Low-dose computed tomography (LDCT) denoising methods based on supervised learning with labeled simulation data have made significant progress. However, these methods usually struggle to directly process unlabeled LDCT images due to inherent biases. While unsupervised methods have been explored to utilize unlabeled LDCT images, they typically involve complex network structures with limited denoising performance. To address these issues, we propose a self-adaptive weight embedded lightweight semi-supervised network (SWELNet) for unlabeled LDCT image denoising, which integrates supervised and unsupervised learning in a lightweight architecture. Unlike other semi-supervised algorithms that only consider the correlations between labeled simulation data and unlabeled data, the proposed SWELNet not only takes into account correlations but also the differences between data. There are three modules in the proposed network respectively for feature extraction, refinement and self-adaptive weight. Specially, the multi-scale convolution feature extraction module (MCFEM) and recursive module (RECM) extract and refine common representations from labeled simulation and unlabeled data with the well-designed. After that, the softmax feature fusion module (SFFM) with self-adaptive weighted learning for forming different feature spaces for two types of data. Extensive experiments using one simulation and two unlabeled datasets demonstrate that the proposed SWELNet outperforms several state-of-the-art baseline network methods in terms of robustness and generalization, as well as inference efficiency. The code is available at https://github.com/nightastars/SWELNet-main.git.

Deep learning (DL) based image processing methods have been successfully applied to low-dose x-ray images based on the assumption that the feature distribution of the training data is consistent with that of the test data. However, low-dose computed tomography (LDCT) images from different commercial scanners may contain different amounts and types of image noise, violating this assumption. Moreover, in the application of DL based image processing methods to LDCT, the feature distributions of LDCT images from simulation and clinical CT examination can be quite different. Therefore, the network models trained with simulated image data or LDCT images from one specific scanner may not work well for another CT scanner and image processing task. To solve such domain adaptation problem, in this study, a novel generative adversarial network (GAN) with noise encoding transfer learning (NETL), or GAN-NETL, is proposed to generate a paired dataset with a different noise style. Specifically, we proposed a method to perform noise encoding operator and incorporate it into the generator to extract a noise style. Meanwhile, with a transfer learning (TL) approach, the image noise encoding operator transformed the noise type of the source domain to that of the target domain for realistic noise generation. One public and two private datasets are used to evaluate the proposed method. Experiment results demonstrated the feasibility and effectiveness of our proposed GAN-NETL model in LDCT image synthesis. In addition, we conduct additional image denoising study using the synthesized clinical LDCT data, which verified the merit of the proposed synthesis in improving the performance of the DL based LDCT processing method.

BACKGROUND:In recent years, low-dose computed tomography (LDCT) has played an important role in the diagnosis CT to reduce the potential adverse effects of X-ray radiation on patients, while maintaining the same diagnostic image quality.PURPOSE:Deep learning (DL)-based methods have played an increasingly important role in the field of LDCT imaging. However, its performance is highly dependent on the consistency of feature distributions between training data and test data. Due to patient's breathing movements during data acquisition, the paired LDCT and normal dose CT images are difficult to obtain from realistic imaging scenarios. Moreover, LDCT images from simulation or clinical CT examination often have different feature distributions due to the pollution by different amounts and types of image noises. If a network model trained with a simulated dataset is used to directly test clinical patients' LDCT data, its denoising performance may be degraded. Based on this, we propose a novel domain-adaptive denoising network (DADN) via noise estimation and transfer learning to resolve the out-of-distribution problem in LDCT imaging.METHODS:To overcome the previous adaptation issue, a novel network model consisting of a reconstruction network and a noise estimation network was designed. The noise estimation network based on a double branch structure is used for image noise extraction and adaptation. Meanwhile, the U-Net-based reconstruction network uses several spatially adaptive normalization modules to fuse multi-scale noise input. Moreover, to facilitate the adaptation of the proposed DADN network to new imaging scenarios, we set a two-stage network training plan. In the first stage, the public simulated dataset is used for training. In the second transfer training stage, we will continue to fine-tune the network model with a torso phantom dataset, while some parameters are frozen. The main reason using the two-stage training scheme is based on the fact that the feature distribution of image content from the public dataset is complex and diverse, whereas the feature distribution of noise pattern from the torso phantom dataset is closer to realistic imaging scenarios.RESULTS:In an evaluation study, the trained DADN model is applied to both the public and clinical patient LDCT datasets. Through the comparison of visual inspection and quantitative results, it is shown that the proposed DADN network model can perform well in terms of noise and artifact suppression, while effectively preserving image contrast and details.CONCLUSIONS:In this paper, we have proposed a new DL network to overcome the domain adaptation problem in LDCT image denoising. Moreover, the results demonstrate the feasibility and effectiveness of the application of our proposed DADN network model as a new DL-based LDCT image denoising method.

