Periodic Reporting for period 4 - 4D-PET (Innovative PET scanner for dynamic imaging)
Période du rapport: 2021-07-01 au 2023-07-31
The main objective of 4D-PET is to develop an innovative PET scanner based in a new detector concept that stores 3D position and time of every single gamma interaction with unprecedented resolution. The combination of scanner geometrical design and high timing resolution will enable developing a full sequence of all gamma-ray interactions inside the scanner, including Compton interactions, like in a 3D movie. 4D-PET fully exploits Time Of Flight (TOF) information to obtain a better image quality and to increase scanner sensitivity, through the inclusion in the image formation of all Compton events occurring inside the detector, which are always rejected in state-of-the-art PET scanners. The new PET design will radically improve state-of-the-art PET performance features, overcoming limitations of current PET technology and opening up new diagnostic venues and very valuable physiological information.
The 4D-PET scanner is characterized essentially by its ability to describe sequentially the development of every single gamma-ray interaction in the detector, including Compton interactions.
The new technology currently being developed is important for society because it will allow:
• Imaging Babies. Imaging babies with PET is important in many critical situations like cardiac arrest or brain injury due to its ability of imaging metabolic status and monitor treatment. However, PET is rarely used with children due to the high-required dose and to its insufficient spatial resolution to image the small structures of the baby body. 4D-PET project will enable imaging babies due to its high sensitivity and outstanding spatial resolution. We aim at a dose for babies of 100μCi, similar to what is routinely injected in mice and 100 times less of a dose in adults. Furthermore, such small dose will allow repeated examinations of the same patient to monitor the evolution to treatment therapy.
• Imaging small structures of the human brain. Sub millimetre resolution is absolutely needed for the visualization of several critical structures in the brain (substantia nigra and raphe nucleus) involved in mental disorders. Brain stem nuclei, substantia nigra and associated areas, as well as subregions of the thalamus have attracted considerable interest in schizophrenia research. A limitation is that the performance of current PET systems does not allow for detailed examination of the dopamine system in the thalamus and its functional implications.
• Imaging tumour heterogeneity. The standard PET/CT (and now also the new PET/MRI) scanners offer PET resolution on a 4-6mm scale and, therefore, cannot image small structures due to the poor resolution but also indirectly due to the decreased efficiency in detecting small features, as the result of the Partial Volume Effect (PVE). Excellent spatial resolution and image quality will allow the study of the geometrical structures of the tumour tissues with high precision. Furthermore, dynamical studies aiming at determining tumour uptake kinetics might also help in determining different components (heterogeneity) inside the tumour (1). Enlightening tumour composition and evolution will ease therapeutic decision, such as the kind of drug to be used in chemotherapy.
• Monitoring Dose in Proton Therapy. Proton therapy is a very useful radiotherapy technique for treating cancer because it optimizes the delivery of dose to the tumour while minimizing the dose in healthy tissues. This technique is especially relevant in children, where vital organs are still under development and might be at risk with conventional RT methods. One way to monitor proton dose delivery is by detecting the scarce gamma rays emitted by the nuclei of the human body after proton irradiation.
The new 4D-PET technology will be validated through the design and development of a preclinical prototype to examine the head of mice. The brain of a mouse has a size of only few millimetres and therefore to be able to visualize the detailed structure of its head and quantify, for instance, the amyloid deposit content represents a very challenging application.
The main performance objectives of this prototype are the following:
• Ultra-High Timing Resolution below 100ps.
• Spatial Resolution Determined by the Physical Limit (0,5mm approx.).
• High Sensitivity over 50% of the mouse brain.
(1) Koolen, Bas B.; … Benlloch J.; et al. Evaluating heterogeneity of primary tumor 18 F-FDG uptake in breast cancer with a dedicated breast PET (MAMMI): a feasibility study based on correlation with PET/CT. NUCLEAR MEDICINE COMM. Volume: 35 Issue: 5 Pages: 446-452
The 4D-PET scanner is characterized essentially by its ability to describe sequentially the development of every single gamma-ray interaction in the detector, including Compton interactions.
The new technology currently being developed is important for society because it will allow:
• Imaging Babies. Imaging babies with PET is important in many critical situations like cardiac arrest or brain injury due to its ability of imaging metabolic status and monitor treatment. However, PET is rarely used with children due to the high-required dose and to its insufficient spatial resolution to image the small structures of the baby body. 4D-PET project will enable imaging babies due to its high sensitivity and outstanding spatial resolution. We aim at a dose for babies of 100μCi, similar to what is routinely injected in mice and 100 times less of a dose in adults. Furthermore, such small dose will allow repeated examinations of the same patient to monitor the evolution to treatment therapy.
