Quantitative imaging of ¹²⁴I and ⁸⁶Y with PET

Abstract

The quantitative accuracy and image quality of positron emission tomography (PET) measurements with ¹²⁴I and ⁸⁶Y is affected by the prompt emission of gamma radiation and positrons in their decays, as well as the higher energy of the emitted positrons compared to those emitted by ¹⁸F. PET scanners cannot distinguish between true coincidences, involving two 511-keV annihilation photons, and coincidences involving one annihilation photon and a prompt gamma, if the energy of this prompt gamma is within the energy window of the scanner. The current review deals with a number of aspects of the challenge this poses for quantitative PET imaging. First, the effect of prompt gamma coincidences on quantitative accuracy of PET images is discussed and a number of suggested corrections are described. Then, the effect of prompt gamma coincidences and the increased singles count rates due to gamma radiation on the count rate performance of PET is addressed, as well as possible improvements based on modification of the scanner’s energy windows. Finally, the effect of positron energy on spatial resolution and recovery is assessed. The methods presented in this overview aim to overcome the challenges associated with the decay characteristics of ¹²⁴I and ⁸⁶Y. Careful application of the presented correction methods can allow for quantitatively accurate images with improved image contrast.

Introduction

Generally positron emission tomography (PET) is considered as a quantitative imaging modality which is able to measure the radioactivity concentration within the patient not just as count rates, but in absolute terms with the unit of becquerel per millilitre. This quantitative measurement is a prerequisite for calculating the radiation dose administered to target and healthy tissues in radionuclide therapy. The quantitative property of PET is based on positron emission and the subsequent annihilation process with the emission of a pair of photons in opposite directions with an energy of 511 keV each. Although the coincident measurement of a pair of annihilation photons by a ring of detectors surrounding the patient is disturbed by random and scattered (Fig. 2) coincidences as well as by events lost due to tissue attenuation, the related errors can be corrected sufficiently so that the final quantitation error is in the range of a few per cent. While this situation holds for the typical positron emitters ¹⁸F or ¹¹C, it is no longer true for isotopes such as ¹²⁴I and ⁸⁶Y. These radionuclides do not only emit positrons, but also additional gamma radiation which may disturb quantitative PET imaging and image quality in a number of ways. Figure 1 shows simplified decay schemes of ¹²⁴I, ⁸⁶Y and, for comparison, ⁸⁹Zr and ¹⁸F. The decay schemes of the nonstandard positron emitters include gamma radiation of different energies and abundances. This gamma radiation, also called prompt or cascade gamma radiation, is often emitted essentially simultaneously with positrons. The amount and the energy of the prompt gamma radiation is different for the various nonstandard positron emitters.

Fig. 1

Simplified decay schemes of ¹²⁴I, ⁸⁶Y, ⁸⁹Zr and, for comparison, the ‘standard’ PET isotope ¹⁸F. Only radiation with abundance >5% is shown. Energy shown for positrons is maximum energy. Data based on [39]

In the decay of ¹²⁴I, with a positron abundance of approximately 23%, about 50% of all positrons (β⁺ ₁ in Fig. 1) are emitted simultaneously with a 603-keV gamma photon (γ₁ in Fig. 1). In the case of ⁸⁶Y, on average about three photons are emitted per decay, compared to a positron abundance of 32%. All ⁸⁶Y positrons are emitted simultaneously with either a 1,077- or a 1,854-keV gamma photon (γ₁ and γ₆), and 628- and 443-keV photons are emitted in 33 and 17% of all decays, respectively. Furthermore, the energy of the positrons emitted by ¹²⁴I and ⁸⁶Y is higher than the positron energy of ¹⁸F. This affects the spatial resolution of the PET images leading to decreased recovery and, consequently, underestimation of radioactivity concentrations in small structures [1].