The purpose of this study is to design and develop a high-resolution handheld gamma camera for thyroid and sentinel lymph nodes imaging. The detector of the camera is based on a continuous NaI(Tl) crystal directly coupled to a Position Sensitive Photo Multiplier Tube (PSPMT) with a highly integrated readout circuit to achieve an intrinsic spatial resolution, Ri, at ~1 mm at a much lower cost than a pixelated CZT detector based camera with similar Ri. A novel readout electronic system based on the scintillation light distribution function and least square estimation (LSE) positioning algorithm implemented on a previously developed FPGA board to provide the high Ri, throughout the entire detector area. Two prototype parallel-hole collimators were designed and fabricated for general purpose and high sensitivity thyroid and sentinel imaging. Preliminary test results show that the Ri of our detector meet the designed target, which has exceeded that of most commercially available handheld gamma cameras. The other imaging characteristics are compatible to the designed parameters. In conclusion, the new handheld gamma camera has the potential for small organ imaging with higher performance characteristics at a lower cost than those that are currently available.

The accelerating complexity and variety of medical imaging devices and methods have outpaced the ability to evaluate and optimize their design and clinical use. This is a significant and increasing challenge for both scientific investigations and clinical applications. Evaluations would ideally be done using clinical imaging trials. These experiments, however, are often not practical due to ethical limitations, expense, time requirements, or lack of ground truth. Virtual clinical trials (VCTs) (also known as in silico imaging trials or virtual imaging trials) offer an alternative means to efficiently evaluate medical imaging technologies virtually. They do so by simulating the patients, imaging systems, and interpreters. The field of VCTs has been constantly advanced over the past decades in multiple areas. We summarize the major developments and current status of the field of VCTs in medical imaging. We review the core components of a VCT: computational phantoms, simulators of different imaging modalities, and interpretation models. We also highlight some of the applications of VCTs across various imaging modalities.

Spherical harmonic (SH) interpolation is a commonly used method to spatially up-sample sparse head related transfer function (HRTF) datasets to denser HRTF datasets. However, depending on the number of sparse HRTF measurements and SH order, this process can introduce distortions into high frequency representations of the HRTFs. This paper investigates whether it is possible to restore some of the distorted high frequency HRTF components using machine learning algorithms. A combination of convolutional auto-encoder (CAE) and denoising auto-encoder (DAE) models is proposed to restore the high frequency distortion in SH-interpolated HRTFs. Results were evaluated using both perceptual spectral difference (PSD) and localisation prediction models, both of which demonstrated significant improvement after the restoration process.

Objective: Dose optimization and pharmacokinetic evaluation of α-particle emitting radium-223 dichloride (223RaCl2) by planar γ-camera or single photon emission computed tomography (SPECT) imaging are hampered by the low photon abundance and injected activities. In this study, we demonstrate SPECT of 223Ra using phantoms and small animal in vivo models. Methods: Line phantoms and mice bearing 223Ra were imaged using a dedicated small animal SPECT by detecting the low-energy photon emissions from 223Ra. Localization of the therapeutic agent was verified by whole-body and whole-limb autoradiography and its radiobiological effect confirmed by immunofluorescence. Results: A state-of-the-art commercial small animal SPECT system equipped with a highly sensitive collimator enables collection of sufficient counts for three-dimensional reconstruction at reasonable administered activities and acquisition times. Line sources of 223Ra in both air and in a water scattering phantom gave a line spread function with a full-width-at-half-maximum of 1.45 mm. Early and late-phase imaging of the pharmacokinetics of the radiopharmaceutical were captured. Uptake at sites of active bone remodeling was correlated with DNA damage from the α particle emissions. Conclusions: This work demonstrates the capability to noninvasively define the distribution of 223RaCl2, a recently approved α-particle-emitting radionuclide. This approach allows quantitative assessment of 223Ra distribution and may assist radiation-dose optimization strategies to improve therapeutic response and ultimately to enable personalized treatment planning.

