Kinetic cluster and α-divergence-based dynamic myocardial factorial analysis of positron-emission computed tomography images / 南方医科大学学报

<p><b>OBJECTIVE</b>We purpose a novel factor analysis method based on kinetic cluster and α-divergence measure for extracting the blood input function and the time-activity curve of the regional tissue from dynamic myocardial positron emission computed tomography(PET) images.</p><p><b>METHODS</b>Dynamic PET images were decomposed into initial factors and factor images by minimizing the α-divergence between the factor model and actual image data. The kinetic clustering as a priori constraint was then incorporated into the model to solve the nonuniqueness problem, and the tissue time-activity curves and the tissue space distributions with physiological significance were generated.</p><p><b>RESULTS</b>The model was applied to the 82RbPET myocardial perfusion simulation data and compared with the traditional model-based least squares measure and the minimal spatial overlap constraint. The experimental results showed that the proposed model performed better than the traditional model in terms of both accuracy and sensitivity.</p><p><b>CONCLUSION</b>This method can select the optimal measure by α value, and incorporate the prior information of the kinetic clustering of PET image pixels to obtain the accurate time-activity curves of the tissue, which has shown good performance in visual evaluation and quantitative evaluation.</p>

Three-dimensional low-dose CT volume reconstruction based on non-local weights optimization and GPU acceleration / 南方医科大学学报

Concerns have been raised over x-ray radiation dose associated with repeated computed tomography (CT) scans for tumor surveillance and radiotherapy planning. In this paper, we present a low-dose CT image reconstruction method for improving low-dose CT image quality. The method proposed exploited rich redundancy information from previous normal-dose scan image for optimizing the non-local weights construction in the original non-local means (NLM)-based low-dose image reconstruction. The objective 3D low-dose volume and the previous 3D normal-dose volume were first registered to reduce the anatomic structural dissimilarity between the two datasets, and the optimized non-local weights were constructed based on the registered normal-dose volume. To increase the efficiency of this method, GPU was utilized to accelerate the implementation. The experimental results showed that this method obviously improved the image quality, as compared with the original NLM method, by suppressing the noise-induced artifacts and preserving the edge information.

Adaptive regularized super-resolution reconstruction for magnetic resonance images / 南方医科大学学报

To increase the resolution and signal-to-noise ratio (SNR) of magnetic resonance (MR) images, an adaptively regularized super-resolution reconstruction algorithm was proposed and applied to acquire high resolution MR images from 4 subpixel-shifted low resolution images on the same anatomical slice. The new regularization parameter, which allowed the cost function of the new algorithm to be locally convex within the definition region, was introduced by the piori information to enhance detail restoration of the image with a high frequency. The experiment results proved that the proposed algorithm was superior to other counterparts in achieving the reconstruction of low-resolution MR images.

A graph cuts-based interactive method for segmentation of magnetic resonance images of meningioma / 南方医科大学学报

For accurate segmentation of the magnetic resonance (MR) images of meningioma, we propose a novel interactive segmentation method based on graph cuts. The high dimensional image features was extracted, and for each pixel, the probabilities of its origin, either the tumor or the background regions, were estimated by exploiting the weighted K-nearest neighborhood classifier. Based on these probabilities, a new energy function was proposed. Finally, a graph cut optimal framework was used for the solution of the energy function. The proposed method was evaluated by application in the segmentation of MR images of meningioma, and the results showed that the method significantly improved the segmentation accuracy compared with the gray level information-based graph cut method.

CT image retrieval of the liver with intrahepatic lesions / 南方医科大学学报

This paper presents a method for global feature extraction and the application of the boostmetric distance metric method for medical image retrieval. The global feature extraction method used the low frequency subband coefficient of the wavelet decomposition based on the non-tensor product coefficient for piecewise Gaussian fitting. The local features were extracted after semi-automatic segmentation of the lesion areas in the images in the database. The experimental verification of the method using 1688 CT images of the liver containing lesions of liver cancer, liver angioma, and liver cyst confirmed that this feature extraction method improved the detection rate of the lesions with good image retrieval performance.

Sparse angular CT projection onto convex set reconstruction using nonlocal means iterative modification / 南方医科大学学报

Based on the fact that nonlocal means (NL-means) filtered image can likely produce an acceptable priori solution, we propose a sparse angular CT projection onto convex set (POCS) reconstruction using NL-means iterative modification. The new reconstruction scheme consists of two components, POCS and NL-means filter. In each phase of the sparse angular CT iterative reconstruction, we first used POCS algorithm to meet the identity and non-negativity of projection data, and then performed NL-means filter to the image obtained by POCS method for image quality improvement. Simulation experiments showed that the proposed POCS scheme can significantly improve the quality of sparse angular CT image by suppressing the noise and removing the streak-artifacts.