• Imaging small structures of the human brain. Sub millimetre resolution is absolutely needed for the visualization of several critical structures in the brain (substantia nigra and raphe nucleus) involved in mental disorders. Brain stem nuclei, substantia nigra and associated areas, as well as subregions of the thalamus have attracted considerable interest in schizophrenia research. A limitation is that the performance of current PET systems does not allow for detailed examination of the dopamine system in the thalamus and its functional implications.
• Imaging tumour heterogeneity. The standard PET/CT (and now also the new PET/MRI) scanners offer PET resolution on a 4-6mm scale and, therefore, cannot image small structures due to the poor resolution but also indirectly due to the decreased efficiency in detecting small features, as the result of the Partial Volume Effect (PVE). Excellent spatial resolution and image quality will allow the study of the geometrical structures of the tumour tissues with high precision. Furthermore, dynamical studies aiming at determining tumour uptake kinetics might also help in determining different components (heterogeneity) inside the tumour (1). Enlightening tumour composition and evolution will ease therapeutic decision, such as the kind of drug to be used in chemotherapy.
• Monitoring Dose in Proton Therapy. Proton therapy is a very useful radiotherapy technique for treating cancer because it optimizes the delivery of dose to the tumour while minimizing the dose in healthy tissues. This technique is especially relevant in children, where vital organs are still under development and might be at risk with conventional RT methods. One way to monitor proton dose delivery is by detecting the scarce gamma rays emitted by the nuclei of the human body after proton irradiation.
The new 4D-PET technology will be validated through the design and development of a preclinical prototype to examine the head of mice. The brain of a mouse has a size of only few millimetres and therefore to be able to visualize the detailed structure of its head and quantify, for instance, the amyloid deposit content represents a very challenging application.
The main performance objectives of this prototype are the following:
• Ultra-High Timing Resolution below 100ps.
• Spatial Resolution Determined by the Physical Limit (0,5mm approx.).
• High Sensitivity over 50% of the mouse brain.
(1) Koolen, Bas B.; … Benlloch J.; et al. Evaluating heterogeneity of primary tumor 18 F-FDG uptake in breast cancer with a dedicated breast PET (MAMMI): a feasibility study based on correlation with PET/CT. NUCLEAR MEDICINE COMM. Volume: 35 Issue: 5 Pages: 446-452
We have developed several technical activities as planned in the methodology section of the grant application.
Several photosensors of the type Silicon Photo Multipliers (SiPMs), Digital (from Philips Digital) and Analog (from Hamamatsu Photonics, SensL, KETEK, ADVANSID) were characterized [9] to achieve the best performance in terms of timing resolution. Also, different crystals (LaBr3, CeBr3, LYSO) and crystal configurations (monolithic and pixelated of different sizes) were tested. Some of the results are presented in publications [1][2][4][7][9][11]. We have used two ASICs from different companies (PETSYS and Weeroc) to read arrays of SiPMs. We have also developed a prototype of a new electronic board to read virtually any ASIC.
During the first part of this grant, several novel technologies have been developed according to the main ideas introduced in the AdG application. Some of these technologies will be used in order to build the small animal PET scanner dedicated to the brain examination of mice [3], which is the final objective of the project.
1) Exploring accurate timing resolution using monolithic crystals and analog SiPM. We have extensively studied the combination of analog SiPMs readout using Application Specific Integration Circuits (ASIC) [11]. We have investigated their performance when using large monolithic blocks (with the aim of differentiating multiple interactions, used for lateral and back readout), but also for pixelated crystals. We reached CTR values of 500 ps FWHM for 50x50x15 mm monolithic crystals, and 180 ps for pixel arrays of 3x3x5 mm. These results have been submitted for publication to Physica Medica [13]. Preliminary studies have been published [11] and also have appeared in indexed conferences proceedings [15][16][17]. Those values of CTR are not good enough for the development of the project. However, we are exploring other ideas to reach the goal of better than 100psec timing resolution. For instance, together with Prof. Paul Lecoq (see section 1.3) we are developing special nanostructures at the surface of scintillating crystals in order to improve the timing resolution of PET detectors.