Apart from ¹²⁴I and ⁸⁶Y as obvious analogues of therapeutic radionuclides, ⁸⁹Zr (half-life 78.4 h) is included here since it has been suggested as an analogue for ⁹⁰Y when labelling monoclonal antibodies [2]. ⁸⁹Zr has much preferred imaging properties compared to ⁸⁶Y. The aim of this chapter is to discuss the quantitation issues related to the use of ¹²⁴I and ⁸⁶Y and, additionally, ⁸⁹Zr with PET, as well as possible corrections and improvements.

Prompt or cascade gamma coincidences

Prompt or (cascade) gamma coincidences occur as coincidences of simultaneously emitted gamma photons with energies accepted within the energy discrimination window (e.g. 400–650 keV) with each other or with annihilation photons and cannot be distinguished from true coincidences, involving two annihilation photons. Detection of these essentially true coincidences therefore introduces a bias in the images which is not corrected for by the standard PET corrections [3–6] (Fig. 2). Since the directions of the gamma photons and the annihilation photons are not correlated, the gamma coincidences are distributed nearly uniformly within the PET field of view (FOV) causing a primarily flat background in the sinograms, as indicated in Fig. 3, and in the reconstructed images. This bias also results in degraded image contrast [4, 7].

Fig. 2

Degrading effects in PET, from left to right random coincidences, scattered radiation and prompt gamma coincidences where one of the annihilation photons is detected in coincidence with a prompt gamma photon. Reprinted from [4] with permission from IOP Publishing

Fig. 3

Normalized line source profiles derived from sinograms of ¹²⁴I or ¹⁸F line sources inside a cylindrical phantom filled with water. Reprinted from [10] with permission from Elsevier

In the case of ¹²⁴I, β⁺ ₁ is always emitted simultaneously with a 603-keV photon whereas β⁺ ₂ emission results directly in the ground state of ¹²⁴Te, which means that 52% of the positrons is emitted simultaneously with a prompt gamma. For ⁸⁶Y, the main two positrons are always emitted simultaneously with at least two photons, and all other positrons are emitted simultaneously with at least one photon. Most of these photons have energies greater than 600 keV. Even if the primary energy of a prompt gamma is above the higher energy level of the energy discrimination window, the prompt gamma may be accepted after being scattered within the patient or septa and having lost part of its energy. For both isotopes, electron capture decays lead to multiple gamma photons emitted simultaneously, which might cause a so-called multiple coincidence. In addition, a multiple coincidence is recorded if a true coincidence is detected simultaneously with a prompt gamma. The probability of multiple coincidences is rather low, but increases with a larger spatial angle of the PET detectors. Such events are discarded in most PET scanners.

For ⁸⁹Zr, on the other hand, prompt gamma coincidences do not occur since the metastable 0.91 MeV level of ⁸⁹Zr has a half-life of 14 s [8], as shown in Fig. 1.

As illustrated in Fig. 3, which compares the background caused by ¹²⁴I with that of ¹⁸F, the background is greater in 3-D PET than in 2-D where the septa limit the acceptance angle for photons not being within a plane perpendicular to the scanner’s axis. Therefore, random and scattered coincidences as well as gamma coincidences are decreased. On the other hand, simulation studies have shown that the principle advantage of 2-D imaging is to some extent counterbalanced by an increase in the relative effect of prompt gamma radiation due to down-scatter of high energy photons in the septa [9].

Earlier generation PET scanners, such as the Scanditronix PC4096 WB (Scanditronix, Uppsala, Sweden), had very long and thick septa so that the recorded rate of gamma coincidences became very low despite possible down-scatter. The advantage of this kind of 2-D PET became obvious by the papers of Pentlow et al., Herzog et al. and Lubberink et al. [4, 10–13], and it can be concluded that early studies using the PC4096 WB scanner for quantitative imaging of nonstandard positron emitters provided valid results even without any corrections for prompt gamma coincidences.

