The
Course: Digital imaging and PACS
Subject Code: HTI 5720
Assignment 1:
Literature review on the image filtering approach
Szucs-Farkas Z, et al. Nonlinear Three-dimensional Noise Filter with Low-Dose CT Angiography: Effect on the Detection of Small High-Contrast Objects in a Phantom Model. Radiology 2011; 258:261-269
Submitted to: Dr. Lawrence WC Chan
Student Name: Cheng Pak Kwan
Student ID: 10673287G
In this century, multi-slice CT (MSCT) angiography is very common tool for the non-invasive evaluation of the aortoiliac system. Because of, ionizing radiation is a well known fact that it has a drawback from the diagnosis. For someone always needs to follow-up the aortic dissection with Post Endo-Vascular Aortic aneurysm Repair (EVAR), could be lead to a large number of cumulative dose. According to the possibility of different species, reducing the CT tube voltage is a way of the common method to reduce patient entrance dose by multi-detector CT angiography. Briefly saying, the low-KV CT images are noisier than the normal-dose images comparatively. However, this phenomenon will inhibit to accept for diagnostic purposes. In spite of the increasing to inject iodine contrast agent at low tube voltage can improve some image qualities, however, it will increase noise and help keep the contrast-to-noise ratio (CNR) constant. Radiologist, who wants to reduce further noise problem by mean of apply noise filter or adaptive statistical iterative reconstruction algorithms to efficiently reduce noise.
The Aim of this journal is to evaluate the clinical practicability of spatial domain filtering as an alternative for supplementary to perform the image reconstruction using different kernels in CT angiography respected by using of 80- and 100-kVp tube voltage in patients with different body weights. They use the spatial domain filtering to generate smooth images from the sharp images and avoid doing several sets of images from reconstruction. It really helps us to reduce the longest processing time. The filter can play a role of a better detection rate for endoleaks at low-dose multi-detector CT angiography compared with using of nonfiltered images in large patient.
For continuous noise reduction point of view, image post-processing by using noise filters, built-in soft reconstruction kernels or adaptive statistical iterative reconstruction algorithms can efficiently reduce noise. Before this phenomenon, these applications have been tested in liver lesions and produced images with reduced spatial resolution and blurred edges of anatomic structures. Besides, using adaptive statistical iterative reconstruction become available in the latest-generation of CT scanners, it is a very time-consuming technique. No matter the detection of small endoleaks, which is adverse effects by the increased quantum noise at low tube voltage of multi-detector CT angiography, can be improved by using noise filters considered as a research topic.
Noise in images appears dominant with decreasing radiation dose on thick patient which, can be reduced by using noise-reduction filters (NRFs). Several approaches, such as linear low-pass filters, non-linear smoothing and non-linear, three-dimensional 3D filters, we have been used to reduce noise in scan datasets. A new algorithm called the nonlinear three-dimensional optimized reconstruction algorithm (ORA) filter is currently used to improve CT image quality since we will be reduced radiation dose. This algorithm note by mean of the comparison with image noise slices sensitivity profile (SSP), contrast-to-noise ratio, and modulation transfer function (MTF) on phantom images processed “with” and “without” the 3D ORA filter, and the affect of the 3D ORA filter on CT images with lower dose. For example on CT head scans the noise reduction was up to 54% compared with typical bone reconstruction algorithms (H70) and a 0.6mm slice thickness; secondly, liver CT scans the noise reduction was up to 30% compared with typical high-resolution reconstruction algorithm (B70) and a 0.6mm slice thickness. MTF and SSP have not significant change compared with applied for 3D ORA filtering (P>0.05), else noise is reduce (P<0.05). On low contrast test the detect ability and MTF of image obtained at a reduced dose and apply 3D ORA filter is equivalent to those of standard dose CT images; there is no significant difference in image noise of scans taken at a lower tube voltage, using 3D ORA and standard dose CT (P>0.05). Since, the 3D ORA filter shows a good potential for reducing images noise without affecting image quality an especially on sharpness. By using this approach, the same image quality can be achieved while gaining an optimum dose reduction.
What have not been done by this study? From this experiment, the scientists tested only one filter strength for each phantom size. Furthermore this phenomenon should be finds the optimal filter settings for clinical application. Secondly, the analyzed CT images are not in series to provide a consistent image order during the reading process. The potentiality of scrolling over a suspicious finding in image series can inevitably facilitate the detection of small objects. Thirdly, the scientists are only simulating static conditions in the phantom and haven’t taken consideration of Aortic lumen and endoleaks with time. Fourthly, the simulation only involved spherical endoleaks shape. Under real conditions, this is an irregular spatial configuration of endoleaks shape which will affect the test result. Finally the scientists haven’t tested the phantom with linear image filter applied.
This research question of this study is to understand the effect of nonlinear noise filter on the detection of simulated endoleaks in a phantom with 80- and 100-kVp multi-detector computed tomographic (CT) angiography. Based on this research, we studied on “nonlinear 3D ORA image filtering” as an alternative reconstruction kernels working in computed tomography and shown us a spatial domain filter function. It shows a good performance between reconstruction images using convolution kernels and filtered images in measuring of both pixel noise and modulation transfer function (MTF). What is the difference in image quality if applied of linear and nonlinear noise filters are used in multi-detector computed tomographic (CT) angiography? The image quality is also affected by the X-ray tube voltage and patient body size. The image reconstruction filter is one of the tools to improve the image diagnostic and quality.
