HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Figures Only Full Text (PDF) Alert me when this article is cited Alert me if a correction is posted Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via Google Scholar Google Scholar Articles by Kubo, T. Search for Related Content PubMed PubMed Citation Articles by Kubo, T. Social Bookmarking                      What's this? Hotlight (NEW!) What's Hotlight?

Use of 3D Adaptive Raw-Data Filter in CT of the Lung: Effect on Radiation Dose Reduction

¹ Department of Radiology, Beth Israel Deaconess Medical Center, Boston, MA.
² Present address: Department of Diagnostic Imaging and Nuclear Medicine, Gradute School of Medicine, Kyoto University, Kyoto, Japan.
³ Department of Radiology, Kobe University Graduate School of Medicine, Kobe, Japan.
⁴ Department of Radiology, German Cancer Research Center, Heidelberg, Germany.
⁵ Present address: Department of Diagnostic and International Radiology, University Clinic Heidelberg, Heidelberg, Germany.
⁶ Department of Radiology, Brigham and Women's Hospital, 75 Francis St., Boston, MA 02115.

Received May 25, 2007; accepted after revision May 1, 2008.

 
Address correspondence to H. Hatabu (hhatabu{at}partners.org).

The iLEAD study project is supported by Toshiba Medical Systems.

H. U. Kauczor has a research grant and receives an honorarium from Toshiba Medical Systems

WEB

This is a Web exclusive article.

Abstract

Top Abstract Introduction Materials and Methods Results Discussion References

 
OBJECTIVE. The purpose of this study was to determine the effectiveness of a 3D adaptive raw-data filter in improving image quality and the role of the filter in radiation dose reduction in lung CT.

MATERIALS AND METHODS. Fifty-eight chest CT examinations were performed with a 16-MDCT scanner. Two acquisitions were performed with different tube current–exposure time settings (50 and 150 mAs, 120 kVp). Four series of lung images were prepared from two sets of raw data with and without application of a 3D adaptive filter (50 mAs, 50 mAs with filter, 150 mAs, 150 mAs with filter). Three blinded readers using a 5-point scale from 1 (nondiagnostic) to 5 (excellent) independently evaluated image quality in five lobes and the lingula. A set of images was considered acceptable when scores in all six regions were 3 (acceptable) or higher. The SD of attenuation was calculated in 24 regions of interest.

RESULTS. The overall mean image quality scores were 3.09, 3.53, 4.02, and 4.38 for the 50 mAs, 50 mAs with filter, 150 mAs, and 150 mAs with filter sets, respectively. Scores were significantly better with filter application (p < 0.001). A significant decrease in SD of attenuation was observed with filter application (p < 0.001). Among the respective series of images, 18, 52, 50, and 58 sets were judged acceptable with no significant difference in acceptability between images obtained at 50 mAs with a filter and at 150 mAs (p = 0.72). With filter application, the acceptability of 50-mAs images became comparable with that of 150-mAs images, making dose reduction to 50 mAs practical.

CONCLUSION. Use of a 3D adaptive raw-data filter improved the quality of lung images, making dose reduction to 50 mAs attainable with use of the filter.

Keywords: artifacts • chest CT • imaging filter • radiation dose

Introduction

Top Abstract Introduction Materials and Methods Results Discussion References

 
Excellent visualization of lung structures can be achieved with CT. Consequently, the number of CT lung examinations has been increasing. An increase in radiation exposure of patients due to this increased use of CT has raised concern about elevating the risk of development of cancer. The contribution of CT examinations to a patient's collective diagnostic radiation dose is estimated to be 67% in the United States, 47% in the United Kingdom, and 52% in Germany [1–3]. Various radiation dose–sparing techniques have been developed to diminish this detrimental effect on patients undergoing CT [4]. Image data processing is one of these techniques [5–12]. Image data processing can improve image quality without increasing the radiation dose to patients. Therefore, with the use of image data processing, we may be able to use CT protocols with reduced radiation doses to attain image quality similar to that achieved with the higher dose.