305 Objectives: We have developed a new set of cardiac motion vector field (CMVF) estimation, analysis and display methods from 4D cardiac-gated (CG) myocardial perfusion (MP) PET images. The goal of this work is to evaluate the methods using 4D CG-MP PET images obtained from different reconstruction and analysis methods from patients with normal and known cardiac motion (CM) abnormalities. We acquired list-mode 4D CG-MP PET data from six patients with normal global ejection fractions (EF) and three patients with known CM abnormalities. The datasets were reconstructed using the standard image reconstruction provided by the vendor and with a new, in-house developed 4D image reconstruction method with respiratory motion (RM) and CM compensation. Specifically, the RM estimation and compensation were applied based on an equal count-based, data-driven gating method. The RM compensated data were reconstructed into CG PET images with a given number of equal time intervals, e.g., 8, 16 and 24, over each cardiac cycle. A reference frame PET image at the center of the heart cycle was then generated from the smoothed CG images using the Groupwise registration method, and the reference frame was transformed back to the individual CG images with the corresponding set of CMVF estimates. The final CM-compensated images showed significantly improved image resolution and lower image noise fluctuations as compared to the CG images obtained from the vendor method. The CMVF estimation is based on the traditional optical flow method for motion estimation and was applied to the 4D CG-MP PET images to estimate the CMVF between selected cardiac frames. The entire heart including both the left and right ventricles was used in the CM estimation. The radial, longitudinal and tangential components of the 3D CMVF over the left ventricle was grouped and displayed in a 13-segment and the standard 17-segment polar plots between each adjacent CG frames over the cardiac cycle. The data were further analyzed and the three components of the 3D CMVF in each segment were averaged and plotted as a function of time over the cardiac cycle. In general, the 3D CMVF estimates from the 4D CG-MP PET images with RM & CM motion compensation show more consistent patterns with less variations over time than those from the images without the compensations. The radial and longitudinal components also show more consistent changes between segments than the tangential component. Further, the 13-segment results show a general CM pattern and the 17-segment results show the CM in more detail, without the effect of noise variations. In patients with known CM abnormalities, the 3D CMVF shows asynchronized beating heart motion in certain segments, which is consistent with clinical findings. In typical normal patients, the components of the 3D CMVF demonstrate synchronized beating heart motion among all segments. In one patient with normal global EF, our results indicate a small but noticeable asynchronized motion among segments along the septal wall. A re-examination of the 4D CG-MP PET images confirms the finding. The developed 3D CMVF estimation, may hold promise for accurate characterization of CM and detection of CM abnormalities from 4D CG-MP PET images. The improved image quality obtained with RM and CM compensation provide more accurate estimations than quality obtained without compensation. The additional quantitative ‘cardiomics’ information of the left ventricular motion, in conjunction with perfusion information, from the same 4D CG-MP PET images may add additional important diagnostic information in subclinical and clinical manifest cardiac diseases.

We evaluated the ability of conventional and corrective image reconstruction methods to detect a myocardial perfusion defect in single-photon emission computed tomography (SPECT) images using the channelized hotelling observer (CHO). Using our previously developed 4-D extended cardiac-torso, we simulated realistic transmural and endocardial perfusion defects in various locations and sizes in the myocardium. Almost noise-free projection datasets were generated separately from the heart, blood pool, lungs, liver, kidneys, stomach, gall bladder, and the remaining body using Monte Carlo simulation techniques that included the effects of collimator detector response, photon attenuation, and scatter. The datasets were then scaled and combined to model 99mTc Sestamibi myocardial perfusion SPECT projection datasets from a typical patient with various reduced count levels, simulating reduced radiation doses. The final projection data were reconstructed using the 3-D filtered back-projection (FBP) without correction and a 3-D ordered-subset expectation-maximization (OS-EM) method with corrections of attenuation, collimator-detector response, and scatter (ADS), followed by smoothing filtering with several different cut-off frequencies. Task-based evaluations on the reconstructed images were performed by using the CHO followed by the receiver operating characteristics (ROC) methodology to evaluate the detectability of a myocardial perfusion defect by the two image reconstruction methods for various defect anatomies. Areas under the ROC curve (AUC) were computed to assess the changes in the detection of myocardial perfusion defects. The results showed that the 3-D OS-EM with ADS correction gave overall higher AUC values than FBP at several post smoothing levels, count levels, and myocardial perfusion defect sizes. The difference in AUC increased as image smoothness decreased, where the 3-D OS-EM with correction was able to provide similar AUC values with 20%–40% reduction in count levels compared to those of FBP. The AUC values for smaller myocardial perfusion defects were lower for both reconstruction methods with smaller differences. We concluded that the 3-D OS-EM with ADS correction provides higher performance in detecting myocardial perfusion defects. At the conventional count level, it allows for less counts or lower radiation doses without loss of defect detection in myocardial perfusion SPECT images compared to the conventional FBP method, especially regarding less smoothed images.