The effect on quantitation due to the background caused by the gamma coincidences depends on the specific nonstandard positron emitter and on the specific tissue or target to be examined. In the case of ¹²⁴I and imaging of thyroid cancer, the radioactivity distribution is limited to a few foci, whereas the backgound is distributed across the entire image and thus contributes little to the activity concentration in a lesion. This situation is similar to that displayed in Fig. 3, where the ratio of counts measured at the maximum of a ¹²⁴I point source in water and of the background counts is more than 10:1. For labelled antibodies, however, the situation is different. Here, a lot of the radioactivity remains in the blood or is distributed to major organs. This effect may be considerable. Figure 4 illustrates the background found in a cold water rod which is located in a phantom homogeneously filled with ¹⁸F, ¹²⁴I or ⁸⁶Y. The radioactivity concentration measured at the cold water rod relative to the surrounding radioactivity is 2% for ¹⁸F, 14% for ¹²⁴I and 56% for ⁸⁶Y. Figure 4 also documents that considerably greater errors are found in bone regions which are simulated by the Teflon rod of the phantom. In this case, the radioactivity concentration measured at the Teflon rod relative to the surrounding radioactivity is 13% for ¹⁸F, 44% for ¹²⁴I and 147% for ⁸⁶Y.

Fig. 4

Images of a NEMA 1994 phantom with cold Teflon (top), water (left) and air (right) inserts. The measurements were done with an ECAT Exact HR+ (Siemens/CTI, Knoxville, TN, USA) PET scanner in 3-D acquisition mode. There is a clear bias in the Teflon and water inserts for ⁸⁶Y and, to a lesser extent, for ¹²⁴I. Reprinted with permission from [16], ©2008 Edizione Minerva Medica

Correction

Several methods have been suggested to correct for the bias caused by prompt gamma coincidences. Firstly, subtraction of a uniform background [4, 14] or a linear background fitted to the sinogram data outside the object [4, 15] (shown for ⁷⁶Br in Fig. 5) has been suggested. In 2-D mode, the assumption can be made that the outermost bins in the sinogram contain only prompt gamma coincidences, since scatter is negligible in these bins. Depending on the size of the imaged object, however, this assumption may be wrong, and especially in 3-D acquisition mode there is a considerable amount of scatter in the edge bins. Herzog et al. [16] found that 75% of the background had to be subtracted for ⁸⁶Y to obtain similar residual correction errors in a cylindrical phantom with cold inserts as for ¹⁸F. Buchholz et al. compared phantom measurements with and without a uniform background subtraction for different scanners [17]. An example of the effect of subtraction of a uniform background on ⁸⁶Y patient data is shown in Fig. 6. Although this subtraction gives a good first approximation for cylindrical objects in the centre of the FOV, the assumption that the background caused by prompt gamma coincidences is linear is generally not valid, as clearly shown in Figs. 5 and 8. Therefore, Kull et al. [15] used a second-order series expansion to describe the shape of the background for 2-D scans with ⁸⁶Y, where they determined the second-order term using a measurement with a body phantom and the linear portion of the background by fitting to the sinogram tails of the individual patient, and also included a recalibration of the scanner based on the trues to singles ratio.

Fig. 5

Projections of a torso phantom filled with ⁷⁶Br and ¹⁸F, along its short axis, as measured on an ECAT Exact HR+ scanner in 3-D mode. The decay scheme of ⁷⁶Br resembles that of ⁸⁶Y. The difference between both projections shows the contribution of gamma coincidences. The red line indicates the linear fit used to correct for prompt gamma coincidences, corresponding to 80% of the counts in the outermost bins. Adapted from [23]

Fig. 6

⁸⁶Y-DOTATOC images as measured with the ECAT Exact in 2-D mode, without (top row) and with (bottom row) subtraction of a uniform sinogram background. The corrected image shows a lower background in the liver and a considerable reduction of radioactivity concentration in the spine. Reprinted from Fig. 4 in [17] with kind permission from Springer Science+Business Media

A convolution subtraction algorithm based on the method suggested by Bergström et al. in the early 1980s for correction for scattered radiation [18] was described by Beattie et al. [6]. This method does take patient-specific variations into account, but has only been described for 2-D acquisitions. Walrand et al. [5] used a patient-dependent correction method based on sinogram tail fitting using an ⁸⁶Y point spread function library, showing promising results. A geometrical correction was suggested by Schweizer and von Busch [19], calculating the contribution of prompt gamma coincidences of a number of source points to each line of response.