The materials and method of this experiment is the scientist Mr. Szucs-Farkas, MD. Who study on ‘spatial domain image filtering in low dose computed tomography CT angiography: possibility study in an aortic aneurysm phantom, including iodinated endoleaks, was constructed. In CT scanning, anatomy information and noise are relevant to the radiation related dose. Radiation dose reduction is always inducing to increase in more noise. The appropriate imaging algorithms can minimize image noise without changing image sharpness and contrast. Thus, the 3D ORA technique was evolve to make low-dose CT images commonly use as routine clinical practice by decreasing image noise.
Many clinical applications may use different convolution kernels B10, B20, B30, B40, B50 and B60 to reconstruct of multiple sets of CT images from difference raw data sets. It may need more time to process the images reconstruction. The author was to reconstruct only three set of image using an ORA 3D Filter. Then, he obtained the smooth images from the sharp images by using a spatial domain filtering. From CT scanning, normally anatomy signal and noise are respected to the related radiation dose. Radiation dose deduction is usually together with increasing the image noise. We hypothesized that suitable imaging algorithms can eliminated the image noise without affecting image sharpness and contrast. Since, the 3D QRA technique was developed to make low-dose CT images their image quality can be accepted in routine clinical practice by eliminating image noise.
Base on the data analysis and visualization techniques used in this study, most of the measurements are done with 10 times in different group to prevent inaccuracies, and take the mean data apply for evaluating the CNR for the aorta namely: CNR= (HUAO – HUThrr)/noise, here HUAO is the attenuation data, which counts from the aorta and HUThrr is the attenuation data, which records from the thrombus. This experiment was designed to proof a 5% different in the sensitivity of endoleak detection between the various data sets distinguish by phantom size, tube voltage, and the existence leak of filtering. Consequently, assay of variance for repeated measurements with Post Hoc Tests and Friedman Analysis of variance make using for comparing findings at various tube voltages in the Aortic Aneurysm Phantom with or without a noise filter. According to size and density of endoleaks, tube energy, phantom size and noise filter on the number of true-positive result, diagnostic confidence, subjective noise, and image quality was analyzed by using a General Linear Model. Statistical tests are performed by using Statistical (StatSoft, Tulsa, Okla) and MedCalc (MedCalc, Mariakerke, Belgium) software. P values <.05 is concerned to show a statistically significant difference. Noise increased apparently when tube voltage is decreased or when the size of the water container is enlarged. Although the noise is increasing, the CNR keeps unchanged when the tube voltage is decreasing from 120 to 100kVp. Images reconstruction with the 3D ORA filter had apparently low noise than the non-filtered images.
The characteristics of these researched techniques, on this phenomenon, noise become dominant when tube voltage decrease or when the size of the water container is enlarged (P value will < 0.0001 for both filtered and non-filtered images). This approach has been passed the noise reduction in scan volume data set. It includes nonlinear smoothing and nonlinear 3D filters. It is well known fact that the linear low-pass filters reduce noise efficiently, they also reduce sharpness and conspicuity of small structures, The nonlinear smoothing kernel is based on two or three-dimensional filtration of image data, which concerns position and orientation of edges at each step of noise reduction. The two-dimensional nonlinear smoothing filter, which is performed filtration in the x-y plan (axial view) only, whereas the position and orientation of edges is determined wantonly. For other one-dimensional filtration is performed along these edges. Since these filters haven’t take into consideration with information in the direction perpendicular to the x-y plane, smaller structures along the x-y plan, which will lose contrast because of partial-volume effect and will be made into ‘invisible”.
The 3D filtration procedure evaluates the present study (3D ORA) generalizes the two-dimensional nonlinear smoothing technique in all three directions (x, y, and z axes) to avoid loss of contrast and sharpness of small structures. Tube energy and noise filter play as an important effect on the diagnostic purpose; subjective noise and image quality both have a very closely bi-direction effect. The image quality and number of identified endoleaks, this point of view, which is highly dependent on phantom size. The aimed of this test, find out the co-relation between applied 3DORA filter with difference phantom size on the detection of small, high-contrast objects with 80- and 100kVp multi-detector CT angiography and endoleaks just represent for an example for the high contrast resolution of lesions.
In conclusion, the nonlinear 3D ORA filter can improve image quality at abdominal multi-detector CT angiography by using 80 and 100kVp in 30- and 40-cm phantoms, which simulate the normal body sized and thick patients. Moreover, the detection rate for endoleaks can be improved with the noise filter at 100kVp in 40-cm phantom size. The spatial domain image filtering in CT image has been confirmed to apply on clinically feasible and should be performed on clinical environment. Both of the filtered images and reconstructed images can improve the diagnostic accuracy in segmental and sub-segmental pulmonary embolism. Future view, we can use different kernels on real image modifications of the sharpness and pixel noise in online mode for multiple reconstructions.
The transmission of electronic images is widely use of different areas in many organization today. Typically it uses for transmission of images to the intensive care unit, operating room, emergency room, and physicians at home or viewing room area. The Picture Archiving and Communication System (PACS) can be determinate as system to encourage image diagnosis, leading to secondary medical interpretation since provided the better treatment for the patients. PACS will sensible of an interdepartmental committee to share the planning of the functional responsibilities of reporting, not only to ensure that the patient-physician-radiologist relationship functions to provide the best patient service but also to protect the integrity of the radiology department. PACS is a good tool for all medical imaging modalities, normally the radiology department should be the major input into the system, its modification, and upgrading is the most necessity. In most cases it will be better to start PACS network system on the limited areas, such as intensive care units, satellite units, operating rooms, and emergency rooms. Aside from interventional radiology, the consultative service is provided by radiologists, which can also provided the direct service to clinician and fast and accuracy diagnosis to the patients.
It is unclear how the PACS is related to the noise filtering.
回覆刪除Grade: C+