Background Pulmonary hypertension (PH) is a known complication of HCM and is a strong predictor of mortality. We aim to investigate the relationship between microvascular dysfunction measured by quantitative PET and PH in HCM patients. Methods Eighty-nine symptomatic HCM patients were included in the study. Each patient underwent two 20-min 13N-NH3 dynamic PET scans for rest and stress conditions, respectively. A 2-tissue irreversible compartmental model was used to fit the segments time activity curves for estimating segmental and global myocardial blood flow (MBF) and myocardial flow reserve (MFR). Echocardiographic derived PASP was utilized to estimate PH. Results Patients were categorized into two groups across PASP: PH (PASP > 36 mmHg) and no-PH (PASP ≤ 36 mmHg). patients with PH had larger left atrium, ratio of higher inflow early diastole (E) and atrial contraction (A) waves, E/A, and ratio of inflow and peak early diastolic waves, E/e', significantly reduced global stress MBF (1.85 ± 0.52 vs. 2.13 ± 0.56 ml/min/g; p = 0.024) and MFR (2.21 ± 0.57 vs. 2.62 ± 0.75; p = 0.005), while the MBFs at rest between the two groups were similar. There were significant negative correlations between global stress MBF/MFR and PASP (stress MBF: r = -0.23, p = 0.03; MFR: r = -0.32, p = 0.002); for regional MBF and MFR measurements, the highest linear correlation coefficients were observed in the septal wall (stress MBF: r = -0.27, p = 0.01; MFR: r = -0.31, p = 0.003). Global MFR was identified to be independent predictor for PH in multivariate regression analysis. Conclusion Echocardiography-derived PASP is negatively correlated with global MFR measured by 13N-NH3 dynamic PET. Global MFR is suggested to be an index of PH in HCM patients.

404 Objectives: The goal is to develop an effective transfer learning (TL) dataset using realistic simulation techniques for the highest possible performance of a deep learning (DL) model in automatic detection of myocardial perfusion (MP) defect in MP single-photon emission computed tomography (SPECT) images. The hypothesis is the use of realistic simulated MP SPECT images with known ‘truth’ in a TL dataset to supplement the limited number of clinical data in training a DL model will enhance its accuracy in detecting MP defects. An upgraded perfusion 4D eXtended CArdiac-Torso (XCAT) phantom, or PXCAT, that included a model of computer-generated coronary artery was used in the study. Models of realistic stenosis-driven MP defect were generated by simulating stenoses with different severities, physiological heterogeneities, and inter-ventricular and inter-layer variations at various coronary arterial segments. To simulate ensembles of patients with anatomical variation, 3 major single vessel diseases (LAD, LCX, RCA) with 3 different degrees of stenosis (50%, 70%, 90%) at 3 different stenosis locations (proximal, distal, diagonal) were simulated. Double vessel diseases were also simulated using the similar combinations as for single vessel disease. Finally, for each MP defect model, SPECT projection datasets of the PXCAT phantoms were generated with combinations of 5 count levels (125%, 100%, 75%, 50%, and 25% of clinical count rate) with multiple noise realizations. The simulated MP SPECT images obtained from the reconstruction of the projection datasets were used to prepare TL dataset with known location and severity about the MP defect status. A classification method of the MP defect images was performed based on the TL technique using a pre-trained deep learning model, Inception V3 from Google. Several methods were used to optimize the DL model for its higher performance including, changing the number and quality of training images, altering the training options, etc. Eighty percent and 10% of the images were used for the main training and validation during the training, respectively, and then a final 10% were used for testing the performance of the DL model in terms of detection accuracy of the MP defects. The results showed that the performance of the DL model depends on the complexity of disease status and acquired count level. That is images of single vessel disease had higher classification accuracies than those of double vessel disease or lower count levels (e.g., such as 25% of clinical count rates). We conclude that realistically simulated clinical images with known ‘truth’ used in the TL datasets was successfully applied to a DL model, and shows potential to address the major difficulty in current application of DL technology to biomedical imaging in acquiring sufficiently large number of clinical datasets with accurately known disease status.