The single-scatter simulation scatter correction applied on all last generation PET(/CT) scanners [20] usually includes a scaling to match the estimated scatter contribution to the actual events measured just outside the body. If this scaling includes both a multiplicative as well as an additive factor, it implicitly performs a crude correction for a uniform bias caused by prompt gamma coincidences as well. This has been shown by Surti et al. for ¹²⁴I on a Gemini PET/CT scanner (Philips Healthcare, Cleveland, OH, USA) [21, 22] (Fig. 7). As Fig. 7 also shows, using only a multiplicative factor leads to overcorrection in the centre of the image.

Fig. 7

Images of a 20-cm diameter phantom containing one 3-cm, one 1.5-cm and three 1-cm diameter spheres. From left to right: ¹⁸F, ¹²⁴I without offset correction and ¹²⁴I with offset correction in the scaling of the scatter estimate. Reprinted with permission from [22], ©2009 IEEE

Finally, it has been shown that the distribution of prompt gamma coincidences matches the distribution of random coincidences rather well [23] (Fig. 8). Therefore, a correction method involving subtraction of a scaled randoms sinogram could be an accurate correction for prompt gamma coincidences [23], possibly incorporated into the single-scatter simulation. This method has been realized by Watson et al. and Hayden et al. [24, 25] for cardiac studies with the nonstandard positron emitter ⁸²Rb, which emits a 777-keV prompt gamma together with positrons in only 14% of decays and has, so far, not been implemented for ¹²⁴I or ⁸⁶Y.

Fig. 8

Projections of a 10-cm off-centre cylindrical phantom with cold inserts filled with ⁷⁶Br (solid black line) and ¹⁸F (solid grey line) as well as the prompt gamma contribution for ⁷⁶Br (dashed line) as measured with an ECAT Exact HR+ in 3-D mode (a) and corresponding delayed coincidence projections (b). Shapes of delayed coincidence and prompt gamma coincidence projections are approximately similar. Reprinted from [23]

Count rate performance

The increased singles rate due to gamma radiation leads to increased random coincidence rates and, consequently, reduced noise equivalent count (NEC) rates [26]. The increased random rate can be accurately corrected for using the standard delayed window method, but correction for a larger random fraction increases image noise. Figure 9 shows NEC rates for ¹²⁴I and ¹¹C as measured using a cylindrical phantom. To calculate these NEC rates, the scatter fraction for ¹¹C was computed according to a method described by de Jong et al. [27]. The total (scatter + prompt gamma) background was measured in a similar way for ¹²⁴I as for ¹¹C, and the prompt gamma contribution to this background was estimated assuming identical scatter fractions for ¹¹C and ¹²⁴I. The NEC rate is then described by:

$$ {\hbox{NEC}} = \frac{{{{\text{T}}^2}}}{{{\hbox{T}} + {\hbox{S}} + {\hbox{G}} + {\hbox{2D}}}} $$

(1)

Here, T are the true coincidences, S and G are the number of scattered and prompt gamma coincidences, respectively, and D is the number of delayed coincidences.