2) First Compton Camera images have been obtained. A Compton Camera was designed, built and characterized using monolithic LYSO crystals. The geometrical configuration was very similar to the one that is going to be used for a module of the final PET scanner prototype. First images were obtained for an array of 5 small size positron emitting radiative sources, forming a straight line and separated 5mm between each other.
Improving 3D impact identification in monolithic crystals at the crystal edges. Due to the edge effect affecting monolithic crystals, gamma rays impinging near the crystal borders are sometimes loss or present a poor resolution [14][18]. We have developed for this project new methodologies based on the so-called Voronoi diagrams that allow one to recover a good resolution over the whole crystal volume (see the image attached to the public summary report). We will be applying this for instance when using a single LYSO block or sectors merged together with a transparent glue of high refractive index (see collaboration with UVa) [3].
Several photosensors of the type Silicon Photo Multipliers (SiPMs), Digital (from Philips Digital) and Analog (from Hamamatsu Photonics, SensL, KETEK, ADVANSID) were characterized [9] to achieve the best performance in terms of timing resolution. Also, different crystals (LaBr3, CeBr3, LYSO) and crystal configurations (monolithic and pixelated of different sizes) were tested. Some of the results are presented in publications [1][2][4][7][9][11]. We have used two ASICs from different companies (PETSYS and Weeroc) to read arrays of SiPMs. We have also developed a prototype of a new electronic board to read virtually any ASIC.
During the first part of this grant, several novel technologies have been developed according to the main ideas introduced in the AdG application. Some of these technologies will be used in order to build the small animal PET scanner dedicated to the brain examination of mice [3], which is the final objective of the project.
1) Exploring accurate timing resolution using monolithic crystals and analog SiPM. We have extensively studied the combination of analog SiPMs readout using Application Specific Integration Circuits (ASIC) [11]. We have investigated their performance when using large monolithic blocks (with the aim of differentiating multiple interactions, used for lateral and back readout), but also for pixelated crystals. We reached CTR values of 500 ps FWHM for 50x50x15 mm monolithic crystals, and 180 ps for pixel arrays of 3x3x5 mm. These results have been submitted for publication to Physica Medica [13]. Preliminary studies have been published [11] and also have appeared in indexed conferences proceedings [15][16][17]. Those values of CTR are not good enough for the development of the project. However, we are exploring other ideas to reach the goal of better than 100psec timing resolution. For instance, together with Prof. Paul Lecoq (see section 1.3) we are developing special nanostructures at the surface of scintillating crystals in order to improve the timing resolution of PET detectors.
2) First Compton Camera images have been obtained. A Compton Camera was designed, built and characterized using monolithic LYSO crystals. The geometrical configuration was very similar to the one that is going to be used for a module of the final PET scanner prototype. First images were obtained for an array of 5 small size positron emitting radiative sources, forming a straight line and separated 5mm between each other.
Improving 3D impact identification in monolithic crystals at the crystal edges. Due to the edge effect affecting monolithic crystals, gamma rays impinging near the crystal borders are sometimes loss or present a poor resolution [14][18]. We have developed for this project new methodologies based on the so-called Voronoi diagrams that allow one to recover a good resolution over the whole crystal volume (see the image attached to the public summary report). We will be applying this for instance when using a single LYSO block or sectors merged together with a transparent glue of high refractive index (see collaboration with UVa) [3].
The following achievements represent progress beyond the state of the art:
1) Large area scintillating crystals with lateral readout. This was one of the important innovations described in the ERC application. Lateral readout makes sense economically only if the area of the slab is large (total number of SiPM detectors of the lateral surface proportional to 4xLxT, where L and T stand for Length and Thickness, respectively, compared to LxL for conventional back readout). We have successfully achieved excellent position resolution with large slabs (51x51mm2) and thickness of 3mm. The results will be published in the present year (2019). The work has been accepted for contribution at the 2019 IEEE NSS-MIC in November (Manchester, UK). A preliminary plot with 11x11 sources is shown in the public summary of this report, reaching an average spatial resolution well below 2 mm FWHM with an energy resolution of around 12%.