Three-dimensional adaptive raw-data filtering is an image data processing method applied to raw data (projection data) before image reconstruction [12]. Raw data can be degraded if an incoming x-ray beam traverses structures that have extremely high attenuation, resulting in streak artifacts that radiate from dense structures such as bone and metallic objects. The 3D raw-data filter removes streak artifacts by modifying parts of the raw data corrupted by the presence of highly attenuating structures. Reduction of streak artifacts with a 3D adaptive filter in low-dose chest CT, in which streak artifacts tend to be prominent, has been reported [7]. It remains to be seen whether this filter is effective for improving the quality of lung images and whether it also is useful for standard-dose images. The purpose of this study was twofold: to investigate the effectiveness of the filter in both standard-dose and reduced-dose imaging and to determine the role of the filter in radiation dose reduction in chest CT.

Materials and Methods

Top Abstract Introduction Materials and Methods Results Discussion References

 
Approval for this study was granted by our institutional review board. The study, including the data collection and review of the medical record and image data, was conducted according to the protocol authorized by the review board. Fifty-eight patients who consecutively underwent low-dose chest CT comprised the study population. All subjects provided written informed consent before the examination. The subjects were 41 men and 17 women with an age range of 56–84 years. All CT examinations were performed with a 16-MDCT scanner (Aquilion 16, Toshiba Medical Systems).

A CT image of the whole chest was obtained with a single breath-hold. With different tube current–exposure time settings, two helical acquisitions were performed for each patient with the same length of helical run and field of view to obtain two volume data sets of the whole chest. The first acquisition was performed with a tube current of 300 mA (current–time product, 150 mAs) and the second with a tube current of 100 mA (current–time product, 50 mAs). Other scan parameters were the same for both the 150-mAs and the 50-mAs scans: peak tube voltage, 120 kV; gantry speed, 0.5 s/rotation; slice collimation, 0.5 mm x 16; table feed, 7.5 mm/rotation; pitch factor 0.94. We therefore had two raw-data files of the same size for all 58 patients.

CT dose index was measured with a 32-cm acrylic dosimetry phantom and 100-mm ionization chamber. After axial imaging of the phantom was performed with a detector collimation of 0.5 mm x 16, the reading of the ionizing chamber was recorded. The measurement was repeated 10 times for each of five slots to diminish the error and effect of the fan angle. The weighted CT dose index (CTDIw) was calculated as one third of the CTDI at the center plus two thirds of CTDI at the periphery.

Qualitative Assessment of Image Quality
Lung images were reconstructed with an algorithm that enhanced high-frequency com ponents of raw data and was used for analysis of image quality in this study. A series of contiguous 2-mm-thick images was reconstructed from each of two raw-data sets (150 mAs and 50 mAs) with this standard lung reconstruction algorithm (FC 51) with and without application of the 3D adaptive filter. Thus we had 232 sets of images, 58 sets of the four image series: 50 mAs without filter processing, 50 mAs with filter processing, 150 mAs without filter processing, 150 mAs with filter processing. We used a 3D adaptive raw-data filter, which was applied only to the detector channels affected by photon starvation; no filtering process was applied otherwise. The filtering process was completed in the raw-data domain. The filter functioned in all three dimensions [12].

All 232 sets of images were made anonymous and stored randomly in a server. These images were presented to three board-certified chest radiologists blinded to scan parameters and use of filter. Images were interpreted on the diagnostic-grade LCD monitors of a PACS viewer (TPC-7200G3, Toshiba Medical Systems). The readers evaluated images with fixed window settings (level, –500 HU; width, 1,500 HU).

