Cryogenic cesium iodide as a potential PET material

Abstract

Total-body PET (TB-PET) scanners, with axial lengths exceeding 1 meter, provide significantly higher sensitivity than conventional PET scanners due to their greater acceptance of gamma pairs. However, this increased sensitivity often comes from gamma rays detected at oblique angles, leading to substantial multiple scattering along the patient’s body. Additionally, the cost of TB-PET scanners scales with their extended detector length, often costing 5 to 10 times more than conventional scanners. Therefore, optimizing TB-PET performance requires crystals that enhance energy resolution and control costs.

Cesium iodide (CsI), though historically less favored for PET due to its lower stopping power and light yield compared to crystals like LYSO, shows remarkable improvement when operated at cryogenic temperatures ($\sim$100 K). Under these conditions, CsI light yield rises dramatically to about 100 photons/keV, providing excellent energy resolution and good coincidence time resolution at a lower cost—typically 3 to 5 times cheaper than other crystals at parity of radiation length.

In our study, cryogenic CsI crystals achieved an energy resolution better than 7% FWHM at 511 keV and a coincidence time resolution of less than 2 ns. These results demonstrate the potential of cryogenic CsI as a cost-effective, high-performance material for TB-PET scanners.

I Introduction

Positron emission tomography (PET) is a powerful imaging technique that plays a pivotal role in medical diagnostics, employing positron-emitting radionuclides to track biologically active molecules. The annihilation of positrons and electrons produces gamma rays of 511 keV, which are detected to create detailed images of functional processes within the body.

Most commercial PET devices employ lutetium-based scintillating crystals, such as lutetium yttrium oxyorthosilicate (LYSO). This material has optimal performances for PET imaging, providing an excellent stopping power, a good light yield and a fast decay time. However, due to its high cost (45-50 €/cm³), as well as the cost of the associated electronics and readout, commercial scanners are typically limited to small axial lengths and are thus able to cover only a small portion of the patient’s body at a time [1].

Although large TB-PET scanners have been built in the last decade, showing an impressive increased performance (e.g. United Imaging uEXPLORER[2], Siemens Biograph Vision Quadra [3]), the cost of such apparatus is beyond the possibilities of most hospitals.

Therefore, developing a cost-effective TB-PET system will require identifying a less expensive alternative to lutetium-based crystals, ideally one that provides similar performance. In this context, the possibility of using scintillating crystals based on cesium iodide (CsI) is worth being explored. Tab. I shows the properties of CsI compared with other crystals commonly used for PET.

TABLE I: Properties of scintillating materials commonly used for PET, compared with scintillators based on cesium iodide.

Material	$Z_{eff}$	$X_{0}$ (cm)	Density (g/cm3)	Light yield ($\gamma$/MeV)	Decay time (ns)	Emission peak (nm)	Cost (€/cm3)
LYSO	66	1.14	7.4	30,000	53	420	45
BGO	74	1.12	7.1	10,000	300	480	25
CsI(Tl)	54	1.86	4.5	55,000	1000	560	5
CsI, T=300 K	54	1.86	4.5	5,000	15	310	5
CsI, T=100 K	54	1.86	4.5	100,000	800	350	5

Pure CsI has been adopted in particle physics for experiments set in high-rate and high-radiation environments, given its fast decay time and radiation hardness [4]. At first glance, its low light yield (5 photons/keV), UV emission, average radiation length, and density make it a sub-optimal candidate for PET.

However, cooling pure CsI to cryogenic temperatures, approximately 100 K, causes its light emission spectrum to shift toward the near ultraviolet range, moving from around 310 nm to 340 nm [5]. This process also increases its light yield by over tenfold to about 100 photons/keV [6, 7, 8].

Unfortunately, its decay time increases also, approximately by a factor of 50. In a PET scanner, the coincidence time resolution (CTR) between pairs of crystals is determined by the material’s emission time profile and amount of light detected. An approximate figure of merit is given by [9]:

$$\mathrm{CTR}\propto\sqrt{\frac{\tau}{N_{\mathrm{ph}}}},$$ (1)

where $\tau$ is the decay time constant and $N_{\mathrm{ph}}$ is the number of detected photons.

Using eq. (1) to compare the CTR for CsI operating at ambient temperature and at 100 K we obtain:

$$\mathrm{CTR}_{100~{}\mathrm{K}}\sim\sqrt{\frac{800/15}{100/5}}\cdot\mathrm{CTR}_{300~{}\mathrm{K}}\sim 1.6\cdot\mathrm{CTR}_{300~{}\mathrm{K}}.$$ (2)

Thus, timing should still be possible at 100 K, in spite of the long decay time.