PURPOSEThe goal of this study was to develop and evaluate four post-reconstruction respiratory and cardiac (R&C) motion vector field (MVF) estimation methods for cardiac 4D PET data.METHODIn Method 1, the dual R&C motions were estimated directly from the dual R&C gated images. In Method 2, respiratory motion (RM) and cardiac motion (CM) were separately estimated from the respiratory gated only and cardiac gated only images. The effects of RM on CM estimation were modeled in Method 3 by applying an image-based RM correction on the cardiac gated images before CM estimation, the effects of CM on RM estimation were neglected. Method 4 iteratively models the mutual effects of RM and CM during dual R&C motion estimations. Realistic simulation data were generated for quantitative evaluation of four methods. Almost noise-free PET projection data were generated from the 4D XCAT phantom with realistic R&C MVF using Monte Carlo simulation. Poisson noise was added to the scaled projection data to generate additional datasets of two more different noise levels. All the projection data were reconstructed using a 4D image reconstruction method to obtain dual R&C gated images. The four dual R&C MVF estimation methods were applied to the dual R&C gated images and the accuracy of motion estimation was quantitatively evaluated using the root mean square error (RMSE) of the estimated MVFs.RESULTSResults show that among the four estimation methods, Methods 2 performed the worst for noise-free case while Method 1 performed the worst for noisy cases in terms of quantitative accuracy of the estimated MVF. Methods 4 and 3 showed comparable results and achieved RMSE lower by up to 35% than that in Method 1 for noisy cases.CONCLUSIONIn conclusion, we have developed and evaluated 4 different post-reconstruction R&C MVF estimation methods for use in 4D PET imaging. Comparison of the performance of four methods on simulated data indicates separate R&C estimation with modeling of RM before CM estimation (Method 3) to be the best option for accurate estimation of dual R&C motion in clinical situation.

Hyaluronic acid (HA) is found naturally in synovial fluid and is utilized therapeutically to treat osteoarthritis (OA). Here, we employed a peptide-polymer cartilage coating platform to localize HA to the cartilage surface for the purpose of treating post traumatic osteoarthritis. The objective of this study was to increase efficacy of the peptide-polymer platform in reducing OA progression in a mouse model of post-traumatic OA without exogenous HA supplementation. The peptide-polymer is composed of an HA-binding peptide (HABP) conjugated to a heterobifunctional poly (ethylene glycol) (PEG) chain and a collagen binding peptide (COLBP). We created a library of different peptide-polymers and characterized their HA binding properties in vitro using quartz crystal microbalance (QCM-D) and isothermal calorimetry (ITC). The peptide polymers were further tested in vivo in an anterior cruciate ligament transection (ACLT) murine model of post traumatic OA. The peptide-polymer with the highest affinity to HA as tested by QCM-D (∼4-fold greater binding compared to other peptides tested) and by ITC (∼3.8-fold) was HABP2-8-arm PEG-COLBP. Biotin tagging demonstrated that HABP2-8-arm PEG-COLBP localizes to both cartilage defects and synovium. In vivo, HABP2-8-arm PEG-COLBP treatment and the clinical HA comparator Orthovisc lowered levels of inflammatory genes including IL-6, IL-1B, and MMP13 compared to saline treated animals and increased aggrecan expression in young mice. HABP2-8-arm PEG-COLBP and Orthovisc also reduced pain as measured by incapacitance and hotplate testing. Cartilage degeneration as measured by OARSI scoring was also reduced by HABP2-8-arm PEG-COLBP and Orthovisc. In aged mice, HABP2-8-arm PEG-COLBP therapeutic efficacy was similar to its efficacy in young mice, but Orthovisc was less efficacious and did not significantly improve OARSI scoring. These results demonstrate that HABP2-8-arm PEG-COLBP is effective at reducing PTOA progression.