Research on algorithms based on Markov random models for diffusion tensor-magnetic resonance images / 南方医科大学学报

With the utilization of diffusion tensor information of image voxels, a novel MRF (Markov Random Field) segmentation algorithm was proposed for diffusion tensor MRI (DT-MRI) images benefitted from the introduction of Frobenius norm. The comparison of the segmentation effects between the proposed algorithm and K-means segmentation algorithm for DT-MRI image was made, which showed that the new algorithm could segment the DT-MRI images more accurately than the K-means algorithm. Moreover, with the same segmentation algorithm of MRF, better outcomes were achieved in DT-MRI than in conventional MRI (T2WI) image.

Cerebral blood flow, cerebral blood volume, oxygen utilization and oxygen extraction fraction: the influence of age / 南方医科大学学报

<p><b>OBJECTIVE</b>To analyze the changes of cerebral blood flow (CBF), cerebral blood volume (CBV), oxygen utilization (CMRO2) and oxygen extraction fraction (OEF) with age.</p><p><b>METHODS</b>The PET images of 7 young (21.0-/+1 years old) and 7 aged volunteers (60.9-/+4.7 years old) were analyzed to identify the areas where CBF, CBV, CMRO2, OEF had significant differences with age. The images were anatomically normalized by statistical parametric mapping (SPM2). A voxel by voxel calculation was performed to obtain the slope with age. Voxels which had statistically significant differences (P<0.05) with age were shown both on global and ROIs brain images.</p><p><b>RESULTS</b>The CBF decreased with age as was consistent with previous reports. The age-related changes in CBV and CMRO2 were similar to CBF, but OEF increased with age.</p><p><b>CONCLUSION</b>CBF, CBV and CMRO2 generally decline with age. The increase in OEF with age suggests a greater reduction in CBF than in CMRO2. The most significant decreases of CBF and CMRO2 occur in the convexity of the frontal cortex and inferior parietal cortex in all the functional images, while in the white matter, the influence of age is minimal.</p>

Adaptive super resolution algorithm for under-sampled images / 南方医科大学学报

A new algorithm of adaptive super-resolution (SR) reconstruction based on the regularization parameter is proposed to reconstruct a high-resolution (HR) image from the low-resolution (LR) image sequence, which takes into full account the inaccurate estimates of motion error, point spread function (PSF) and the additive Gaussian noise in the LR image sequence. We established a novel nonlinear adaptive regularization function and analyzed experimentally its convexity to obtain the adaptive step size. This novel algorithm can effectively improve the spatial resolution of the image and the rate of convergence, which is verified by the experiment on optical images.

Hybrid interpolation for CT metal artifact reducing / 南方医科大学学报

Numerous interpolation-based methods have been described for reducing metal artifacts in CT images, but due to the limit of the interpolation methods, interpolation alone often fails to meet the clinical demands. In this paper, we describe the use of quartic polynomial interpolation in reconstruction of the images of the metal implant followed by linear interpolation to eliminate the streaks. The two interpolation methods are combined according to their given weights to achieve good results.

Graph-based interactive three-dimensional segmentation of magnetic resonance images of brain tumors / 南方医科大学学报

We propose a graph-based three-dimensional (3D) algorithm to automatically segment brain tumors from magnetic resonance images (MRI). The algorithm uses minimum s/t cut criteria to obtain a global optimal result of objective function formed according to Markov Random Field Model and Maximum a posteriori (MAP-MRF) theory, and by combining the expectation-maximization (EM) algorithm to estimate the parameters of mixed Gaussian model for normal brain and tumor tissues. 3D segmentation results of brain tumors are fast achieved by our algorithm. The validation of the algorithm was tested and showed good accuracy and adaptation under simple interactions with the physicians.

Cone-beam pseudo lambda tomography under FDK framework / 南方医科大学学报

The medical CT scanner is rapidly evolving from the fan-beam mode to the cone-beam geometry mode. In this paper, a new cone-beam pseudo Lambda tomography was proposed based on the Noo's fan beam super-short scan formula and FDK framework. The proposed pseudo-LT algorithm, which avoids the computation of any PI line and any differential operation, has a significant practical implementation, thus leading to the images with quality improvement and reduced artifacts. The results in the simulation studies confirm the observation that the new algorithm can improve the image resolution over the traditional algorithms with noise projection data.

Medical image segmentation based on Gibbs morphological gradient and distance map Snake model / 南方医科大学学报

<p><b>OBJECTIVE</b>To propose a new algorithm for medical image segmentation based on Gibbs morphological gradient and distance map (DM) Snake model, which allows identification of the correct contours of the objects when processing medical images with noises and pseudo-edges.</p><p><b>METHODS</b>Gibbs morphological gradient was deduced and the method for image segmentation based on Gibbs morphological gradient and distance map Snake model was presented.</p><p><b>RESULTS</b>This new medical image segmentation algorithm proved to effectively suppress the noises and pseudo-edges when calculating distance map.</p><p><b>CONCLUSION</b>The proposed algorithm is robust for image noise suppression and allows easy implementation in clinical image segmentation without the need of user interventions.</p>

A fast 3-D medical image registration algorithm using principal component analysis / 南方医科大学学报

This paper presents a new 3-D image registration method based on the principal component analysis (PCA). Compared with intensity-based registration methods using the whole volume intensity information, our approach utilizes PCA to estimate the centroid and principal axis, and completes the registration by aligning the centroid and principal axis. We evaluated the effectiveness of this approach by applying it to simulated and actual brain image data (MR, CT, PET, and SPECT). The experimental results indicate that the algorithm is effective, especially for registration of 3-D medical images.