In this paper, we demonstrate that such high light yield translates into an energy resolution better than 7% and a coincidence time resolution below 2 ns at 511 keV.

Figure 1: Schematic of the experimental apparatus. The crystals and photosensors are placed on a copper plate mounted inside a cryostat. The copper plate is connected to a cold finger immersed in liquid nitrogen. An infrared shield is placed on the copper plate to minimize radiative losses.

II Performance assessment of cryogenic CsI crystals

II-A Experimental setup

We set up a table-top experiment, illustrated Fig. 1, employing a pair of pure CsI crystals of dimensions $3\times 3\times 20$ mm³ provided by AMCRYS and polished on all sides. The crystals were wrapped with 3 layers of 80 $\mathrm{\SIUnitSymbolMicro m}$-thick polytetrafluoroethylene (PTFE) tape and mounted on a copper plate in contact with a liquid nitrogen dewar through a cold finger. A Na-22 radioactive source with an activity of 40 kBq was placed between the two crystals. The light was detected by a pair of $3\times 3$ mm² Hamamatsu S13360-3025CS MPPCs [10]. An aluminum infrared shield was placed above the copper plate to minimize radiative losses. The signals were acquired by a Tektronix MSO44 oscilloscope triggering in coincidence with a sampling rate of 3.125 Gs/s. Full waveforms were saved to disk with a trigger rate of approximately 5 Hz.

A PT100 temperature sensor and two heating strips were placed behind the copper plate and connected to a CN32PT-224-C24 Omega PID controller, which allowed to keep the setup at the desired temperature. The setup was placed in a stainless steel chamber connected to a ConFlat 6-way cross. One flange featured 4 BNC connectors (2 per MPPC) and another one a 25-pin electronic feedthrough, used for the connection of the heating strips and the PT100 sensor. The chamber was then evacuated with an Agilent TwisTorr 305 fs turbopump to reach a vacuum level of approximately $10^{-4}$ mbar.

II-B Light yield and energy resolution

The charge spectra of the MPPCs signals were obtained by integrating the full waveforms acquired by the oscilloscope. In order to convert the integral to the equivalent number of photoelectrons, the charge spectra were normalized by the single photoelectron charge peak, obtained by amplifying the dark counts of the MPPC with a custom amplifier with a voltage gain of 900.

The intrinsic light yield of the crystal is obtained by the following formula:

$$\mathrm{LY}(\mathrm{photons/MeV})=\frac{N_{\mathrm{pe}}/\mathrm{MeV}}{\mathrm{PDE}\times\epsilon_{\mathrm{det}}},$$ (3)

where $N_{\mathrm{pe}}$ is the number of photoelectrons, $\mathrm{PDE}$ is the photon detection efficiency of the MPPC, and $\epsilon_{\mathrm{det}}$ is the detection efficiency of the setup. The weighted average PDE for the room temperature CsI emission spectrum is 12%, compared with the 17% at cryogenic temperature (see Fig. 2).

Figure 2: CsI scintillation spectrum at room and cryogenic temperatures [5] (in black, left axis), compared with the photon detection efficiency of the Hamamatsu S13360-3025CS MPPC [10] (in red, right axis).

The detection efficiency $\epsilon_{\mathrm{det}}$ has been estimated with a Geant4 [11] Monte Carlo simulation using the LUT optical model [12] at 58%. This relatively low efficiency is caused by the narrow form factor of the crystals (almost 7 times longer than wide), since the scintillation photons are reflected inside the crystals several times before reaching the photosensor. Thus, even if the PTFE reflectivity is approximately 95% [13], the probability for a scintillation photon of being absorbed before reaching the MPPC is not negligible.

A room temperature measurement was performed first, obtaining a light yield of $4500\pm 500$ photons/MeV and $4300\pm 500$ photons/MeV for the two crystals, respectively. The energy resolution was measured to be approximately 35% FWHM at 511 keV for both.

After filling the dewar with liquid nitrogen, the system reached an equilibrium temperature of 104 K after approximately 5 hours.

