In secondary hyperparathyroidism (SHPT) disease, preoperatively localizing hyperplastic parathyroid glands is crucial in the surgical procedure. These glands can be detected via the dual-modality imaging technique single-photon emission computed tomography/computed tomography (SPECT/CT) since it has high sensitivity and provides an accurate location. However, due to possible low-uptake glands in SPECT images, manually labeling glands is challenging, not to mention automatic label methods. In this work, we present a deep learning method with a novel fusion network to detect hyperplastic parathyroid glands in SPECT/CT images. Our proposed fusion network follows the convolutional neural network (CNN) with a three-pathway architecture that extracts modality-specific feature maps. The fusion network, composed of the channel attention module, the feature selection module, and the modality-specific spatial attention module, is designed to integrate complementary anatomical and functional information, especially for low-uptake glands. Experiments with patient data show that our fusion method improves performance in discerning low-uptake glands compared with current fusion strategies, achieving an average sensitivity of 0.822. Our results prove the effectiveness of the three-pathway architecture with our proposed fusion network for solving the glands detection task. To our knowledge, this is the first study to detect abnormal parathyroid glands in SHPT disease using SPECT/CT images, which promotes the application of preoperative glands localization.

Dynamic PET imaging provides superior physiological information than conventional static PET imaging. However, the dynamic information is gained at the cost of a long scanning protocol; this limits the clinical application of dynamic PET imaging. We developed a modified Logan reference plot model to shorten the acquisition procedure in dynamic PET imaging by omitting the early-time information necessary for the conventional reference Logan model. The proposed model is accurate theoretically, but the straightforward approach raises the sampling problem in implementation and results in noisy parametric images. We then designed a self-supervised convolutional neural network to increase the noise performance of parametric imaging, with dynamic images of only a single subject for training. The proposed method was validated via simulated and real dynamic [Formula: see text]-fallypride PET data. Results showed that it accurately estimated the distribution volume ratio (DVR) in dynamic PET with a shortened scanning protocol, e.g., 20 minutes, where the estimations were comparable with those obtained from a standard dynamic PET study of 120 minutes of acquisition. Further comparisons illustrated that our method outperformed the shortened Logan model implemented with Gaussian filtering, regularization, BM4D and the 4D deep image prior methods in terms of the trade-off between bias and variance. Since the proposed method uses data acquired in a short period of time upon the equilibrium, it has the potential to add clinical values by providing both DVR and Standard Uptake Value (SUV) simultaneously. It thus promotes clinical applications of dynamic PET studies when neuronal receptor functions are studied.

Positron Emission Tomography (PET) has become a preferred imaging modality for cancer diagnosis, radiotherapy planning, and treatment responses monitoring. Accurate and automatic tumor segmentation is the fundamental requirement for these clinical applications. Deep convolutional neural networks have become the state-of-the-art in PET tumor segmentation. The normalization process is one of the key components for accelerating network training and improving the performance of the network. However, existing normalization methods either introduce batch noise into the instance PET image by calculating statistics on batch level or introduce background noise into every single pixel by sharing the same learnable parameters spatially. In this paper, we proposed an attentive transformation (AT)-based normalization method for PET tumor segmentation. We exploit the distinguishability of breast tumor in PET images and dynamically generate dedicated and pixel-dependent learnable parameters in normalization via the transformation on a combination of channel-wise and spatial-wise attentive responses. The attentive learnable parameters allow to re-calibrate features pixel-by-pixel to focus on the high-uptake area while attenuating the background noise of PET images. Our experimental results on two real clinical datasets show that the AT-based normalization method improves breast tumor segmentation performance when compared with the existing normalization methods.