2) Distinguishing several interactions inside a full crystal block. An incident gamma ray is scattered within the detector (Compton scattering) and/or absorbed by a photoelectric effect [1][2]. Both the Compton scattering and the photoelectric effect produce scintillation photons (optical photons) of the order of a few thousand per MeV that are uniformly distributed in a 4π solid angle (spherical distribution). The electronic readout of the scintillated optical photons is performed with a pixelated photo-detector. By combining the energy deposition and time-stamp information from each pixel, the detector (by using the novel proposed method) can precisely determine the 2D coordinates of various Compton interactions (of an incident gamma ray) and/or a photoelectric effect. Moreover, using the energy distribution it can also calculate the DOI of each interaction and as the result of a fit. It can also estimate the corresponding deposited energy (in keV) of each hit. This whole information can be used for an improved PET apparatus where the coordinates of the first interaction are precisely given and therefore the lines of response are determined more accurately. In this case only the total energy would be needed. By additionally using the energy information for each individual interaction (hit) coming from one incident gamma ray, one can even use the detector as a Compton camera.
3) Image Quality Improvement through Identifying First Interaction in Compton Sequence.
We have realized that an excellent TOF resolution will greatly impact sensitivity, spatial resolution, and image quality in PET scanners [1][2], not only through the very well-known direct assessment of the positron annihilation position along the line of response (LOR) of the pair of detected gamma rays, but also due to the ability to better treat Compton events. This has come as a nice unexpected result from realistic simulations.
GE Healthcare has reported an increase in PET scanner sensitivity of almost 20% by including events which experience interactions at neighboring detectors. Compton Scatter Recovery (CSR) detects scatter of 511 keV photons between adjacent detector blocks, and reconstitutes events in which the summed energy falls within the energy window. However, this nice increase in sensitivity comes at the cost of some image blurring since the scanner is not able to unequivocally identify which interaction came first and, therefore, the LOR provided by the system is not the real one. Obviously, a good timing resolution will enable the scanner to include those events without any image blurring.
4) Planar Gamma Camera with 3D capability.
While trying to increase the sensitivity of a Compton Camera module in the first part of the project, we realized that additional gamma ray detectors placed perpendicular to and at the edges of the standard ones would cover the whole angular area, hence achieving 100% sensitivity. Then, we applied this idea to a conventional gamma camera. Therefore, this idea is a by-product of the present grant.
Maintaining the configuration of a multi-pinhole collimator with a high degree of overlapping (providing thus a high sensitivity) we add a new element, an “active collimator” which is a radiation sensitive detector that also acts as a collimator, preventing the gamma rays to reach the overlapping region. Thus, besides acting as a collimator, it is also able to measure the impact coordinates of the incident photon. This way all the needed information is retrieved and we are able to unambiguously identify through which pinhole any gamma ray passes before being detected. The result is a high resolution and sensitivity gamma camera device that eliminates the overlapping problem without producing image truncation or any artifacts. Maintaining the high sensitivity property of a multi-pinhole collimator, we obtain a high resolution reconstructed image for an arbitrarily large FOV. This gamma camera can be used for obtaining two-dimensional images (in vivo) and also three-dimensional images as in Single Photon Emission Computed Tomography, for animals or humans.
5) A new crystal block configuration that allows simultaneous 4D excellent resolution.
This idea is the result of what we have learned throughout the duration of this project so far. It is not possible with the current technology to achieve simultaneously excellent X-Y position resolution, DOI resolution, energy resolution and CRT resolution [11][13]. We obtain excellent energy and timing resolution with pixelated crystals but no DOI information at all. We obtain excellent X-Y and DOI resolution with continuous crystals but no good enough timing resolution with current SiPM technology. The new design overcomes these problems with a new crystal block configuration that simultaneously achieves excellent resolution in all the aforementioned quantities [15][16]. A detailed explanation of the idea will be described in the next report of the ERC AdG.
Until the end of the project we expect to achieve the following results:
• Validate the new 4D-PET technology through the development of a prototype of a PET scanner enhance with Compton layers to examine the head of mice.
• Reconstruct the Compton events and reject scatter and random coincidences with the Compton information.
• Use the timing information to provide an attenuation map of the brain of the mouse.
• Compare the performance of such scanner with Compton information with a standard scanner with respect to spatial and timing resolution, signal to noise ratio and image quality.
1) Large area scintillating crystals with lateral readout. This was one of the important innovations described in the ERC application. Lateral readout makes sense economically only if the area of the slab is large (total number of SiPM detectors of the lateral surface proportional to 4xLxT, where L and T stand for Length and Thickness, respectively, compared to LxL for conventional back readout). We have successfully achieved excellent position resolution with large slabs (51x51mm2) and thickness of 3mm. The results will be published in the present year (2019). The work has been accepted for contribution at the 2019 IEEE NSS-MIC in November (Manchester, UK). A preliminary plot with 11x11 sources is shown in the public summary of this report, reaching an average spatial resolution well below 2 mm FWHM with an energy resolution of around 12%.