Fig. 9

NEC rates for a 20-cm diameter phantom filled with ¹²⁴I (solid line) and ¹¹C (dashed line) for the Gemini TF PET/CT scanner. Data for ¹¹C were based on Surti et al. [40] with radioactivity concentrations divided by 0.225 to account for the difference in positron abundance between ¹²⁴I and ¹¹C [31]

For ¹²⁴I, and even more so for ⁸⁶Y, the fraction of detected photons with energy outside the scanner’s energy window increases considerably compared to positron-only emitters. Rejection of photons outside the energy window does contribute to dead time, but these photons are, on most scanners, not counted in the singles rate. Since the dead time correction is usually implemented as a function of singles rate, it may become inaccurate [3, 15, 28, 29]. This effect has previously been shown for ⁸⁶Y [15], ¹²⁴I [28] as well as for ⁷⁶Br [29], which has a decay scheme somewhere in between that of ¹²⁴I and ⁸⁶Y in terms of the number of emitted prompt gamma photons.

Energy window

One option to improve image quality for isotopes emitting high-energy gamma radiation besides positrons may be the use of a narrower energy window, which reduces random coincidence rates involving higher-energy photons, such as the 603-keV photon emitted by ¹²⁴I [7, 30] (Figs. 10 and 11). Furthermore, the amount of scattered higher-energy gamma photons which may coincide with annihilation photons will be decreased. Gregory et al. found an increase in NEC rates of 48% for ¹²⁴I when changing the energy window from 409–665 keV to 455–588 keV on a Gemini Dual GS PET/CT scanner [28]. Due to its higher standard lower level discriminator, a more limited improvement was found for the Gemini TF scanner when changing the energy window from 440–665 keV to 440–560 keV [31]. In addition to a different energy window, a narrower coincidence window decreases random coincidence rates which is relatively of more importance for ¹²⁴I and ⁸⁶Y than for ¹⁸F. Use of a narrower energy window, however, may increase the inaccuracy of dead time corrections as described in the previous paragraph, and energy window-specific scanner normalization may be required.

Fig. 10

a NEC rates, which are a measure of the signal to noise ratio, of the Gemini TF-64 with ¹¹C using the standard 440–665 keV energy window (black), for ¹²⁴I using this same window (blue) and for ¹²⁴I using a narrow 440–560 keV window (red). Activity concentrations were normalized for positron abundance. b Improvement in recovery of ¹²⁴I using the narrower energy window [31]

Fig. 11

PET images of a patient with metastatic thyroid cancer at 24 h after administration of 37 MBq ¹²⁴I acquired on a Gemini TF-64 PET/CT scanner (a) 440–665 keV and (b) 440–560 keV energy window. The narrower energy window results in a 15% improvement in image contrast in the largest metastasis (arrow) due to the decreased image background [31]

Resolution and recovery

One of the parameters influencing the accuracy of quantitation is image resolution. The lower the image resolution is, the greater is the partial volume effect which compromises the analysis of small structures. As a first approximation, the image resolution of PET can be estimated by the following equation:

$$ {\hbox{r = 1}}{.25}\sqrt {{{{\left( {\frac{\text{d}}{{2}}} \right)}^2}{ + }{{\left( {{0}{.0022} \cdot {\hbox{D}}} \right)}^{{2}}}{ + }{{\hbox{R}}^2}{ + }{{\hbox{b}}^{{2}}}}} $$

(2)

where the image resolution r is expressed as full-width at half-maximum (FWHM), d is the detector width, D is the detector ring diameter and R is the effective positron range [32]. The factor b equals 0 for detectors individually coupled to photomultiplier tubes and 2 for a block detector design.

This equation takes into account the positron energy as an important factor influencing the image resolution. Both ⁸⁶Y and ¹²⁴I emit positrons with different energies with average energies ranging from 550 to 898 keV and from 686 to 973 keV, respectively [33]. For comparison, ¹⁸F emits positrons with mean energy of 250 keV, ⁸⁹Zr with 389 keV, ¹⁵O with 735 keV and ⁶⁸Ga (mainly) with 836 keV. Thus, the positron ranges given in the literature for ¹⁵O and ⁶⁸Ga (2.0 and 2.2 mm [34]) can be regarded as appropriate estimates for the mean positron ranges of ⁸⁶Y and ¹²⁴I, respectively. The mean positron range of ¹⁸F is 0.64 mm. The consequences of the greater positron ranges of ⁸⁶Y and ¹²⁴I can be estimated with Eq. 2 and are summarized in Table 1 for typical design parameters of human and small animal PET.