We present a direct (noniterative) algorithm for 1-D quadratic data fitting with neighboring intensity differences penalized by the Huber function. Applications of such an algorithm include 1-D processing of medical signals, such as smoothing of tissue time concentration curves in kinetic data analysis or sinogram preprocessing, and using it as a subproblem solver for 2-D or 3-D image restoration and reconstruction. dynamic programming was used to develop the direct algorithm. The problem was reformulated as a sequence of univariate optimization problems, for , where is the number of data points. The solution to the univariate problem at index is parameterized by the solution at , except at . Solving the univariate optimization problem at yields the solution to each problem in the sequence using back-tracking. Computational issues and memory cost are discussed in detail. Two numerical studies, tissue concentration curve smoothing and sinogram preprocessing for image reconstruction, are used to validate the direct algorithm and illustrate its practical applications. In the example of 1-D curve smoothing, the efficiency of the direct algorithm is compared with four iterative methods: the iterative coordinate descent, Nesterov's accelerated gradient descent algorithm, FISTA, and an off-the-shelf second order method. The first two methods were applied to the primal problem, the others to the dual problem. The comparisons show that the direct algorithm outperforms all other methods by a significant factor, which rapidly grows with the curvature of the Huber function. The second example, sinogram preprocessing, showed that robustness and speed of the direct algorithm are maintained over a wide range of signal variations, and that noise and streaking artifacts could be reduced with almost no increase in computation time. We also outline how the proposed 1-D solver can be used for imaging applications.

14 Objectives: We have previously developed respiratory and cardiac motion compensation methods that significantly improve both the resolution and noise of 4D cardiac-gated (CG) myocardial perfusion (MP) PET images. The aim of this work is to investigate the estimation of cardiac motion vector field (CMVF) information from the improved 4D CG MP PET images. We acquired the 4D CG MP PET data in listmode. The previously developed respiratory motion (RM) estimation and compensation were applied based on an equal count-based data-driven gating Methods: The RM compensated data were reconstructed into CG PET images with a given number of equal time intervals, e.g., 8, 16 and 24, over each cardiac cycle. A reference frame PET image at the center of the heart cycle was generated from the smoothed CG images using the Group-wise B-spline non-rigid image-based registration method. The reference frame was transformed back to the individual CG images with the corresponding set of CMVF estimates. Parameters of the Group-wise registration method was selected based on a simulation study using the 4D XCAT phantom for minimum motion artifact. The final CM compensated images showed significantly improved image resolution and noise compared to the original CG images. The processed 4D CG MP PET images were used to estimate the final CMVF between selected cardiac frames. The entire heart including both the left and right myocardium is used in CM estimation. In an IRB approved patient study, we have processed the acquired listmode data of 4D CG MP PET studies from five patients that are considered normal as the time of submission. We compared the CMVF estimates between the selected cardiac frames using the clinical images, and the RM and CM compensated datasets with 8, 16 and 24 CG frames. The results show consistent CMVF patterns from the RM and CM compensated datasets with 8, 16 and 24 CG frames, indicating normal beating heart motion. They showed marked differences when compared to the CMVF derived from the clinical images without RM and CM compensation. Further study is underway to determine the accuracy of the derived CMVF using realistic simulated CG MP PET data from the 4D XCAT phantom with known CMVF between the CG frames. We conclude the application of RM and CM compensation results in much improved 4D CG MP PET images in terms of both image resolution and noise. The processed data can potentially provide improved visualization and more accurate quantitative analysis of the beating heart. The additional quantitative cardiomics information of the beating heart may result in early detection and diagnosis of cardiac motion abnormalities for improved patient care.

Objective: Dose optimization and pharmacokinetic evaluation of alpha emitting Radium-223 dichloride ( 223 RaCl 2 ) by planar gamma camera or single photon emission computed tomographic (SPECT) imaging are hampered by the low photon abundance and injection activities. Here, we demonstrate SPECT of 223 Ra using phantoms and small animal in vivo models. Methods: Line phantoms and mice bearing 223 Ra were imaged using a next generation dedicated small animal SPECT by detecting the low energy photon emissions from 223 Ra. Localization of the therapeutic agent was verified by whole body and whole limb autoradiography and its effect determined by immunofluorescence. Results: A state-of-the-art commercial small animal SPECT system equipped with a highly sensitive collimator enables collection of sufficient counts for three-dimensional reconstruction. Line sources of 223 Ra in both air and in a water scattering phantom gave linear response functions with provide full-width-at-half-maximum of 1.45 mm. Early and late phase imaging of the pharmacokinetics of the radiopharmaceutical were captured. Uptake at sites of active bone remodeling were correlated with DNA damage from the alpha particle emissions. Conclusions: This work demonstrates the capability to noninvasively define the distribution of 223 Ra, a recently approved alpha emitting radionuclide. This approach allows quantitative assessment of 223 Ra distribution and may provide radiation dose optimization strategies to improve therapeutic response and ultimately to enable personalized treatment planning.

