Iterative reconstruction techniques (IRTs) are used to reduce the radiation dose significantly and suppress the noise in
computed tomography (CT) imaging. However, the image from IRTs is unacceptable to the human visual system due to
the presence of outlier and non-Gaussian noise structure. The conventional noise measurements, such as mean and
standard deviation, are limited to provide the information of noise characteristics of an image comprehensively when the
undesirable noise structure happens on the image. In this study, the images reconstructed by using Weighted filtered
back-projection (WFBP) and image-based IRTs (SAFIRE and SafeCT) were compared in terms of conventional noise
statistics, high-order noise statistics, modulation transfer function (MTF), and slice sensitivity profile (SSP) with
different levels of radiation dose. The results showed that the noise characteristics, which were considered with
conventional and high-order noise statistics, and spatial resolution characteristics were different for the IRTs,
reconstruction parameters, and levels of radiation dose. This study can contribute to the optimization of image quality
from IRT with the reduction of radiation dose and development of IRT which overcomes the issue caused by undesirable
noise structures.
Iterative reconstruction techniques (IRTs) has been shown to suppress noise significantly in low dose CT imaging.
However, medical doctors hesitate to accept this new technology because visual impression of IRT images are
different from full-dose filtered back-projection (FBP) images. Most common noise measurements such as the
mean and standard deviation of homogeneous region in the image that do not provide sufficient characterization
of noise statistics when probability density function becomes non-Gaussian. In this study, we measure L-moments
of intensity values of images acquired at 10% of normal dose and reconstructed by IRT methods of two state-of-art
clinical scanners (i.e., GE HDCT and Siemens DSCT flash) by keeping dosage level identical to each other.
The high- and low-dose scans (i.e., 10% of high dose) were acquired from each scanner and L-moments of noise
patches were calculated for the comparison.
KEYWORDS: Systems modeling, Image restoration, Imaging systems, Data modeling, CT reconstruction, Image quality, 3D modeling, Computed tomography, Medical imaging, In vivo imaging
Total variation (TV) based iterative image reconstruction has been shown to possess desirable noise suppression
and edge preservation characteristics. However, such approaches also produce "staircase effects" where intensity
ramps are discretized into steps, resulting in images which appear blocky or patchy. In this paper, we present an
improved regularization technique by incorporating higher-order derivatives to reduce staircase artifacts without
sacrificing edge sharpness. In this preliminary investigation we demonstrate our approach using both phantom
and clinical images acquired at 25% of conventional radiation dosage (i.e., 75% dose reduction).
CT imaging is useful and ubiquitous. There is, however, a desire to reduce imaging artifacts, improve resolution, while reducing radiation. Iterative reconstruction algorithms have been proposed as one approach towards achieving these goals. In this paper we compare phantom images produced using commercial FBP-based reconstruction to three different iterative algorithms. We focus specifically on statistical characterizations of the noise, both at full radiation dose and at 50% dose. An iterative algorithm which segregates the image into two components (soft tissue and dense object), and imposes different constraints on these components, yielded better noise characteristics than ART, total variation, and FBP.
A common problem arising in medical imaging is the suppression of undesired image artifacts with the simultaneous
preservation of salient clinical information. Often the proposed processing "cure" introduces its own artifacts in other
parts of the image that confound reliable diagnosis. A canonical example is the suppression of artifacts from hyperdense
objects, such as metal and calcium. In this paper we propose a new decomposition-based approach to the combined
image formation and the suppression of localized image artifacts which is motivated by recent results on image
inpainting. The approach, which we term Model-Based Algebraic Iteration (MBAI) processing, decomposes an image
into a collection of homogeneous components, each of which can be reconstructed in the manner most appropriate to its
underlying nature. Because each component is localized, the effects of processing on that component do not contaminate
other areas of the image. Our specific motivation is the mitigation of artifacts in cardiac multi-detector computed
tomography (MDCT) images.
Recently, MDCT has offered the promise of a non-invasive alternative to invasive coronary angiography to evaluate
coronary artery disease. An impediment preventing its utilization as a routine clinical replacement for angiography is the
presence of image "blooming" artifacts due to the presence vascular calcium. We develop MBAI for the purpose of
ameliorating artifacts in cardiac images and thus increase the applicability of MDCT for the evaluation of at-risk patient
population. We demonstrate preliminary results in the reduction of the calcium blooming-effect in software simulation,
phantom, ex-vivo, and in-vivo MDCT data.
