Yongjin Chang

Denoising of polychromatic CT images based on their own noise properties

PURPOSE Because of high diagnostic accuracy and fast scan time, computed tomography (CT) has been widely used in various clinical applications. Since the CT scan introduces radiation exposure to patients, however, dose reduction has recently been recognized as an important issue in CT imaging. However, low-dose CT causes an increase of noise in the image and thereby deteriorates the accuracy of diagnosis. In this paper, the authors develop an efficient denoising algorithm for low-dose CT images obtained using a polychromatic x-ray source. The algorithm is based on two steps: (i) estimation of space variant noise statistics, which are uniquely determined according to the system geometry and scanned object, and (ii) subsequent novel conversion of the estimated noise to Gaussian noise so that an existing high performance Gaussian noise filtering algorithm can be directly applied to CT images with non-Gaussian noise. METHODS For efficient polychromatic CT image denoising, the authors first reconstruct an image with the iterative maximum-likelihood polychromatic algorithm for CT to alleviate the beam-hardening problem. We then estimate the space-variant noise variance distribution on the image domain. Since there are many high performance denoising algorithms available for the Gaussian noise, image denoising can become much more efficient if they can be used. Hence, the authors propose a novel conversion scheme to transform the estimated space-variant noise to near Gaussian noise. In the suggested scheme, the authors first convert the image so that its mean and variance can have a linear relationship, and then produce a Gaussian image via variance stabilizing transform. The authors then apply a block matching 4D algorithm that is optimized for noise reduction of the Gaussian image, and reconvert the result to obtain a final denoised image. To examine the performance of the proposed method, an XCAT phantom simulation and a physical phantom experiment were conducted. RESULTS Both simulation and experimental results show that, unlike the existing denoising algorithms, the proposed algorithm can effectively reduce the noise over the whole region of CT images while preventing degradation of image resolution. CONCLUSIONS To effectively denoise polychromatic low-dose CT images, a novel denoising algorithm is proposed. Because this algorithm is based on the noise statistics of a reconstructed polychromatic CT image, the spatially varying noise on the image is effectively reduced so that the denoised image will have homogeneous quality over the image domain. Through a simulation and a real experiment, it is verified that the proposed algorithm can deliver considerably better performance compared to the existing denoising algorithms.

Super-Resolution Reconstruction of 3D PET Images Using Two Respiratory-Phase Low-Dose CT Images

Positron emission tomography (PET) images usually suffer from incorrect quantification of the radioactive uptake of small lesions due to low spatial resolution. To improve the spatial resolution, we previously proposed super-resolution (SR) algorithms based on wobble scanning. The proposed algorithms, however, require mechanical motion of the patient bed or a system gantry for wobble scanning. In this paper, we propose a framework for SR reconstruction of 3D PET images based on the use of respiratory motion rather than wobble motion. As in the conventional protocol of PET imaging, gated list-mode PET data are acquired in a free breathing condition. In addition, we acquire two low-dose CT images in a breath-hold manner at exhale and inhale phases, without increasing the radiation burden to a patient. Using the two low-dose CT images, we estimate the 4D motion vector field (MVF) and correspondingly generate a virtual 4D CT image that are matched to the 4D PET image. The 3D CT images have much better spatial resolution than PET images and therefore the corresponding estimated 3D MVFs can be considered reliable for PET SR reconstruction. We then estimate space-variant point spread functions (PSFs) in the imaging field of view using a minimum number of PSFs obtained through Monte-Carlo simulations. Finally, SR reconstruction is performed by incorporating the estimated 3D MVFs and space-variant PSFs. In the SR reconstruction, to avoid the resolution degradation in the evenly spaced parallel-beam rebinning and to reduce the computational time on the graphics processing unit, we introduce a parallel-friendly spanned line of response reconstruction technique based on fan-beam reordering. The proposed framework is evaluated via Monte-Carlo simulations with the digital XCAT phantom and via experiments with several patient datasets including moving lung lesions. Both the simulation and experiment results show that the proposed framework provides much clearer organ boundaries as well as more accurate quantitative lesion information than the conventional methods, with a considerable reduction of computational time.

LOR-Based Reconstruction for Super-Resolved 3D PET Image on GPU

Positron emission tomography (PET) images usually suffer from low spatial resolution mainly because of the finite dimension of crystals. To improve the spatial resolution based on wobble scanning, we previously proposed a sinogram-based super-resolution (SR) algorithm using a space-variant blur matrix. However, the algorithm may cause unwanted resolution loss owing to an inevitable interpolation process for preparing evenly spaced projections. In this article, we propose a novel one-step line of response (LOR)-based SR framework for 3D PET images. In the framework, we efficiently determine a large number of space-variant point spread functions (PSFs) in the image space by using the scanner symmetries and the proposed PSF interpolation scheme based on nonrigid registration. Furthermore, to minimize the resolution degradation in the evenly spaced parallel-beam rebinning and to reduce the computational time in the graphics processing unit (GPU) implementation, we introduce parallel-friendly LOR reconstruction based on cone-beam reordering. We then obtain a high resolution image via a one-step super-resolved 3D PET image reconstruction with the determined PSFs. The proposed framework provides noticeable improvement on the spatial resolution of PET images with a considerable reduction of computational time.