68Ga-DOTATATE PET-CT is routinely used for imaging neuroendocrine tumor (NET) somatostatin receptor subtype 2 (SSTR2) density in patients, and is complementary to FDG PET-CT for improving the accuracy of NET detection, characterization, grading, staging, and predicting/monitoring NET responses to treatment. Performing sequential18F-FDG and68Ga-DOTATATE PET scans would require 2 or more days and can delay patient care. To align temporal and spatial measurements of18F-FDG and68Ga-DOTATATE PET, and to reduce scan time and CT radiation exposure to patients, we propose a single-imaging session dual-tracer dynamic PET acquisition protocol in the study. A recurrent extreme gradient boosting (rXGBoost) machine learning algorithm was proposed to separate the mixed18F-FDG and68Ga-DOTATATE time activity curves (TACs) for the region of interest (ROI) based quantification with tracer kinetic modeling. A conventional parallel multi-tracer compartment modeling method was also implemented for reference. Single-scan dual-tracer dynamic PET was simulated from 12 NET patient studies with18F-FDG and68Ga-DOTATATE 45-min dynamic PET scans separately obtained within 2 days. Our experimental results suggested an18F-FDG injection first followed by68Ga-DOTATATE with a minimum 5 min delayed injection protocol for the separation of mixed18F-FDG and68Ga-DOTATATE TACs using rXGBoost algorithm followed by tracer kinetic modeling is highly feasible.

Tensor fields are useful for modeling the structure of biological tissues. The challenge to measure tensor fields involves acquiring sufficient data of scalar measurements that are physically achievable and reconstructing tensors from as few projections as possible for efficient applications in medical imaging. In this paper, we present a filtered back-projection algorithm for the reconstruction of a symmetric second-rank tensor field from directional X-ray projections about three axes. The tensor field is decomposed into a solenoidal and irrotational component, each of three unknowns. Using the Fourier projection theorem, a filtered back-projection algorithm is derived to reconstruct the solenoidal and irrotational components from projections acquired around three axes. A simple illustrative phantom consisting of two spherical shells and a 3D digital cardiac diffusion image obtained from diffusion tensor MRI of an excised human heart are used to simulate directional X-ray projections. The simulations validate the mathematical derivations and demonstrate reasonable noise properties of the algorithm. The decomposition of the tensor field into solenoidal and irrotational components provides insight into the development of algorithms for reconstructing tensor fields with sufficient samples in terms of the type of directional projections and the necessary orbits for the acquisition of the projections of the tensor field.

Convolutional neural networks (CNNs) have been widely used in single image super-resolution (SR) and obtained remarkable performance. However, most CNN-based SR models require heavy computation, which limits their real-world applications. In this paper, we address the computation problem of SR by network binarization, which converts the full-precision network into the binary network, thus intensively reducing computation. We propose the pixel-correlation distillation for SR network binarization, which distills the knowledge of pixel relationship from the original full-precision network to the binary network. In addition, we further reduce the quantization errors of the binary network by introducing trainable scaling factors to replace the fixed scaling factors in most existing binarization methods. We carry out extensive experiments on SRResNet [1] and VDSR [2], which are two commonly used SR networks. It is shown that the proposed method generates more visually pleasing SR images, and consistently outperforms other state-of-the-art methods in PSNR and SSIM.

Deep convolutional neural networks (DCNNs) have achieved state-of-the-art performance in classification, natural language processing (NLP), and regression tasks. However, there is still a great gap between DCNNs and the human brain in terms of computation efficiency. Inspired by neural synaptic plasticity and stochastic computing (SC), we propose neural synaptic plasticity-inspired computing (NSPC) to simulate the human brain's neural network activity for inference tasks with simple logic gates. The multiplication and accumulation (MAC) is transformed by the wire connectivity in NSPC, which only requires bundles of wires and small width adders. To this end, the NSPC imitates the structure of neural synaptic plasticity from a circuit wires connection perspective. Furthermore, from the principle of NSPC, we use a data mapping method to convert the convolution operations to matrix multiplications. Based on the methodology of NSPC, fully-pipelined and low latency architecture is designed. The proposed NSPC accelerator exhibits high hardware efficiency while maintaining a comparable network accuracy level. The NSPC based DCNN accelerator (NSPC-CNN) processes DCNN at 1.5625M images/s with a power dissipation of 15.42 W and an area of 36.4 mm2. The NSPC based deep neural network (DNN) accelerator (NSPC-DNN) that implements three fully connected layers DNN consumes only 6.6 mm2 area and 2.93 W power, and achieves a throughput of 400M images/s. Compared with conventional fixed-point implementations, the NSPC-CNN achieves 2.77× area efficiency, 2.25× power efficiency; the proposed NSPC-DNN exhibits 2.31× area efficiency and 2.09× power efficiency.

