by Dr. Vittori Franco
Let’s start with the explanation of the HIRESPET Collaboration work titled "Crystals and light collection in Nuclear Medicine" (transparency 1). As suggested by the title, I’m going to talk exclusively about scintillating crystals used in Nuclear Medicine applications. After this short introduction, I’ll just give you a short literature overview of the works concerned with this issue and published in the last two years. We’ll see that a great attention has been paid to the development of gamma detectors dedicated to scintimammography applications. I’ll point out the main characteristics of this new detectors and then we’ll give a look at the recent clinical results obtained with the actual HIRESPET detector for scintimammography. In order to improve the actual detector performance, we have developed a Montecarlo code with the purpose of studying light collection efficiency as a function of crystal geometry and surfaces treatment. First, we’ll discuss some Montecarlo results and then the expected performance of the improved detector. Finally, we’ll jot down the conclusions.
First of all, let’s have a look at the main characteristics of the most used scintillating crystals in Nuclear Medicine (transparency 2) that is sodium iodide thallium doped, caesium iodide thallium and sodium doped, yttrium aluminium perovskite caesium doped and BGO. Sodium iodide thallium doped is used on the Anger camera which is the most widely available instrument in Nuclear Medicine capable of performing SPE analysis even if it suffers from poor spatial resolution. BGO instead is mainly used for PET applications also in multipillar structure thanks to its short attenuation length at 511 keV. Recently, particular attention has been paid to YAP :Ce and CsI(Tl) in multipillar configuration for SPE detectors. YAP :Ce has high hardness, is not hygroscopic, has an emission peak that well matches with the quantum response of bialkali photocathodes and is very fast. Although CsI(Tl) is quite soft, it has a relatively short attenuation length at 140 keV and has a light yield comparable to that of YAP :Ce when coupled to a PMT. Because its emission is peaked at 550 nm, its light yield increases considerably when coupled to a photodiode. CsI(Na) instead shows a light yield of 85 % when coupled to a PMT.
Now, let’s have a look at a short literature summary concerning with the use of scintillating crystals in Nuclear Medicine applications. Moses et al. (transparency 3) developed, for a scintimammography application, a detector made up of a CsI(Tl) scintillating matrix coupled to a photodiodes array. This solution is very attractive because the detector shows an excellent energy resolution of about 11 % FWHM at 140 keV, but the main drawback is the small FOV. In order to widen the FOV it’s necessary to increase considerably the number of electronics channels as well as the total cost of the detector. Levin et al., for a scintimammography application, demonstrated that it’s possible to arrange a NaI(Tl) planar crystal coupled to a PSPMT with good results for both energy and spatial resolution, but the FOV is smaller than the crystal size due to the wider scintillating light distribution obtainable with a planar crystal. Majewski et al., for a scintimammography application, studied the behaviour of YAP :Ce, CsI(Na) and NaI(Tl) multipillar matrices coupled to a R3292 PSPMT obtaining a good spatial resolution. However the energy resolution seems quite poor.
Tornai et al. (transparency 4) studied the use of mercuric iodide photodetectors coupled to a CsI(Tl) scintillating matrix for a scintimammography application. They obtained an excellent 10-12 % energy resolution at 122 keV and a good spatial resolution. Also in this case the main drawback is represented by the small FOV. HIRESPET Collaboration, for a scintimammography application, developed a detector with large FOV by exploiting a CsI(Tl) matrix coupled to a R3292 PSPMT obtaining a good spatial resolution equal to 1.8 mm FWHM even if this represents the spatial response of a single pillar. In this condition the true spatial resolution is determined by the pillars pitch equal to 2.3 mm. We obtained also a nearly optimal energy resolution of 20 % FWHM at 122 keV over the entire FOV. This detector is now under clinical experimentation at the Institute of tumours Regina Elena of Rome and I’ll show you some clinical results in a couple of minutes. Similar results have been obtained by Pani et al.
HIRESPET Collaboration has moreover developed a small FOV detector (transparency 5) particularly suitable for radiopharmaceuticals research which is now under clinical experimentation at University of Padua. The imager is characterised by an excellent spatial resolution of 1.2 mm and 0.7 or 0.5 mm FWHM depending on the crystal matrix used, for standard resistive chain readout and single wires readout respectively. In the end, we should mentioned the results of a YAP :Ce matrix made up of 0.3 mm x 0.3 mm size pillars coupled to the ISPA tube, whose last results were presented yesterday by Dr. Puertolas. The detector showed an unbelievable spatial resolution and a very good energy resolution even in the light of the low light yield of a pillar with a 0.3x0.3 square millimetre area and a 10 mm thickness.
