Kyeong Yun Kim

Simultaneous Multiparametric PET/MRI with Silicon Photomultiplier PET and Ultra-High-Field MRI for Small-Animal Imaging

Visualization of biologic processes at molecular and cellular levels has revolutionized the understanding and treatment of human diseases. However, no single biomedical imaging modality provides complete information, resulting in the emergence of multimodal approaches. Combining state-of-the-art PET and MRI technologies without loss of system performance and overall image quality can provide opportunities for new scientific and clinical innovations. Here, we present a multiparametric PET/MR imager based on a small-animal dedicated, high-performance, silicon photomultiplier (SiPM) PET system and a 7-T MR scanner. Methods: A SiPM-based PET insert that has the peak sensitivity of 3.4% and center volumetric resolution of 1.92/0.53 mm3 (filtered backprojection/ordered-subset expectation maximization) was developed. The SiPM PET insert was placed between the mouse body transceiver coil and gradient coil of a 7-T small-animal MRI scanner for simultaneous PET/MRI. Mutual interference between the MRI and SiPM PET systems was evaluated using various MR pulse sequences. A cylindric corn oil phantom was scanned to assess the effects of the SiPM PET on the MR image acquisition. To assess the influence of MRI on the PET imaging functions, several PET performance indicators including scintillation pulse shape, flood image quality, energy spectrum, counting rate, and phantom image quality were evaluated with and without the application of MR pulse sequences. Simultaneous mouse PET/MRI studies were also performed to demonstrate the potential and usefulness of the multiparametric PET/MRI in preclinical applications. Results: Excellent performance and stability of the PET system were demonstrated, and the PET/MRI combination did not result in significant image quality degradation of either modality. Finally, simultaneous PET/MRI studies in mice demonstrated the feasibility of the developed system for evaluating the biochemical and cellular changes in a brain tumor model and facilitating the development of new multimodal imaging probes. Conclusion: We developed a multiparametric imager with high physical performance and good system stability and demonstrated its feasibility for small-animal experiments, suggesting its usefulness for investigating in vivo molecular interactions of metabolites, and cross-validation studies of both PET and MRI.

Evaluation of a silicon photomultiplier PET insert for simultaneous PET and MR imaging

PURPOSE In this study, the authors present a silicon photomultiplier (SiPM)-based positron emission tomography (PET) insert dedicated to small animal imaging with high system performance and robustness to temperature change. METHODS The insert consists of 64 LYSO-SiPM detector blocks arranged in 4 rings of 16 detector blocks to yield a ring diameter of 64 mm and axial field of view of 55 mm. Each detector block consists of a 9 × 9 array of LYSO crystals (1.2 × 1.2 × 10 mm(3)) and a monolithic 4 × 4 SiPM array. The temperature of each monolithic SiPM is monitored, and the proper bias voltage is applied according to the temperature reading in real time to maintain uniform performance. The performance of this PET insert was characterized using National Electrical Manufacturers Association NU 4-2008 standards, and its feasibility was evaluated through in vivo mouse imaging studies. RESULTS The PET insert had a peak sensitivity of 3.4% and volumetric spatial resolutions of 1.92 (filtered back projection) and 0.53 (ordered subset expectation maximization) mm(3) at center. The peak noise equivalent count rate and scatter fraction were 42.4 kcps at 15.08 MBq and 16.5%, respectively. By applying the real-time bias voltage adjustment, an energy resolution of 14.2% ± 0.3% was maintained and the count rate varied ≤1.2%, despite severe temperature changes (10-30 °C). The mouse imaging studies demonstrate that this PET insert can produce high-quality images useful for imaging studies on the small animals. CONCLUSIONS The developed MR-compatible PET insert is designed for insertion into a narrow-bore magnetic resonance imaging scanner, and it provides excellent imaging performance for PET/MR preclinical studies.

Prototype pre-clinical PET scanner with depth-of-interaction measurements using single-layer crystal array and single-ended readout

In this study, we developed a proof-of-concept prototype PET system using a pair of depth-of-interaction (DOI) PET detectors based on the proposed DOI-encoding method and digital silicon photomultiplier (dSiPM). Our novel cost-effective DOI measurement method is based on a triangular-shaped reflector that requires only a single-layer pixelated crystal and single-ended signal readout. The DOI detector consisted of an 18  ×  18 array of unpolished LYSO crystal (1.47  ×  1.47  ×  15 mm3) wrapped with triangular-shaped reflectors. The DOI information was encoded by depth-dependent light distribution tailored by the reflector geometry and DOI correction was performed using four-step depth calibration data and maximum-likelihood (ML) estimation. The detector pair and the object were placed on two motorized rotation stages to demonstrate 12-block ring PET geometry with 11.15 cm diameter. Spatial resolution was measured and phantom and animal imaging studies were performed to investigate imaging performance. All images were reconstructed with and without the DOI correction to examine the impact of our DOI measurement. The pair of dSiPM-based DOI PET detectors showed good physical performances respectively: 2.82 and 3.09 peak-to-valley ratios, 14.30% and 18.95% energy resolution, and 4.28 and 4.24 mm DOI resolution averaged over all crystals and all depths. A sub-millimeter spatial resolution was achieved at the center of the field of view (FOV). After applying ML-based DOI correction, maximum 36.92% improvement was achieved in the radial spatial resolution and a uniform resolution was observed within 5 cm of transverse PET FOV. We successfully acquired phantom and animal images with improved spatial resolution and contrast by using the DOI measurement. The proposed DOI-encoding method was successfully demonstrated in the system level and exhibited good performance, showing its feasibility for animal PET applications with high spatial resolution and sensitivity.

