interaction events in certain crystal elements. Crystals in the all layers can be clearly distinguished in the three position histograms. Non-uniformities of the spot positions are caused by insensitive regions of the MPPC array. From comparison with the three position histograms, no degradation of the crystal identification performance was observed.

Fig. 7
figure 7

Position histograms for uniform irradiation of 511 keV gamma rays in the three experimental conditions: a outside the MRI room, b on the RF coil without MRI measurement, and c on the RF coil with MRI measurement

Full size image

Figure 8 shows energy spectra for uniform irradiation of gamma rays from the 22Na point source under the three experimental conditions: (a) outside the MRI room, (b) on the RF coil without MRI measurement, and (c) on the RF coil with MRI measurement (simultaneous measurement). Shoulders above 400 channels are caused by 1275 keV gamma rays from 22Na. The insensitive areas of the MPPC array broaden the photo-peaks of the 511 keV gamma rays. The photo-peaks of the crystals on the insensitive areas are lower than those on the MPPC pixels due to loss of scintillation photons. Overlapping of the photo-peaks for the crystals on different positions of the MPPC leads to the broad photo-peaks of the spectra in Fig. 8.

Fig. 8
figure 8

Energy spectra for uniform irradiation of gamma rays from the 22Na point source in the three experimental conditions, outside the MRI room, in the MRI without MRI measurement, and on the RF coil with MRI measurement (simultaneous measurement)

Full size image

Setting the region of interest on the position histograms, we obtained energy spectra for single crystal elements. Examples of the energy spectra for single crystal elements in all four layers under the three experimental conditions are shown in Fig. 9. Photo-peaks of 511 and 1275 keV clearly appear for each energy spectrum. The ratios of the two photo-peak channels are ~2.4 and nearly equal to the ratio of the gamma ray energies of 2.50 (=1.275/0.511). This means that the linearity of the MPPC outputs around 511 keV is sufficient for evaluation of the energy resolution. Therefore, no saturation correction needs to be applied in the analysis of the energy resolution.

Fig. 9
figure 9

Examples of energy spectra of single crystal elements in each layer for uniform irradiation of gamma rays from the 22Na point source in the three experimental conditions: a outside the MRI room, b on the RF coil without MRI measurement, and c on the RF coil with MRI measurement

Full size image

Tables 1 and 2 summarize the photo-peak positions and energy resolutions for 511 keV gamma rays. Although a shift of the pulse height of the PET detectors was observed, no significant degradation of the energy resolution was noted. This means that the reason for the peak shift of the energy spectra is not the influence of the MRI, but room temperature variation in the MRI room. The room temperature increased by approximately 1° during the experiment. Variation of the temperature around the PET detector slightly increases the temperature of the MPPCs and decreases the internal gains of the MPPCs. The several-percent decrease of the internal gain generally does not degrade the energy resolution. Therefore, we concluded that the MRI measurement did not influence the energy performance of the four-layer DOI–PET detector.

Table 1 Photo-peak positions for the single crystal elements (ch)
Full size table
Table 2 Energy resolutions for the single crystal elements (%, FWHM)
Full size table

3.2 Influence of the DOI–PET detector on the MRI

Figure 10 shows MRI images obtained (a) without the PET detector, (b) with the PET detector (turned-off state), and (c) in the simultaneous measurement. Their signal-to-noise ratios (SNRs) were 36.8, 34.2, and 34.0, respectively. The SNR value did not change between the measurement with the PET detector turned off and the simultaneous measurement, compared with the measurement without the PET detector. This indicates that the electrical noises from the MPPC array, the data acquisition system, and the power supplies were shielded well by the aluminum shielding box. On the other hand, the SNR was decreased from 36.8 to 34.2 by attachment of the PET detector, compared with Fig. 10a and b. It was caused by environmental noise entering the MRI room from outside the room through the cables.