In this short overview we have seen that great efforts have been made to develop a compact and reliable detector for scintimammography. In order to understand the importance of such a research, I would like to show you the comparison between the characteristics of X-ray mammography and scintimammography with Tc99m the most widely diffused monoclonal antibody for scintimammography diagnosis, which has been recently accepted by the Foods and Drugs Administration of USA (transparency 6). X-ray mammography represents the most widely used diagnostic technique for detecting breast cancer thanks to its good spatial resolution of about one hundred microns and low cost. It has a high sensitivity for the detection of breast cancer, which decreases significantly for radiologically dense breast, but it suffers from a low specificity for distinguishing malignant from benign tumours. This means that many biopsies are needed in order to evaluate the true lesions nature. On the other hand, scintimammography has a higher specificity but it has a sensitivity that decreases to about 50 % when considering lesions smaller than 1 cm. So, the main objective of scintimammography is to improve the sensitivity for lesions smaller than 1 cm in order to detect tumours in earlier growth stages as well as reducing the number of biopsies. The solution is to develop compact gamma cameras designed specifically for breast imaging and characterised by an optimal geometry and an improved spatial resolution.
In few words, the main characteristics of the new gamma cameras dedicated to scintimammography are (transparency 7): multipillar structure of the scintillating crystal, light detection performed by PSPMTs or photodiodes arrays and a small detection head. The advantages over the traditional Anger camera are : the multipillar structure of scintillating crystals provides improved intrinsic spatial resolution ; the small gamma camera size allows shorter imaging distance, thus improving collimator resolution ; the improved total spatial resolution allows to observe tumours in earlier growth stages (lesser than 1 cm) ; the compact design allows a great variety of viewing angles ; thanks to a similar detectors geometry, it is possible to perform a direct comparison between X-ray and scintigraphy images, thus providing the basis of a multi-modality system for breast imaging ; and finally low cost.
In this transparency (transparency 8), we analyse what optimal geometry means for a scintimammographic detector. First of all, I would like to apologise to you for this ugly sketch. It should represent the chest and the breast. As you can see, the measure conditions are completely similar to those of X-ray mammography. By using this geometry it’s possible to reduce the strong background coming from the heart. As a matter of fact the heart exhibits a strong uptake of Sestamibi. I would like to recall your attention on the breast compressor that is used in the same way of X-ray mammography. It allows to reduce the lesion distance to 2.5 cm at maximum.
In this transparency (transparency 9), a sketch of the HIRESPET detector dedicated to scintimammography is represented. Thanks to its optimal geometry it is possible to perform acquisition from several viewing angles. This detector is under clinical experimentation at the Institute of tumours Regina Elena of Rome. 29 patients with breast masses underwent scintimmammography with this detector and a state of art Anger camera using 740 MBq of Tc99m-sestamibi.
This transparency (transparency 10) shows the comparison between the Anger camera and the new detector image relative to an interesting case. It’s clearly visible that the Anger camera detected two not well distinguished lesions whereas the SPEM detected and clearly well separated the two lesions. Moreover it shows another smaller nodule beyond the larger one not detected by the Anger camera.
This is a very impressive case (transparency 11). The case appeared negative in both Anger Camera and X-ray mammography investigation while SPEM was able to detect a 0.4 cm size lesion, in good agreement with MRI and echography images. Probably, in the Anger image the lesion is hidden by the strong heart gamma emission.
Although these results are really interesting, we think that it’s necessary to improve further the detector performance. So, let’s have a look at what we think are the optimum expected performance of a detector dedicated to scintimammography (transparency 12). It should have an energy resolution lesser than or equal to 15 % in order to reduce the Compton background coming mainly from the heart, a spatial resolution of 1 or 2 mm at maximum in order to detect small lesions, a FOV not greater than 15 cm x 15 cm, a low cost and a short acquisition time. I would say that the HIRESPET detector characteristics are nearly optimal but we are implementing further improvements. In order to improve the detector characteristics it is necessary to increase the light output of the scintillating crystal and to design a large FOV optical photons detector also in the light of the poor performance shown by larger PSPMT’s. A MonteCarlo code has been developed in order to study the light collection efficiency as a function of the pillars dimensions and the surfaces treatment. As regards the optical photons detector, the solution ISPA tube is being implemented.