Post-filtering of PET image based on noise characteristic and spatial sensitivity distribution

Positron emission tomography (PET) images suffer from a noticeable amount of statistical noise. In order to reduce the noise, a post smoothing process is usually adopted in the conventional PET systems. However, its performance is limited because the process is mostly based on a Gaussian random noise which is quite distinct from the noise of PET images. It has been reported that noise variance of each voxel is proportional to the square of the mean value in a PET image reconstructed by expectation-maximization (EM). In addition, we also observe that the variance varies with the spatial sensitivity distribution in a PET system. Based on those properties, we determine a unique formula representing a relationship between the mean and variance for a given PET system. Meanwhile, a block matching 3D (or 4D) algorithm is known as the state of the art in Gaussian noise reduction. To effectively apply it for noise reduction of PET images, we first perform a noise characteristic conversion from the PET image noise to Gaussian random noise using a pre-determined relationship. We then apply a block matching 4D (BM4D) algorithm and reconvert the result. Using the Monte Carlo simulation, we demonstrate that proposed algorithm can effectively reduce the noise in the whole image region while minimizing the image resolution degradation.

Motion-compensated reconstruction based on filtered backprojection for helical head CT

In head CT imaging, it is assumed that the patient’s head does not move during the CT acquisition. In clinical practice, however, the head sometimes moves and thereby causes considerable motion artifacts on the reconstructed image. To solve this motion artifact problem, motion estimation (ME) and motion-compensated (MC) reconstruction are needed. Reliable MC reconstruction is especially critical, because it is usually used for ME in addition to motion compensation. In this work, we propose a novel MC reconstruction algorithm for helical head CT, under the assumption that the head motion is rigid. CT acquisition of a rigidly moving object in a helical scan geometry can be considered as the acquisition of a static object in the scan geometry virtually transformed according to the motion. Based on this consideration, we propose a MC reconstruction algorithm by assuming that the head motion is already estimated. The algorithm consists of three consecutive steps, namely, MC rebinning, tangential filtering, and weighted backprojection. In the rebinning step, a virtually transformed helical geometry according to the motion is carefully taken account of, and a new weighting function is introduced to the backprojection step to minimize unwanted artifacts. To evaluate the proposed algorithm, we perform simulations by using a numerical phantom with pre-defined motion in a helical scan geometry. The proposed MC algorithm well restores a reconstructed image that is corrupted by motion, and thereby achieves the image quality comparable to that of the phantom with no motion.

Motion compensated 4D PET-CT-MR image generation for respiratory synchronized multi-modal image display

Positron emission tomography (PET) image has been widely used for early detection of malignant lesion(s) and its treatment, because it can provide functional information. Since computed tomography (CT) and magnetic resonance (MR) images can provide high resolution anatomical information, their simultaneous display with PET image may improve the clinical value of PET-image-based applications. Meanwhile, temporal information regarding anatomical changes due to respiration may be important in the radiation therapy planning of a thoracic or abdominal region. To provide such important information, we propose a framework for motion compensated 4D PET-CT-MR image generation. To realize the framework, we present a MR-driven motion compensation algorithm for respiratory-gated 4D PET imaging. We then introduce a respiratory-synchronized multi-modal image generation method. We expect that the respiratory-synchronized multi-modal image display can be useful for clinicians to efficiently utilize the temporal information as well as the complementary spatial information in their applications.

LOR-based reconstruction for super-resolved 3D PET image

PET images usually suffer from low spatial resolution due to positron range, photon non-collinearity, scatters inside scintillating crystals, finite dimension of crystals, and so on. To improve the spatial resolution based on wobble scanning, we previously proposed a sinogram-based super-resolution (SR) algorithm based on a space-variant blur matrix. However, the algorithm may cause unwanted resolution loss due to an inevitable interpolation process for preparing even-spaced sinograms. In this paper, we propose a novel and efficient one-step line of response (LOR) based SR framework for 3D PET images. In the framework, we efficiently determine a large number of space-variant PSFs in an image space by using the scanner symmetries and the proposed PSF interpolation scheme based on non-rigid registration. We then obtain a HR image via one-step super-resolved 3D PET image reconstruction with the determined PSFs. Furthermore, we reduce the computational time of GPU-based reconstruction by introducing a parallel-friendly cone-beam based LOR system matrix. The proposed framework provides noticeable improvement on the spatial resolution of PET images with a considerable reduction of computational time.