Improving accuracy of simultaneously reconstructed activity and attenuation maps using deep learning

Simultaneous reconstruction of activity and attenuation using the maximum-likelihood reconstruction of activity and attenuation (MLAA) augmented by time-of-flight information is a promising method for PET attenuation correction. However, it still suffers from several problems, including crosstalk artifacts, slow convergence speed, and noisy attenuation maps (μ-maps). In this work, we developed deep convolutional neural networks (CNNs) to overcome these MLAA limitations, and we verified their feasibility using a clinical brain PET dataset. Methods: We applied the proposed method to one of the most challenging PET cases for simultaneous image reconstruction (18F-fluorinated-N-3-fluoropropyl-2-β-carboxymethoxy-3-β-(4-iodophenyl)nortropane [18F-FP-CIT] PET scans with highly specific binding to striatum of the brain). Three different CNN architectures (convolutional autoencoder [CAE], Unet, and Hybrid of CAE) were designed and trained to learn a CT-derived μ-map (μ-CT) from the MLAA-generated activity distribution and μ-map (μ-MLAA). The PET/CT data of 40 patients with suspected Parkinson disease were used for 5-fold cross-validation. For the training of CNNs, 800,000 transverse PET and CT slices augmented from 32 patient datasets were used. The similarity to μ-CT of the CNN-generated μ-maps (μ-CAE, μ-Unet, and μ-Hybrid) and μ-MLAA was compared using Dice similarity coefficients. In addition, we compared the activity concentration of specific (striatum) and nonspecific (cerebellum and occipital cortex) binding regions and the binding ratios in the striatum in the PET activity images reconstructed using those μ-maps. Results: The CNNs generated less noisy and more uniform μ-maps than the original μ-MLAA. Moreover, the air cavities and bones were better resolved in the proposed CNN outputs. In addition, the proposed deep learning approach was useful for mitigating the crosstalk problem in the MLAA reconstruction. The Hybrid network of CAE and Unet yielded the most similar μ-maps to μ-CT (Dice similarity coefficient in the whole head = 0.79 in the bone and 0.72 in air cavities), resulting in only about a 5% error in activity and binding ratio quantification. Conclusion: The proposed deep learning approach is promising for accurate attenuation correction of activity distribution in time-of-flight PET systems.

Proof-of-concept prototype time-of-flight PET system based on high-quantum-efficiency multianode PMTs

Purpose: Time‐of‐flight (TOF) information in positron emission tomography (PET) scanners enhances the diagnostic power of PET scans owing to the increased signal‐to‐noise ratio of reconstructed images. There are numerous additional benefits of TOF reconstruction, including the simultaneous estimation of activity and attenuation distributions from emission data only. Exploring further TOF gains by using TOF PET scanners is important because it can broaden the applications of PET scans and expand our understanding of TOF techniques. Herein, we present a prototype TOF PET scanner with fine‐time performance that can experimentally demonstrate the benefits of TOF information. Methods: A single‐ring PET system with a coincidence resolving time of 360 ps and a spatial resolution of 3.1/2.2 mm (filtered backprojection/ordered‐subset expectation maximization) was developed. The scanner was based on advanced high‐quantum‐efficiency (high‐QE) multianode photomultiplier tubes (PMTs). The impact of its fine‐time performance was demonstrated by evaluating body phantom images reconstructed with and without TOF information. Moreover, the feasibility of the scanner as an experimental validator of TOF gains was verified by investigating the improvement of images under various conditions, such as the use of joint estimation algorithms of activity and attenuation, erroneous data correction factors (e.g., without normalization correction), and incompletely sampled data. Results: The prototype scanner showed excellent performance, producing improved phantom images, when TOF information was employed in the reconstruction process. In addition, investigation of the TOF benefits using the phantom data in different conditions verified the usefulness of the developed system for demonstrating the practical effects of TOF reconstruction. Conclusions: We developed a prototype TOF PET scanner with good performance and a fine‐timing resolution based on advanced high‐QE multianode PMTs and demonstrated its feasibility as an experimental validator of TOF gains, suggesting its usefulness for investigating new applications of PET scans and clarifying TOF techniques in detail.