Before showing the simulation results, let’s give a look at this simple light collection analysis reported in this transparency (transparency 13). This is the crystal pillar and this is the PSPMT window. You well know that if the photon incidence angle, and then the angle between the photon trajectory and the main pillar axis, is smaller than the critical angle (defined by the refraction indexes of the crystal and the window material), the photon is refracted into the PSPMT window (with a probability given by the Fresnel formulas) and hits the photocathode. On the contrary, if the incidence angle is greater than the critical angle, the photon is reflected, trapped into the pillar and, after many internal reflections, absorbed into the bulk or onto the surfaces. If we hypothesise perfectly reflective surfaces and an infinite absorption length, it is possible to calculate the maximum theoretical collection efficiency with this formula. It is equal to 0.37 for YAP, 0.46 for CsI, 0.41 for NaI and 0.28 for BGO.
However, this is an optimistic estimation because we disregarded the absorption effect into the crystal bulk and onto the pillars surfaces. YAP :Ce pillars are covered by a special 10 microns thick reflective layer and a recent work showed (transparency 14) the optical properties of both the reflective layer and the crystal. The reflectance of the layer ranges between 90 % and 100 % depending on the incidence angle and the absorption length of the YAP :Ce crystal at 370 nm is 14 cm.
We put these experimental results into the simulations obtaining the results showed in this transparency (transparency 15). This graphics represents the dependence of the photon collection efficiency on the length and the cross-sectional area of the pillars. The black lines represent the montecarlo results for a 2x2, 1x1, 0.6x0.6 and 0.3x0.3 square millimetre size pillars with thickness ranging from 1 mm to 30 mm. The red and the blue dots represent the experimental results of light collection efficiency obtained by YAP :Ce pillars with 0.6x0.6 and 1x1 square millimetres area respectively. There is a good agreement between the experimental and the simulation results. In the graphic upper part, it’s visible the theoretical limit equal to 0.37 for YAP-glass interface due to the critical angle.
We have just seen the strong dependence of the collection efficiency on the pillars length. This effect is mainly due to the absorption effect on the lateral surface as it is possible to see in this transparency (transparency 16) where the dependence on pillar length of the percentage of detected and absorbed photons is reported. In spite of the high reflectance shown by the reflective layer used here, the high number of internal reflections makes increase considerably the number of photons absorbed onto the surfaces. As a reference, a 1x1x5 cube millimetres pillar shows a 25 % of photons detected, a 71 % and a 4 % of photons absorbed onto the surfaces and into the crystal bulk respectively.
Our simulation showed that the highest collection efficiency of scintillating photons occurred when the distal surface from the PMT was of diffusive type and the sides of the pillars were of reflective type. In this transparency (transparency 17) we see the collection efficiency comparison between the latter condition, the blue lines, and the standard one, the red ones, that is with all the surfaces of reflective type. As a reference, a 2x2x5 mm3 YAP pillar shows a 28 % collection efficiency with top reflective surface and 41 % with diffusive top surface. It is worth noting that a CsI pillar with all diffusive surfaces shows a 79 % and a 65 % collection efficiency with 2x2x3 mm3 and 1x1x3 mm3 dimensions respectively. Regarding collection efficiency of YAP pillars as a function of their length, it has to be noted that the longer is the pillar the lesser is the difference in collection efficiency. Moreover, the smaller is the cross sectional area the shorter is the length where this difference disappears. For a pillar with a 0.3x0.3 square millimetre area the difference disappears at about a 7 mm length whereas for a pillar with a 2x2 square millimetre area it disappears for lengths greater than 30 mm.
This behaviour is well understood by seeing the incidence angle distribution of the detected photons relative to a 10 mm length pillar (transparency 18). The diffusive top surface of the pillar randomises the direction of photons promoting the direction of PMT. Nevertheless, this effect is not visible with a 0.3x0.3 pillar 10 mm length, because of the high number of internal reflections which the photons undergo, before arriving to the output surface. On the contrary, the positive effect given by the diffusive top surface is visible with a 2x2 square millimetre area pillar 10 mm length. Indeed, in this geometrical condition the number of photons that reach the output surface with an incidence angle smaller than 20 degrees, is considerably increased.