Recently there has been an increase of clinical cases of osteoporosis due to an increase in the ageing population. As an established diagnostic tool for osteoporosis, dual-energy x-ray absorptiometry (DEXA) is widely used due to its high accuracy, precision and image quality. In recent years, a new type of silicon-based photomultipliers, SiPMs, has been developed which has high potential to be used in photon counting DEXA detectors for its compact size and reliable performance. In this work, we designed a DEXA test platform with a SiPM and evaluated its performance using various bone density imaging parameters. The experimental DEXA platform is constructed with a fan-beam geometry consisting of an x-ray source from Spellman and a self-developed scanning motion control system. We studied the performance characteristics of a 64-channel YSO/SiPM array detector and compared them with a cadmium zinc telluride (CZT) detector. The maximum-likelihood (ML) estimation method was used to optimize the surface fitting algorithm to achieve high material decomposition accuracy. The results show the average error of the YSO/SiPM detector is 1.24% for PMMA and 1.72% for Al. We conclude the YSO/SiPM detector can achieve material decomposition accuracy close to CZT detector and it has the potential in DEXA imaging with good imaging performance characteristics at low cost.

Over the past decades, significant improvements have been made in the field of computational human phantoms (CHPs) and their applications in biomedical engineering. Their sophistication has dramatically increased. The very first CHPs were composed of simple geometric volumes, e.g., cylinders and spheres, while current CHPs have a high resolution, cover a substantial range of the patient population, have high anatomical accuracy, are poseable, morphable, and are augmented with various details to perform functionalized computations. Advances in imaging techniques and semiautomated segmentation tools allow fast and personalized development of CHPs. These advances open the door to quickly develop personalized CHPs, inherently including the disease of the patient. Because many of these CHPs are increasingly providing data for regulatory submissions of various medical devices, the validity, anatomical accuracy, and availability to cover the entire patient population is of utmost importance. This paper is organized into two main sections: the first section reviews the different modeling techniques used to create CHPs, whereas the second section discusses various applications of CHPs in biomedical engineering. Each topic gives an overview, a brief history, recent developments, and an outlook into the future.

We have significantly improved four data-driven respiratory motion (RM) extraction methods for various activity distributions in clinical myocardial perfusion (MP) and 18 F-FDG PET datasets. They are activity distributions: (1) with high myocardial uptake, (2) same as (1) but with portion of the heart outside the image, and with high image intensity (3) in the liver and (4) in the lung area without attenuation compensation. In Method #1, a 3D volume-of-interest (VOI) was placed over the heart region of the PET image obtained from the total acquisition time period. The surrogate RM signals were obtained from the centroids of the image intensity of the myocardial activity uptake within the same VOI of PET images obtained from rebinned list-mode data in short time intervals. The Fourier Transform (FT) of the time sequence of surrogate RM signals and smoothing reveal the RM peak and its average period, Pav. In Methods #2, #3, and #4, specially-shaped 3D VOIs were placed over the heart, the top of the liver, and the bottom of the lungs, respectively. Then, the same procedures used in Method #1 were employed except using the total counts within the corresponding VOI. The location, sizes and shapes of the VOIs were optimized for the highest signal-to-noise (S/N) in the RM peak extraction. The improved RM extraction methods were evaluated using 14 patient datasets. Method #1 was shown to work well for 79% of the datasets, and Pav showing high S/N and excellent agreement (Pearson correlation coefficient 0.997) with those obtained from an external RM monitoring belt system. Method #2 was applied successfully to 14%, and Methods #3 and #4 to the rest of datasets. Excellent agreements were also found in cross comparison between the methods. We conclude that the improved data-driven RM extraction methods which showed successful results in various PET image datasets will provide an important first step for the motion compensation application in commercial PET scanners.