Self-adapted segmentation of pulmonary nodule based on region growing / 南方医科大学学报

<p><b>OBJECTIVE</b>To improve the accuracy and efficiency of pulmonary nodule segmentation of thoracic CT image for computer-aided diagnostic (CAD) system, especially for those nodules adhering to the pleural or blood vessels.</p><p><b>METHODS</b>We proposed the automatic process of pulmonary nodule segmentation, and using region growing method based on the contrast and gradient, the pulmonary nodule images were acquired. A self-adapted morphologic segmentation algorithm was presented for the unsuccessful nodule segmentation using region growing.</p><p><b>RESULTS AND CONCLUSIONS</b>Experiments on clinical 2D pulmonary images showed that the solitary pulmonary nodules and those adhering to the pleural or blood vessels could all be segmented. This pulmonary nodule segmentation algorithm is feasible for automated segmentation of the thoracic CT images.</p>

A new algorithm for magnetic resonance image segmentation based on fuzzy kerne1 clustering / 南方医科大学学报

Fuzzy clustering technique is a popular model widely used in the segmentation of magnetic resonance (MR) images. However, when the conventional fuzzy clustering algorithm is used for image segmentation, the algorithm strictly depending on the current pixels works only on images with less noise. In the paper, we presented a modified fuzzy kernel clustering algorithm for MR image segmentation. The new algorithm incorporates a kernel-induced distance mertric and a penalty term that controls the neighborhood effect to the objective function. The results of experiment on both the synthetic images and simulated MR images show that the proposed algorithm is more robust to noise than the standard fuzzy image segmentation algorithms.

Elimination of direct derivative of the projection data for image reconstruction based on hybrid filter in helical cone-beam CT / 南方医科大学学报

<p><b>OBJECTIVE</b>We present an alternative approach for precise reconstruction of the images from helical cone-beam projections combining Hilbert filter and Ramp filter.</p><p><b>METHODS</b>Based on the Katsevich algorithm framework, the proposed algorithm combined the FDK-type algorithms and Katsevich algorithm for their respective advantages, to completely avoid the direct derivatives with respect to the coordinates on the detector plane.</p><p><b>RESULTS</b>The experimental results validated the accuracy of the new algorithm, and this approach significantly improved the resolution of the reconstructed images with much reduced artifacts.</p><p><b>CONCLUSION</b>The proposed reconstruction formula based on hybrid Hilbert-Ramp filter is an important development of Katsevich reconstruction formula, and the different forms of the Ramp filters can be designed to realize frequency modulation according to the actual clinical application.</p>

Improved reconstruction for PROPELLER data in magnetic resonance imaging based on fuzzy enhancement / 南方医科大学学报

To improve the conventional reconstruction algorithm for PROPELLER MRI data, we propose a new algorithm based on fuzzy enhancement. The motion parameters were extracted from fuzzy enhanced images reconstructed through zero-padding strips. After motion compensation, the image was obtained through gridding reconstruction. The experiment results showed that this algorithm could estimate and compensate the motion more robustly and precisely, and the motion artifacts could be better suppressed to obtain improved image quality.

A new Bayesian reconstruction algorithm for PET images based on correction of the detected sinogram data / 南方医科大学学报

<p><b>OBJECTIVE</b>To improve Bayesian reconstruction of positron-emission tomography (PET) images by devising a novel coupled feedback (CF) iterative model.</p><p><b>METHODS</b>The CF iterative algorithm was applied to update the noisy detected emission sinogram data using the latest reconstructed image in the iterative process of PET reconstruction. The relevant operations included linear filtering, wiener filtering, and projection of the reconstructed images. The sinogram data used in the study was obtained from simulated phantom data.</p><p><b>RESULTS</b>The experiments and corresponding visional and quantitative comparisons showed that the new method had better performance than the traditional Bayesian reconstruction of PET images for improvement of the PET images.</p><p><b>CONCLUSIONS</b>The proposed sinogram-correcting method allows improvement on the original measurement data, and is applicable for PET image reconstruction or other reconstruction tasks with high noise level.</p>

A new unsupervised algorithm for image segmentation based on an inhomogeneous Markov random field model / 南方医科大学学报

A new unsupervised algorithm for image segmentation is proposed using an inhomogeneous Markov random field (MRF) model, in which the parameter is estimated in fuzzy spel affinities. The proposed algorithm improved the accuracy of segmentation. Simulated brain MR image with different noise levels and clinical brain MR image were presented in the experiments. The results showed that the proposed algorithm was more powerful than conventional homogeneous MRF model-based ones and than the fuzzy c-means clustering algorithm as well.