Computed tomography super-resolution using deep convolutional neural network

The objective of this study is to develop a convolutional neural network (CNN) for computed tomography (CT) image super-resolution. The network learns an end-to-end mapping between low (thick-slice thickness) and high (thin-slice thickness) resolution images using the modified U-Net. To verify the proposed method, we train and test the CNN using axially averaged data of existing thin-slice CT images as input and their middle slice as the label. Fifty-two CT studies are used as the CNN training set, and 13 CT studies are used as the test set. We perform five-fold cross-validation to confirm the performance consistency. Because all input and output images are used in two-dimensional slice format, the total number of slices for training the CNN is 7670. We assess the performance of the proposed method with respect to the resolution and contrast, as well as the noise properties. The CNN generates output images that are virtually equivalent to the ground truth. The most remarkable image-recovery improvement by the CNN is deblurring of boundaries of bone structures and air cavities. The CNN output yields an approximately 10% higher peak signal-to-noise ratio and lower normalized root mean square error than the input (thicker slices). The CNN output noise level is lower than the ground truth and equivalent to the iterative image reconstruction result. The proposed deep learning method is useful for both super-resolution and de-noising.

Joint estimation of activity distribution and attenuation map for TOF-PET using alternating direction method of multiplier

Recent advances in TOF PET joint estimation of activity and attenuation showed that activity and attenuation can be determined up to a global constant scale without severe crosstalk. MLAA was first proposed to estimate activity and attenuation map simultaneously, and then MLACF was developed to estimate activity and attenuation compensation factor (ACF). MLAA incorporated prior knowledge on the zero attenuation value outside body area to determine global scalar, but was slow to converge. MLACF converged much faster than MLAA, but required knowing total activity level in advance. We propose a new optimization method based on variable splitting and alternating direction method of multiplier (MLADMM). Our proposed MLADMM achieved fast convergence rate comparable to MLACF without knowing total activity level. MLADMM also has a potential to use more sophisticated MR-based prior for attenuation in PET-MR.

Simultaneous positron emission tomography and magnetic resonance imaging

Positron emission tomography (PET) scanners provide quantitative information about various physiological and biochemical processes that occur in a living body (e.g., glucose metabolism, gene expression, and drug occupancy). In general, scintillation detectors—consisting of inorganic scintillation crystal and photosensor arrays—are used in PET scanners to measure the gamma rays that are emitted from radiopharmaceuticals. In this process, visible or UV photons are emitted when gamma rays are detected by the scintillation crystal and are then measured by the photosensors. There are, however, several problems associated with these traditional PET detection systems (e.g., limited efficiency and high cost) and alternative options are therefore required. The rapid development of silicon photomultipliers (SiPMs) means that these photosensors can now offer extremely high count rate detection and photon number resolution.1, 2 One of the most promising applications of these innovative devices is the simultaneous acquisition of PET and magnetic resonance imaging (MRI) scans for the study of animal models of human diseases. However, there are several technical challenges connected with this approach. For instance, there is only limited space for PET detectors and electronics inside the MRI bore, and mutual interference between the PET and MRI components tends to occur.3–5 Most previously developed small-animal PET/MRI systems have been based on avalanche photodiode (APD) technology.6, 7 The relatively poor physical performance of APD-based PET systems, however, has further heightened the demand for the next generation of PET/MRI systems (i.e., based on more advanced semiconductor photosensors). Figure 1. Illustration of previously developed prototype positron emission tomography (PET) systems that include silicon photomultipliers (SiPMs). (a) Photograph of the first SiPM PET prototype and a corresponding PET/computed tomography image. (b) The SiPM PET insert that is compatible with magnetic resonance imaging (MRI) inside a 3Tesla MRI scanner, and a PET/MRI image acquired with this system.8, 9

Lymph node imaging using novel simultaneous PET/MRI and dual-modality imaging agent

Lymph node (LN) imaging has clinical significance because the invasion status of the LN is crucial information for disease stratification, staging, and management. Compared with conventional method, simultaneous PET/MRI using the multi-modal imaging marker can offer synergistic advantages. Here, we present LN mapping using selfdeveloped PET/MRI and novel bimodal biomarker. The SiPM-based PET insert which has peak sensitivity of 3.4% and center volumetric resolution of 0.57 cubic mm was developed. The PET insert was placed between the RF and gradient coil of Bruker 7T MRI. For LN targeting, 64Cu-NOTA-ironoxide-mannose, a new LN targeted dualmodality probe yielding superior T2 contrast was used. The relaxivity of the probe was measured by phantom study. Before the simultaneous imaging, MRI of the left and right popliteal lymph node of an anaesthetized BALB/c mouse was acquired as the control. 10 μL of the tracer was then injected into the left hindpaw of same mouse. Simultaneous PET/MRI was acquired for 10 minutes after 10-min and 120-min uptake period. The r2 of the probe was 845.3 at 7T magnetic field. The simultaneous PET/MRI has sufficient resolution and sensitivity to imaging tiny organ like mouse LN. A left popliteal LN in the 120-min post-injection MRI resulted in a remarkable signal decrease compared to those in the pre-injection MRI and that was good agreement with PET. The 10-min post-injection PET also showed clear regional activity in the LN, but the 10-min uptake period was not sufficient to generate MR contrast. It is due to the large difference in the sensitivities of the two modality. The simultaneous PET/MRI is useful for in vivo imaging and bimodal imaging probe development particularly to generating negative T2 contrast.