Single photon emission computed tomography (SPECT) is based on the measurement of radiation emitted by a radiotracer injected into the patient. Because of photoelectric absorption and Compton scatter, the gamma photons are attenuated inside the body before arriving at the detector. A quantitative reconstruction must consider the attenuation, which is usually nonuniform. Novikov has derived an explicit inversion formula for the nonuniformly attenuated Radon transform for SPECT reconstruction of parallel-beam collimated projections. In this paper, we extend his research to variable focal-length fan-beam VFF collimator geometry. A ray-driven analytical formula for VFF reconstruction with nonuniform attenuation was derived. As a unified framework, this formula can be used for parallel-beam, fan-beam, and VFF collimators. Its accuracy is demonstrated by computer simulation experiments.
This paper investigates an accurate reconstruction method to invert the attenuated Radon transform in nonparallel beam (NPB) geometries. The reconstruction method contains three major steps: 1) performing one-dimensional phase-shift rebinning; 2) implementing nonuniform Hilbert transform; and 3) applying Novikov's explicit inversion formula. The method seems to be adaptive to different settings of fan-beam geometry from very long to very short focal lengths without sacrificing reconstruction accuracy. Compared to the conventional bilinear rebinning technique, the presented method showed a better spatial resolution, as measured by modulation transfer function. Numerical experiments demonstrated its computational efficiency and stability to different levels of Poisson noise. Even with complicated geometries such as varying focal-length and asymmetrical fan-beam collimation, the presented method achieved nearly the same reconstruction quality of parallel-beam geometry. This effort can facilitate quantitative reconstruction of single photon emission computed tomography for cardiac imaging, which may need NPB collimation geometries and require high computational efficiency.
SPECT is one of the nuclear medicine imaging techniques and widely used in the clinical applications. Different from CT, SPECT achieved the functional image of the organ of interest, and the diseases can be found much earlier. Conebeam SPECT reconstruction can improve the photon density and spatial resolution of the reconstructed image, but it is time consuming. In clinic, doctors usually just care about the region of interest (ROI), such as heart, not whole body. Local reconstruction can reduce the reconstruction time. In this paper, based on Novikov's analytical SPECT reconstruction algorithm, we built a framework for local cone-beam SPECT reconstruction with non-uniform attenuation. The simulation results show our reconstruction framework is feasible.
It is still a challenge to differentiate space-occupying brain lesions such as tumefactive demyelinating lesions (TDLs), tumefactive primary angiitis of the central nervous system (TPACNS), primary central nervous system lymphoma (PCNSL), and brain gliomas. Convolutional neural networks (CNNs) have been used to analyze complex medical data and have proven transformative for image-based applications. It can quickly acquire diseases' radiographic features and correct doctors' diagnostic bias to improve diagnostic efficiency and accuracy. The study aimed to assess the value of CNN-based deep learning model in the differential diagnosis of space-occupying brain diseases on MRI.We retrospectively analyzed clinical and MRI data from 480 patients with TDLs (n = 116), TPACNS (n = 64), PCNSL (n = 150), and brain gliomas (n = 150). The patients were randomly assigned to training (n = 240), testing (n = 73), calibration (n = 96), and validation (n = 71) groups. And a CNN-implemented deep learning model guided by clinical experts was developed to identify the contrast-enhanced T1-weighted sequence lesions of these four diseases. We utilized accuracy, sensitivity, specificity, and area under the curve (AUC) to evaluate the performance of the CNN model. The model's performance was then compared to the neuroradiologists' diagnosis.The CNN model had a total accuracy of 87% which was higher than senior neuroradiologists (74%), and the AUC of TDLs, PCNSL, TPACNS and gliomas were 0.92, 0.92, 0.89 and 0.88, respectively.The CNN model can accurately identify specific radiographic features of TDLs, TPACNS, PCNSL, and gliomas. It has the potential to be an effective auxiliary diagnostic tool in the clinic, assisting inexperienced clinicians in reducing diagnostic bias and improving diagnostic efficiency.
Computed tomography is an advanced imaging technology. It determines the attenuation coefficient distribution of the measured object based on known projection data and reconstruction algorithms. The filtered back-projection algorithm is one of the traditional algorithms commonly used in image reconstruction. It can quickly obtain the image to be reconstructed. The quality of the reconstructed image relies on the selection of the filter designed manually. Besides, the image reconstructed will contain serious artifacts when the projection data is incomplete. In this paper, a fully connected layer is used to simulate the filtered back-projection algorithm by designing a neural network model, it can adaptively learn a suitable filter for the current projection system. Meanwhile, the artifacts generated in the case of limited-angle reconstruction are corrected through the convolution layer. The results indicate that the method proposed in this paper can obtain the image with better quality compared to the traditional filtered back-projection algorithm. Moreover, it is easy to obtain the training set, with only the point source for training, and the network can learn the reconstruction process of the image with any pixel distribution.