The glomerular filtration rate (GFR) is a crucial index to measure renal function. In daily clinical practice, the GFR can be estimated using the Gates method, which requires the clinicians to define the region of interest (ROI) for the kidney and the corresponding background in dynamic renal scintigraphy. The manual placement of ROIs to estimate the GFR is subjective and labor-intensive, however, making it an undesirable and unreliable process. This work presents a fully automated ROI detection method to achieve accurate and robust GFR estimations. After image preprocessing, the ROI for each kidney was delineated using a shape prior constrained level set (spLS) algorithm and then the corresponding background ROIs were obtained according to the defined kidney ROIs. In computer simulations, the spLS method had the best performance in kidney ROI detection compared with the previous threshold method (threshold) and the Chan-Vese level set (cvLS) method. In further clinical applications, 223 sets of 99mTc-diethylenetriaminepentaacetic acid renal scintigraphic images from patients with abnormal renal function were reviewed. Compared with the former ROI detection methods (threshold and cvLS), the GFR estimations based on the ROIs derived by the spLS method had the highest consistency and correlations (r = 0.98, p <; 0.001) with the reference estimated by experienced physicians. The results indicate that the proposed automated ROI detection method has great potential in automated ROI detection for accurate and robust GFR estimation in dynamic renal scintigraphy.

Saliency detection on images has experienced substantial progress in recent years on the basis of deep neural network (DNN). However, there may exist secondary saliency in the background that distracts DNN learning and mistakes the secondary salient regions as saliency. To address this issue, we propose a dual-term energy to improve the inference of saliency on top of DNN estimation, where dense term smoothens salient regions in pixel scale and sparse term extracts prior knowledge to differentiate saliency and non-saliency superpixels. Our prior knowledge including extra- and intra-region priors, contributes to improving overall saliency detection. The extra-region prior knowledge estimates the saliency probabilities for different pre-partitioned regions to eliminate the secondary saliency. The intra-region prior knowledge helps to group the salient regions that otherwise could be ignored by DNN predictor, and thus to provide more complete saliency definition. We evaluated our model on 8,465 images from four well-recognized saliency detection benchmarking datasets, and compared our model to six state-of-the-art comparative methods. Experimental results demonstrated that our model outperformed the state-of-the-art counterpart with improvements of up to 2.51% in terms of F-measure.

Increasing the image quality of positron emission tomography (PET) is an essential topic in the PET community. For instance, thin-pixelated crystals have been used to provide high spatial resolution images but at the cost of sensitivity and manufacture expense. In this paper, we proposed an approach to enhance the PET image resolution and noise property for PET scanners with large pixelated crystals. To address the problem of coarse blurred sinograms with large parallax errors associated with large crystals, we developed a data-driven, single-image super-resolution (SISR) method for sinograms, based on the novel deep residual convolutional neural network (CNN). Unlike the CNN-based SISR on natural images, periodically padded sinogram data and dedicated network architecture were used to make it more efficient for PET imaging. Moreover, we included the transfer learning scheme in the approach to process cases with poor labeling and small training data set. The approach was validated via analytically simulated data (with and without noise), Monte Carlo simulated data, and pre-clinical data. Using the proposed method, we could achieve comparable image resolution and better noise property with large crystals of bin sizes of thin crystals with a bin size from to . Our approach uses external PET data as the prior knowledge for training and does not require additional information during inference. Meanwhile, the method can be added into the normal PET imaging framework seamlessly, thus potentially finds its application in designing low-cost high-performance PET systems.

An algorithm was developed for the 2-D reconstruction of truncated and nontruncated uniformly attenuated data acquired from single photon emission computed tomography (SPECT). The algorithm is able to reconstruct data from half-scan (180 ( degrees ) ) and short-scan (180 ( degrees ) +fan angle) acquisitions for parallel- and fan-beam geometries, respectively, as well as data from full-scan (360 ( degrees )) acquisitions. The algorithm is a derivative, backprojection, and Hilbert transform (DBH) method, which involves the backprojection of differentiated projection data followed by an inversion of the finite weighted Hilbert transform. The kernel of the inverse weighted Hilbert transform is solved numerically using matrix inversion. Numerical simulations confirm that the DBH method provides accurate reconstructions from half-scan and short-scan data, even when there is truncation. However, as the attenuation increases, finer data sampling is required.