The same effect is visible when considering a 1x1 square millimetre area pillar with different lengths (transparency 19). Whereas a 20 mm length pillar doesn’t show any difference between reflective and diffusive top surface, the increase of detected photons is impressive for a 1 mm length pillar.
From this analysis it seems that the collection efficiency improvement provided by the diffusive top surface is really effective when using pillars with particular length by side ratios. In this transparency (transparency 20), we can see the collection efficiency as a function of the length by side ratio. It is worth noting that the diffusive top surface is useful for a h/a ratio lesser than 20 but is really effective when the h/a ratio is less than or equal to 5. Moreover, given a h/a ratio, the collection efficiency is almost constant. This is quite reasonable. Indeed, from simple geometrical considerations, it is possible to deduce that the number of hits on the lateral surfaces, and then the probability of being absorbed on them, is proportional to the h/a ratio, whereas the travel distance, and then the probability of being absorbed into the bulk, is proportional to the pillar length. Because, as already mentioned, the main photons absorption contribution is given by the lateral surfaces, this behaviour is justified.
The last parameter to be taken into account is the Full Energy Peak detection efficiency (transparency 21). Thanks to its higher interaction coefficient, CsI shows a better FEP efficiency and we have a value of about 60 % for a single 2x2x3 cube millimetre CsI pillar. As a reference, a 2x2x10 cube millimetre YAP :Ce pillar shows a FEP efficiency of about 48 %.
Now, let’s have a final look at the typical configuration of YAP :Ce and CsI(Tl) matrices (transparency 22). The YAP pillars have a special reflective layer only 10 microns thick, which allows a dead zone as low as 1 %. On the contrary, CsI needs a diffusive layer typically 0.3 mm thick which leads the dead zone to about 25 %.
Now, we can evaluate the improvements that are being implemented on the HIRESPET detector dedicated to scintimammography (transparency 23). The new detector should consist of a scintillating matrix coupled to an ISPA tube with a 15x15 cm2 FOV, a focusing system with a demagnification ratio of 5:1 and a detector chip composed by a diodes array and an electronics plane of 3x3 square centimetre area. Depending on the crystal matrix type coupled to the tube, we can evaluate, starting from the experimental results and the simulation ones, the following detector performance. With a YAP :Ce matrix made up of 1x1x5 cube millimetre pillars, the dead zone is very low, the energy resolution is 20 % and the spatial resolution is 1 millimetre. The main drawback is given by its poorer detection efficiency. With 2x2x5 cube millimetre YAP :Ce pillars, the energy resolution increases slightly whereas the spatial resolution is 2.0 millimetre. Using 2x2x3 cube millimetre CsI(Na) pillars instead, the dead zone is about 25 % due to the thicker diffusive layer used with CsI pillars, the detection efficiency is good, the energy resolution is an excellent 10 %, whereas the spatial resolution is 2.2 millimetre due to a thinner inter-pixel spacing as claimed by a CsI matrices manufacturer. With 1x1x3 cube millimetre pillars, the dead zone is 20-30 % depending on the inter-pixel spacing, the detection efficiency is 59 %, the energy resolution is about 12 % and the spatial resolution is 1.1-1.2 millimetre. The latter seems to be the best solution even if it shows a relatively high dead zone. However this drawback can be overcome using a parallel holes collimator with holes openings that exactly match the pillars sizes, and with a septa thickness which corresponds with the inter-pixel spacing, in order to maximise the signal for any given detector element of the array.
In conclusion (transparency 24), we can say that scintimammography represents a sensitive, specific, and non invasive method to define the nature of radiologically described breast masses and would be very useful as a complement to X-ray mammography in screening programs for breast cancer. In order to observe tumours in early growth stages it’s necessary to develop new detectors characterised by optimal geometry and good spatial and energy resolution. Recently, particular attention has been paid to the development of new detectors for scintimammography by exploiting new techniques like scintillating crystal arrays, PSPMT’s and solid state detectors technologies. HIRESPET Collaboration has developed a new large FOV detector which is now under clinical experimentation with impressive results. New technological improvements, such as an optimal pillars geometry and surfaces treatments together with the use of a large FOV ISPA tube as scintillating photons detector, are being implemented on the actual HIRESPET detector in order to achieve the optimum characteristics for an imager dedicated to scintimammography. The expected performance of the new HIRESPET detector are a FOV of 15x15 square centimetre, an energy resolution of about 12 % and a spatial resolution of about 1.2 mm.
That’s all. Thank you for your attention.