The voltage breakdown of MPPCs is proportional to the temperature with a slope of approximately 54 mV/K [10]. At T=104 K, we adjusted the bias voltage by reducing it by 10.3 V to compensate for this effect. We then repeated the single photon measurement. The integration window was increased to 5.5 $\mathrm{\SIUnitSymbolMicro s}$, due to the longer decay time of CsI at low temperatures (see Sec. II-C). The MPPC dark count rate is halved every 5 K [14], becoming negligible at T=104 K. Under these low-temperature conditions, the number of photoelectrons ($N_{\mathrm{pe}}$) increased by over an order of magnitude. This corresponds to light yields of $91,000\pm 2,000$ photons/MeV and $87,000\pm 2,000$ photons/MeV for the respective crystals.

This significant enhancement in light output translates into an improved energy resolution for the crystals, with values of 7.07% and 6.30%, as illustrated in Fig. 3.

Figure 3: Na-22 charge coincidence spectra of the two CsI crystals operating at T=104 K. The 511 keV photopeak has been fitted with a Gaussian distribution (solid and dashed lines).

The setup was gradually warmed with the PID controller. The light yield and the single photoelectron peak for the two MPPCs were measured at various temperatures during this process. Each run lasted approximately 30 minutes and the MPPC bias voltage was adjusted accordingly for each temperature.

Fig. 4 shows the non-linear increase in light yield as the temperature decreases, with the yield reaching a plateau at approximately 110 K, in good agreement with the existing literature [5]. The approximate 20-fold increase between room temperature and T=104 K is mainly due to the larger amount of scintillation light being emitted. In addition, the light emission at cryogenic temperature is shifted towards larger wavelengths, where the quantum efficiency of the photosensor is higher (see Fig. 2). This effect contributes to an additional 30% increase in the amount of light detected.

Figure 4: Light yield of pure CsI crystals as a function of temperature (black dots), compared with the existing literature [5] (white squares). The heating strips were able to warm the setup up to 180 K, thus the absence of measurements between this value and the room temperature.

II-C Emission time profile and coincidence time resolution

Figure 5: An example of two waveforms acquired by triggering in coincidence, with an equivalent energy near the 511 keV photopeak. The dotted lines correspond to the threshold used for the CTR measurement.

(a)

(b)

Figure 6: At cryogenic temperatures, pure CsI exhibits a time emission profile dominated by a single component of approximately 0.81 $\mathrm{\SIUnitSymbolMicro s}$. At higher temperatures two further components appear. (a) Average waveforms at three different temperatures fitted with one (dotted line), two (dashed line) or three (solid line) exponential functions. (b) Fast (white squares), medium (black circles) and slow (white triangles) decay times as a function of temperature.

The CTR for events near the 511 keV photopeak was measured in our setup using a baseline-corrected leading edge time pick-off method [15]. The threshold was set at 5 times the electronic noise, defined as the standard deviation of the voltage values in a region with no signal. The events were selected by requiring a charge equivalent to the energy window $511~{}\mathrm{keV}-\mathrm{FWHM}/2<E<511~{}\mathrm{keV}+\mathrm{FWHM}/2$. An example of a triggered event with an energy near the photopeak is shown in Fig. 5.

At room temperature, a CTR of $1.31\pm 0.06$ ns was achieved. The time emission profile is dominated by a fast component of approximately 15 ns, with two further components of approximately 50 ns and 2 $\mathrm{\SIUnitSymbolMicro s}$ contributing to less than 10% of the total light yield.

(a)

(b)

Figure 7: The coincidence time resolution increases with decreasing temperature, due to the larger decay time of CsI. The effect is partially compensated by the larger light yield. (a) Gaussian fit of the time difference at T=104 K. (b) CTR as a function of the temperature.

Cryogenic CsI, instead, is a relatively slow emitter, with a single decay time component of approximately 800 ns at 104 K. At this temperature, the measured CTR was $1.84\pm 0.07$ ns, which is qualitatively consistent with the value extrapolated from the room temperature measurement using eq. (1). Thus, the effect of the increase in decay time is partially compensated by an increase in the number of detected photons, mitigating the impact on the CTR. The CTR as a function of temperature and the time difference distribution at 104 K are shown in Fig. 7.

III Summary and future prospects

We have measured the energy resolution and coincidence time resolution for a pair of pure CsI crystals at cryogenic temperatures read out by solid state photosensors (Hamamatsu MPPCs). At T=104 K, an energy resolution better than 7% and a coincidence time resolution lower than 2 ns have been achieved. The measurements have been repeated at different temperatures, with results in good agreement with the existing literature. The performances of the material prove its potential as a candidate for affordable TB-PET scanners.