| Περίληψη: | A valid definition of molecular imaging could be the noninvasive, real-time visualization of biochemical events at the cellular and molecular level within living cells, tissues, and/or intact subjects.The words molecular imaging mean different things to various groups, and thus the areas of research and medicine that fall under the umbrella of molecular imaging are incredibly vast and varied.Generally speaking, molecular imaging involves specialized instrumentation, used alone or in combination with targeted imaging agents, to visualize tissue characteristics and/or biochemical processes. The data provided from molecular imaging studies can be used to help understand biological phenomena, identify regions of pathology, and provide insight regarding the mechanisms of pathogens.
The PET-MRI combination requires the implementation of four technologic achivements that influence current state-of-the-art PET and MRI.
First, the photomultiplier technology must be replaced with magnetic field–insensitive avalanche photodiodes.
Second, compact PET detectors must be constructed in such a way to be transparent to the MRI and so to not interfere with the field gradients or MR radiofrequency.
Third, the MRI scanner must be adapted to accommodate the PET detectors and to allow simultaneous and concurrent data acquisition.
Finally, investigation on the optimum reconstruction strategies to accompany such a system incorporating completely new procedures for PET attenuation correction, based solely on MRI information, have to be performed.
The development of integrated PET-MRI is, therefore, a comprehensive endeavor that requires a significant advancement of PET detector technology, MRI system integration, and new software approaches.
Historically, PET systems have generally developed as circular “rings”. The earliest tomographs consisted of few detectors that rotated and translated to obtain a complete set of projection data, but soon full ring systems were developed. Yet, dual head PET scanners, due to their smaller size, compact geometry and closer proximity to the source can provide optimum dedicated scanning. In other cases imaging can be performed where convensional full ring geometries cannot be used. The have been proven valuable tools in preclinical imaging and are emerging in clinical cases like in PEM (Positron Emission Mammography).
For the current Ph.D. thessis a planar dual head PET system was used for the evaluation of the reconstruction algorithms as well as the validation of the simulation models. It was developed by the Detector and Imaging Group of the Thomas Jefferson National Accelerator Facility (USA) in collaboration with the Medical Instruments Technology dep. of the Technological Educational Institute of Athens and is currently installed at the Institute of Radioisotopes - Radiodiagnostic prod. at the National Center for Scientific Research “Demokritos”.
The system has two planar detectors. Each head contains one Hamamatsu H8500 PSPMT with 50$\times$50 mm$^2$ active size; an LSO:Ce crystal array with 20x20 pixels, 2x2x10 mm^3 in size. The septa between the crystals are 0.2 mm.
The two detector heads were mounted on a gantry made initially from wood and afterwards from plastic. The materials were selected for their magnetic tolerance and low cost. In addition, their construction allow the easy adjustment of the head separation distance according to the needs of the experiment. The minimum separation distance between the two heads can be 7 cm and the maximum 14 cm. Moreover, it is capable to accept two additional heads in order to support a quad head system.
The system is able to provide images without rotation using the Focal Plane Tomography algorithm. While using step and shoot acquisition it can provide tomographic images based on the acquired planar projection data or data obtained in listmode format and sorted in 3D sinograms similar to cylindrical PET systems.
Evaluation of the system under planar imaging showed that for head separation distance 5 cm, the system maintains its linear performance for activities up to 3.5 MBq, which is sufficient for mice applications. For larger separations distances this value is well above 4 MBq.
It is fully capable of providing fast planar coincidence images as well as non-kinetic tomographic images using a step and shoot rotation. The main drawback of the rotating head approach remains of low sensitivity compared to the full-ring systems. The best spatial resolution, in the center of the FOV, is 2.5 mm in the planar mode and 1.9 mm in the tomographic mode.
For head separation distances below 8 cm the FOV appears to be uniform in the central 4x4 cm^2 area in planar and in tomographic acquisitions. Further on the edges the sensitivity is reduced to the 10%. The performance of the system in imaging small animals, despite any limitations on the reconstruction, is considered satisfactory.Fast planar images, for pharmaceutical kinetic analysis, can be obtained. While using the rotating capabilities of the gantry, all the important anatomical structures are imaged in detail.