2) Distinguishing several interactions inside a full crystal block. An incident gamma ray is scattered within the detector (Compton scattering) and/or absorbed by a photoelectric effect [1][2]. Both the Compton scattering and the photoelectric effect produce scintillation photons (optical photons) of the order of a few thousand per MeV that are uniformly distributed in a 4π solid angle (spherical distribution). The electronic readout of the scintillated optical photons is performed with a pixelated photo-detector. By combining the energy deposition and time-stamp information from each pixel, the detector (by using the novel proposed method) can precisely determine the 2D coordinates of various Compton interactions (of an incident gamma ray) and/or a photoelectric effect. Moreover, using the energy distribution it can also calculate the DOI of each interaction and as the result of a fit. It can also estimate the corresponding deposited energy (in keV) of each hit. This whole information can be used for an improved PET apparatus where the coordinates of the first interaction are precisely given and therefore the lines of response are determined more accurately. In this case only the total energy would be needed. By additionally using the energy information for each individual interaction (hit) coming from one incident gamma ray, one can even use the detector as a Compton camera.
3) Image Quality Improvement through Identifying First Interaction in Compton Sequence.
We have realized that an excellent TOF resolution will greatly impact sensitivity, spatial resolution, and image quality in PET scanners [1][2], not only through the very well-known direct assessment of the positron annihilation position along the line of response (LOR) of the pair of detected gamma rays, but also due to the ability to better treat Compton events. This has come as a nice unexpected result from realistic simulations.
GE Healthcare has reported an increase in PET scanner sensitivity of almost 20% by including events which experience interactions at neighboring detectors. Compton Scatter Recovery (CSR) detects scatter of 511 keV photons between adjacent detector blocks, and reconstitutes events in which the summed energy falls within the energy window. However, this nice increase in sensitivity comes at the cost of some image blurring since the scanner is not able to unequivocally identify which interaction came first and, therefore, the LOR provided by the system is not the real one. Obviously, a good timing resolution will enable the scanner to include those events without any image blurring.
4) Planar Gamma Camera with 3D capability.
While trying to increase the sensitivity of a Compton Camera module in the first part of the project, we realized that additional gamma ray detectors placed perpendicular to and at the edges of the standard ones would cover the whole angular area, hence achieving 100% sensitivity. Then, we applied this idea to a conventional gamma camera. Therefore, this idea is a by-product of the present grant.
Maintaining the configuration of a multi-pinhole collimator with a high degree of overlapping (providing thus a high sensitivity) we add a new element, an “active collimator” which is a radiation sensitive detector that also acts as a collimator, preventing the gamma rays to reach the overlapping region. Thus, besides acting as a collimator, it is also able to measure the impact coordinates of the incident photon. This way all the needed information is retrieved and we are able to unambiguously identify through which pinhole any gamma ray passes before being detected. The result is a high resolution and sensitivity gamma camera device that eliminates the overlapping problem without producing image truncation or any artifacts. Maintaining the high sensitivity property of a multi-pinhole collimator, we obtain a high resolution reconstructed image for an arbitrarily large FOV. This gamma camera can be used for obtaining two-dimensional images (in vivo) and also three-dimensional images as in Single Photon Emission Computed Tomography, for animals or humans.
5) A new crystal block configuration that allows simultaneous 4D excellent resolution.
This idea is the result of what we have learned throughout the duration of this project so far. It is not possible with the current technology to achieve simultaneously excellent X-Y position resolution, DOI resolution, energy resolution and CRT resolution [11][13]. We obtain excellent energy and timing resolution with pixelated crystals but no DOI information at all. We obtain excellent X-Y and DOI resolution with continuous crystals but no good enough timing resolution with current SiPM technology. The new design overcomes these problems with a new crystal block configuration that simultaneously achieves excellent resolution in all the aforementioned quantities [15][16]. A detailed explanation of the idea will be described in the next report of the ERC AdG.
Until the end of the project we expect to achieve the following results:
• Validate the new 4D-PET technology through the development of a prototype of a PET scanner enhance with Compton layers to examine the head of mice.
• Reconstruct the Compton events and reject scatter and random coincidences with the Compton information.
• Use the timing information to provide an attenuation map of the brain of the mouse.
• Compare the performance of such scanner with Compton information with a standard scanner with respect to spatial and timing resolution, signal to noise ratio and image quality.