Table 1 Theoretical spatial resolution (mm)

Thus, for small animal PET the effect of high positron range is more severe than for human whole-body PET, both in the relative and absolute sense. Pentlow et al. evaluated the resolution of ¹²⁴I using phantoms containing hot spheres in a number of different scanners [11, 12], but did not report actual resolution values. Herzog et al. [10] measured an image resolution of 6.1 mm with ¹²⁴I and of 5.1 mm with ¹⁸F in a line source centrally located in an HR+ scanner and reconstructed with filtered backprojection and a Shepp filter of 2.5 mm. Vandenberghe [30] examined simulated line sources filled with ¹⁸F, ⁸⁶Y or ¹²⁴I and placed centrally in an Allegro PET scanner (Philips Healthcare, Cleveland, OH, USA). The line spread functions showed FWHM values of 4.8, 5.7 and 6.1 mm, respectively. Comparative resolution measurements were performed by González Trotter et al. [35] using line sources filled with either ¹⁸F or ¹²⁴I and placed in the centre of a Discovery LS PET/CT scanner (GE Healthcare, Milwaukee, WI, USA) operated in 2-D mode. After reconstruction with filtered backprojection using a Gaussian filter of 7 mm the measured resolution was 8.71 and 9.74 mm, respectively. Bading et al. [36] examined the degradation of image resolution, when ¹²⁴I was used comparatively with ¹⁸F in a microPET R4 small animal scanner (Concorde/Siemens, Knoxville, TN, USA). In the scanner’s centre the FWHM was 2.3 mm for ¹²⁴I and 1.9 mm for ¹⁸F. Using a Discovery STE PET/CT scanner Zhu and El Fakhri [37] studied the spatial resolution of ¹⁸F and ⁸⁶Y with a line source placed centrally in air and reported values of 5.56 and 6.06 mm. Gregory et al. reported a resolution degradation of 0.7 mm for ¹²⁴I compared to ¹⁸F for the Gemini Dual GS PET/CT scanner [28]. Although the numbers just summarized differ depending on the scanner and reconstruction method applied in the respective studies, the effect of the higher positron energies of ⁸⁶Y and ¹²⁴I is obvious: concordantly the image resolution is reported to be 0.5–1 mm inferior to that when using ¹⁸F. Considering the small increase in positron energy between ¹⁸F and ⁸⁹Zr, only a minor degradation in resolution of about 0.1 mm is expected for ⁸⁹Zr.

The effect of this degradation of image resolution for ¹²⁴I on recovery, that is, the ability of the PET scanner to quantify the radioactivity concentration in small structures, was measured extensively by Jentzen et al. [38] (Fig. 12). Recovery for ¹²⁴I was considerably worse than for ¹²⁴I, even for spheres as large as 37 mm in diameter. This has to be accounted for when quantifying tumour uptake of ¹²⁴I for dose estimations in thyroid cancer therapy, and the authors conclude that recovery correction is mandatory for ¹²⁴I, even for large structures.

Fig. 12

Recovery for ¹²⁴I and ¹⁸F as measured with an ECAT Exact HR+ scanner in 3-D mode. Reprinted from Fig. 2 in [38] with kind permission from Springer Science+Business Media

Summary

The decay characteristics of ¹²⁴I and ⁸⁶Y set a challenge to quantitative imaging of these isotopes. The methods presented in this overview aim to overcome this as well as to improve image quality, and careful application of presented correction methods can allow for quantitatively accurate images with improved image contrast.