In USA, breast cancer is a most frequent cause of deaths for women. It is important to detect the cancer in its early stage. X-ray three-dimensional (3D) mammography can provide a good image resolution and contrast However, the associated radiation is relatively high. Reduction of the soft X-ray radiation for 3D mammography has been a research focus in the past years. In a typical 3D mammography system, the X-ray source and detector rotate around the object (breast) beneath the table, on which the patient lies in a prone position. In order to sample the data as close as possible to the chest base, a circular orbit with half cone-beam geometry has been investigated. It can provide very good reconstruction if the X-ray source is far away from the object. For a relatively short distance between the source and the object for an improved spatial resolution, the circular orbit may not be an optimal choice. In this case, the portion far away from the circular orbit wouldn't be well reconstructed because of the missing of projection data in that region. In this work, we investigated five possible orbits, attempting to find an optimal orbit that can reconstruct satisfactorily the whole object with least projections (less radiation). The results showed that two near half-circular orbits may be a choice, one near the chest base and the other near the breast tip. The redundant samplings beyond 180/spl deg/ were eliminated by our algorithm, rendering very good reconstructions.
Abstract Purpose Focal cortical dysplasia (FCD) is a common cause of epilepsy; the only treatment is surgery. Therefore, detecting FCD using noninvasive imaging technology can help doctors determine whether surgical intervention is required. Since FCD lesions are small and not obvious, diagnosing FCD through visual evaluations of magnetic resonance imaging (MRI) scans is difficult. The purpose of this study is to detect and segment histologically confirmed FCD lesions in images of normal fluid‐attenuated inversion recovery (FLAIR)‐negative lesions using convolutional neural network (CNN) technology. Methods The technique involves training a six‐layer CNN named Net‐Pos, which consists of two convolutional layers (CLs); two pooling layers (PLs); and two fully connected (FC) layers, including 60 943 learning parameters. We employed activation maximization (AM) to optimize a series of pattern image blocks (PIBs) that were most similar to a lesion image block by using the trained Net‐Pos. We developed an AM and convolutional localization (AMCL) algorithm that employs the mean PIBs combined with convolution to locate and segment FCD lesions in FLAIR‐negative patients. Five evaluation indices, namely, recall, specificity, accuracy, precision, and the Dice coefficient, were applied to evaluate the localization and segmentation performance of the algorithm. Results The PIBs most similar to an FCD lesion image block were identified by the trained Net‐Pos as image blocks with brighter central areas and darker surrounding image blocks. The technique was evaluated using 18 FLAIR‐negative lesion images from 12 patients. The subject‐wise recall of the AMCL algorithm was 83.33% (15/18). The Dice coefficient for the segmentation performance was 52.68. Conclusion We developed a novel algorithm referred to as the AMCL algorithm with mean PIBs to effectively and automatically detect and segment FLAIR‐negative FCD lesions. This work is the first study to apply a CNN‐based model to detect and segment FCD lesions in images of FLAIR‐negative lesions.
SPECT (single photon emission computed tomography) is a non-invasive, cost-effective means for assessment of tissue/organ functions in nuclear medicine. For more accurate diagnosis, quantitative reconstruction of radiotracer concentration at any location inside the body is desired. To achieve this goal, we have to address a number of factors that significantly degrade the acquired projection data. The cone-beam SPECT system has higher resolution comparing with parallel-beam and fan-beam SPECT, which is highly advantageous in small object detection. In this paper, we used four analytical reconstruction schemes for cone-beam SPECT that allow simultaneous compensation for non-uniform attenuation and distance-dependent resolution variation (DDRV), as well as accurate treatment of Poisson noise. The simulation results show that the reconstruction scheme 1 and 4 both can obtain good reconstruction results.
Single photon emission computed tomography (SPECT) is a nuclear medicine imaging technique that is widely used in the clinical applications. For more accurate diagnosis, quantitative reconstruction of radiotracer concentration at any location inside the body is desired, which requires accurate compensation for the non-uniform attenuation and accurate treatment of the Poisson noise. However, we found that the treatment of Poisson noise in projection data (sinogram) would possibly affect the result of analytical SPECT reconstruction. In this paper, we used several de-noising methods to study the effect of de-noising for analytical SPECT reconstruction with non-uniform attenuation.
Computed tomography technology has wide applications in industrial non-destructive tests, medical fields, and public security. In practice, incomplete projection is often acquired because of limitations on the detection environment, time and cost. Iterative reconstruction algorithm has a major potential to achieve superior results on the reconstruction with incomplete projection data. In this paper, we presented an iterative approach utilizing improved Particle Swarm Optimization, which can efficiently reconstruct the image from incomplete projection data by solving global optimization problem. Simulation results showed that the proposed algorithm has better convergence speed and reconstruction quality.
'/%3e%3cuse xlink:href='%23a' transform='rotate(23.036 .093 25.536)'/%3e%3cuse xlink:href='%23a' transform='rotate(45.87 1.273 16.18)'/%3e%3cuse xlink:href='%23a' transform='rotate(69.945 .996 12.078)'/%3e%3cuse xlink:href='%23a' transform='rotate(20.66 -19.689 31.932)'/%3e%3c/svg%3e)