Fig. 10
figure 10

Magnitude images. a Without the PET detector, b with the PET detector (turned-off state), and c in the simultaneous measurement

Full size image

To evaluate the influence of the PET detectors on the static magnetic field of the MRI, we calculated ΔB0 maps with and without the PET detector. The ΔB0 map shows the difference of the B0 from the standard magnetic field in parts per million (ppm). The ΔB0 maps were calculated from the two phase images of the cylindrical phantom by the gradient-echo method operated by two TEs (TE = 5, 30 ms) [23]. The ΔB0 maps for the phantom positioned at the center of the FOV without and with the PET detectors are shown in Fig. 11a, b. Figure 11c shows the difference between the ΔB0 maps without and with the PET detectors. The positions of the PET detector, the head coil elements, and the cylindrical phantom in this measurement are shown in Fig. 11d. The difference of the ΔB0 maps is lower than ±0.1 ppm. The ΔB0 maps for the phantom near the PET detector without and with the PET detectors are shown in Fig. 12a, b. The difference between the ΔB0 maps and the position relationship are also shown in Fig. 12c, d. The ΔB0 dropped to ~1.0 ppm near the PET detector. The PET detector included some magnetic materials, such as the contact pins of the MPPC array and sockets for them, which were generally nickel coated. As a result, the static magnetic field was decreased by the presence of the PET detector.

Fig. 11
figure 11

ΔB0 maps for the phantom positioned at the center of the FOV. a Without the PET detector and b with the PET detector (turned-off state). c Subtracted image between (a) and (b). d Positional relationship of the PET detector, cylindrical phantom, and RF coil elements

Full size image
Fig. 12
figure 12

ΔB0 maps for the phantom positioned near the PET detector position. a Without the PET detector and b with the PET detector (turned-off state). c Subtracted image between (a) and (b). d The positional relationship of the PET detector, cylindrical phantom, and RF coil elements

Full size image

4 Discussion and conclusion

We have proposed a new PET–MRI system of a birdcage-type RF coil integrated with PET detectors to realize high spatial resolution and sensitivity of the PET scanner using a four-layer DOI detector. As the first step, we constructed a prototype four-layer DOI–PET detector for the PET–MRI system and evaluated the reciprocal influence between the PET and MRI measurements with the commercial birdcage-type RF coil. As a result, we demonstrated that the DOI–PET detector operated normally in simultaneous measurements of MRI with no influence from the MRI measurement. The influence of the PET detector on the static magnetic field of the MRI was observed, whereas no influence of noise from the electronics of the PET detector on the MRI images was obtained, even though the PET detector was positioned near the RF coil. On the other hand, some groups reported performance degradation of PET detectors or MRI images in simultaneous measurements [24–27]. The primary reason for differences between these reports and the proposed system is the shielding methods of the PET detectors. Although we used the aluminum shielding box which had the sufficient thickness for noise reduction (>1 mm), they used thin copper foils or no shielding. As a result, there was no influence on the SNR of the MRI image from the PET detector and on the PET signals from the RF pulse of the MRI.

However, the large volume of the shielding material may increase the eddy current by field gradient pulses, compared with the shielding by a copper foil. It is possible to cause artifacts in some kinds of MRI sequences, such as N half artifact in echo-planar imaging [28, 29]. We are evaluating the shielding performance and the influence of the eddy current for various shielding materials. We will change the shielding material to a material that achieved the best performance.

We expect that the performance of the PET–MRI system can be improved by the following modifications: First, removal of magnetic materials in the PET detectors will suppress the influence on the static magnetic field of the MRI. The static magnetic field was decreased by the contact pins of the MPPC array and the sockets for them in the experiment, as shown in Fig. 12b. Use of surface-mounted MPPCs which have no contact pins, such as S11828-3344M, will improve the homogeneity of the static magnetic field. Second, decreasing the insensitive area of the MPPC array will improve the performance of the PET detector, especially the energy resolution. Energy resolutions of the PET detector above 20 % are caused by scintillation light loss in the insensitive area of the MPPC array, because the energy resolution depends on the number of photons collected in each MPPC pixel. By use of an MPPC array with a small insensitive area, the energy resolution will be improved and can be achieved to be better than 20 %, and the crystal identification performance will also be improved. Finally, the use of an RF coil dedicated to the proposed PET–MRI will improve the quality of the MRI images. The dedicated RF coil tuned with the PET detectors can achieve a higher imaging performance, compared with the commercial RF head coil, which is not dedicated to the presented system.

As a next step, we are developing a one-ring prototype of the integrated PET–MRI system which consists of eight PET detectors and a prototype RF coil dedicated for PET-MRI. The prototype RF coil will be tuned with the PET detectors attached to the RF coil elements. In addition, surface-mount MPPCs which have a small insensitive area between the MPPC pixels (e.g., S11828-3344M) will be used instead of the S11064-050 which was used in the present study. We expect that these modifications will improve the MRI images obtained and the performance of the PET detector in simultaneous measurements.