Modern CT systems have advanced at a dramatic rate. Algebraic iterative reconstruction techniques have shown
promising and desirable image characteristics, but are seldom used due to their high computational cost for complete
reconstruction of large volumetric datasets. In many cases, however, interest in high resolution reconstructions is
restricted to smaller regions of interest within the complete volume. In this paper we present an implementation of a
simple and practical method to produce iterative reconstructions of reduced-sized ROI from 3D helical tomographic
data. We use the observation that the conventional filtered
back-projection reconstruction is generally of high quality
throughout the entire volume to predict the contributions to
ROI-related projections arising from volumes outside the
ROI. These predictions are then used to pre-correct the data to produce a tomographic inversion problem of
substantially reduced size and memory demands. Our work expands on those of other researchers who have observed
similar potential computational gains by exploiting FBP results. We demonstrate our approach using cardiac CT cone-beam
imaging, illustrating our results with both ex vivo and in vivo multi-cycle EKG-gated examples.
This paper presents an automatic tissue segmentation methodology for High-Resolution Ultrasonic Transmission Tomography (HUTT) imagery of biological organs. This method combines a recent segmentation approach: the L-level set active contours algorithm with unsupervised clustering using the agglomerative hierarchical k-means algorithm. The active contours algorithm has been recently explored as a powerful tool for image segmentation since it automatically decomposes a given image into 2L segment classes by utilizing L level set functions and finding the optimal boundaries of the 2L segment classes so that the pixel feature values of each segment are as homogeneous as possible. Unfortunately, the algorithm is often trapped at local minima due to the intrinsic non-convexity of the cost function, especially for noisy data. To overcome this problem, we introduce a multi-stage multi-resolution analysis that optimizes the active contours at successive resolutions of the image data. The resulting segments are then re-clustered by subsequent agglomerative hierarchical k-means clustering that seeks the optimal clusters yielding the minimum within-cluster distance in the feature space. The preliminary studies reported here indicate that this proposed methodology can enhance the accuracy of soft tissue segmentation and provide fully automatic tissue differentiation without any user intervention except for specifying the number of level set functions L.
Recently it was shown that soft tissue can be differentiated with spectral unmixing and detection methods that utilize multi-band information obtained from a High-Resolution Ultrasonic Transmission Tomography (HUTT) system. In this study, we focus on tissue differentiation using the spectral target detection method based on Constrained Energy Minimization (CEM). We have developed a new tissue differentiation method called “CEM filter bank”. Statistical inference on the output of each CEM filter of a filter bank is used to make a decision based on the maximum statistical significance rather than the magnitude of each CEM filter output. We validate this method through 3-D inter/intra-phantom soft tissue classification where target profiles obtained from an arbitrary single slice are used for differentiation in multiple tomographic slices. Also spectral coherence between target and object profiles of an identical tissue at different slices and phantoms is evaluated by conventional cross-correlation analysis. The performance of the proposed classifier is assessed using Receiver Operating Characteristic (ROC) analysis. Finally we apply our method to classify tiny structures inside a beef kidney such as Styrofoam balls (~1mm), chicken tissue (~5mm), and vessel-duct structures.
In diagnostic ultrasound, tissue differentiation is essential to detect lesions or cancerous tissues from normal tissue. The attenuation characteristics of various tissues will be different at different frequencies, since the propagating ultrasonic pulse undergoes frequency-dependent attenuation, that is characteristic of the material it traverses. These vectors of attenuation values at different frequency bands represent multi-band characteristics of individual pixels (termed “multispectral”) that can be used for tissue differentiation akin to color. In this study, we have developed tissue differentiation methods that utilize the multispectral signatures of different materials in multi-band images produced by a newly built high-resolution ultrasonic transmission tomography (HUTT) system. The HUTT system obtains 3-D multi-band sinograms through FFT analysis of the first arriving pulse (snippet). A filtered backprojection algorithm is utilized to reconstruct a stack of multi-band attenuation images that contain multispectral signatures for each pixel and represent a multispectral augmentation of the 2-D conventional tomographic slice. To differentiate each tissue type according to its characteristic multispectral signature, we adopt the methods of spectral unmixing and spectral target detection. We demonstrate the feasibility of tissue differentiation using multi-band/multispectral signatures of different tissue objects in initial data collected from soft animal tissue phantoms.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.