The geometry of the prototype system was simulated using GATE 6.0. Two simulation models were implemented and validated. With and without the ^176Lu radioactivity, since the LSO intrinsic radioactivity is not included by default to GATE simulations. The two models were validated with reference experimental data in terms of dark count rate, count rate performance (cps) and scatter fraction (sf). In addition the effect of the low level discriminator (LLD) threshold on signal as well as image quality is compared to the effect of the software energy window.
The intrinsic radioactivity concentration of the ^176Lu was found in literature as 277Bq/cm^3. The intrinsic activity was uniformly distributed within the volume of the crystal array, accounting for the septa volume between the crystals.
Close investigation on the origin of the detected events in the simulated data, concluded that the use of high LLD thresholds and a wide energy window improves the sensitivity of the system in terms of NECR, since greater number of true events are detected while randoms and scatters are early rejected. Investigation on the SNR properties, using a additional water phantom,to approximate the small animal body, showed that the value peaks when the low energy window limit is at 350 keV. Below that limit the scatters are strongly increased and above a portion of the trues is rejected.
The minimum detectable activity of the system was assessed to 12.4 KBq, under the aforementioned imaging conditions. Using a more complex phantom, rather than a capillary source, the minimum detectable activity is expected to take higher values.
Simulation were carried out incorporating the influence of a static magnetic field. The results suggested great improvement on the minimum detectable activity, in the case were there is not sufficient medium around the source for positron annihilation. Hence, improvements on the detectability of small lesions in the lungs of near the skin, are to be expected in an PET/MR module. This is a positive side effect of the magnetic field which has not been stressed out in literature. In addition, the results showed that the spatial resolution of the system got improved, as expected.
In order to address the limitations of the rotating planar reconstruction, STIR reconstruction toolkit was introduced. STIR is a well validated reconstruction toolkit providing the OSEM algorithm, accompanied with a great variety of applicable options and filters. For the current studies only OSEM with 2 iterations was used. Possible image improvements on the image quality with the use of filters and priors was out of the scopes of this thesis.
In addition, a component based normalization technique and an attenuation correction approximation were applied during the reconstruction. In order to produce the normalization sinograms two different source were simulated. First, an ideal cylindrical source, covering the entire FOV for the extraction of the axial geometric factors and the detector efficiencies. Second, a planar rotating source in order to calculate the transverse geometric factors and crystal interference functions. For the calculation of the experimental PET's detector efficiencies an plastic planar source was constructed rotated mathematically around the FOV, in order to approximate the ideal cylindrical of the simulation.
The components of the normalization were geometric (transverse and axial), detector efficiencies and crystal interference functions. The normalized reconstructed images images, simulated as well as experimental, demonstrate uniform sensitivity inside the FOV.
The final, part of a small animal imaging PET system, compatible with strong magnetic fields, which was under investigation, was the part of the detector. Current trends lead to the SiPMs as the next generation of PET detectors due to the magnetic tolerance.
SiPM detectors purchased from SensL were evaluated in terms of their output pulse and noise characteristics, photon detection efficiency and linearity over the bias voltage and the energy of the irradiating source. Two SiPM detectors were evaluated SPMM-3020 and SPMM-3035. The differences reside on the difference size, wherefore and total number, of SPAD elements. In order to overcome limitations on the manufacturer's electronics a custom amplifier was designed and implemented. The amplifier was able to condition the signals from both SiPM to be acquired correctly from the DAQ.
SPMM-3020, which had more and smaller SPAD elements showed a more linear response under a wide variety of conditions ranging to different operational voltages to crystals with higher light output irradiated from sources with different γ-photon energy. In addition, under normal room temperature the noise propertied were superior over SiPM-3035. The results indicate that this detector would be the preferable choice for a SPECT system, which the imaging protocols require the linear and accurate detect many different $\gamma$-photon energies.
SPMM-2035, which had less and larger SPAD elements displayed better energy resolution and a narrower but adjustable (through the operating voltage) linear area. The provided signal was higher, hence less amplification was demanded for it to be recorded, even after long transition though cables. These merits make it a suitable candidate for PET-MR scanners since in PET imaging the energy of the detected γ-photons is only 511 keV and the linear area can be adjusted according to the crystal's light output.
|
|---|