The three blinded readers used a 5-point scale (1, nondiagnostic; 2, poor; 3, acceptable; 4, good; 5, excellent) to independently evaluate image quality in the five lobes and lingula of all 232 image sets. Medians of the three individual scores were used for lobar quality analysis. Mean scores from six locations were calculated to represent overall image quality. To assess the adequacy of a series of images as a whole, acceptability of a set of images was defined as a consistent score of 3 (satisfactory) or greater for all six locations.

View larger version (13K): [in this window] [in a new window] [as a PowerPoint slide]   Fig. 1 —Drawing shows regions of interest. Images at four levels (lung apices, aortic arch, left ventricle, and lung bases) were selected from image sets, and six regions of interest were placed on each image: two in central part of image (1), two in anterior and posterior parts of image (2), and two in right and left parts of image (3).

We chose regions that have homogeneous structures for analysis of noise level of the images by calculation of the SD of attenuation. Circular regions of interest (ROIs) with a 16-pixel radius and con taining 709 pixels were placed within the images for analysis. Two ROIs were placed in the central part of the image and four ROIs at the periphery (Fig. 1). Central ROIs were placed within the mediastinal structures adjacent to the perihilar regions of the lungs. Two ROIs were placed in the bilateral soft-tissue structures within the chest wall that are adjacent to the peripheral part of the lungs. An additional two ROIs were placed in the anterior and posterior aspects of the body for assessment of the effect of the direction of the x-ray path on the efficiency of the adaptive noise filter. Thus six ROIs were placed on four images at the selected levels (apices, aortic arch, left ventricle, and lung bases). A total of 5,568 regions (58 cases x 4 sets x 4 levels x 6 regions) were used for analysis. The SD of attenuation in the ROIs was measured. The SDs of filter-processed images and unprocessed image were calculated in all ROIs at both 150 and 50 mAs.

Statistical Analysis
Difference in overall acceptability of images in the four series was analyzed with McNemar's test. The difference in the scores between images without and with filtering (unprocessed 50 mAs vs processed 50 mAs and unprocessed 150 mAs vs processed 150 mAs) was tested with the Wilcoxon's signed rank test. For detection of regional differences in image quality, pairs of two regions from the six total ROIs were analyzed with generalized estimating equations. The difference in SD of attenuation between filter-processed images and unprocessed images was analyzed with the paired Student's t test. A value of p < 0.05 was considered significant for all statistical tests.

Results

Top Abstract Introduction Materials and Methods Results Discussion References

 
Radiation Dose Measurement
The CT dose index was 24.5 mGy for the standard-dose (150 mAs) scans and 8.2 mGy for the low-dose (50 mAs) scans.

View larger version (140K): [in this window] [in a new window] [as a PowerPoint slide]   Fig. 2A —56-year-old man with lung nodule. Compared with low-dose (50 mAs) unprocessed CT scan (A, scored 2), filter-processed low-dose CT scan (B, scored 3) shows remarkable decrease in streak artifacts, and streak artifacts appear less severe than in standard-dose (150 mAs) image without filter processing (C, scored 4).

View larger version (132K): [in this window] [in a new window] [as a PowerPoint slide]   Fig. 2B —56-year-old man with lung nodule. Compared with low-dose (50 mAs) unprocessed CT scan (A, scored 2), filter-processed low-dose CT scan (B, scored 3) shows remarkable decrease in streak artifacts, and streak artifacts appear less severe than in standard-dose (150 mAs) image without filter processing (C, scored 4).

View larger version (127K): [in this window] [in a new window] [as a PowerPoint slide]   Fig. 2C —56-year-old man with lung nodule. Compared with low-dose (50 mAs) unprocessed CT scan (A, scored 2), filter-processed low-dose CT scan (B, scored 3) shows remarkable decrease in streak artifacts, and streak artifacts appear less severe than in standard-dose (150 mAs) image without filter processing (C, scored 4).

View larger version (122K): [in this window] [in a new window] [as a PowerPoint slide]   Fig. 2D —56-year-old man with lung nodule. Filter-processed standard-dose (150 mAs) CT scan (scored 4) also shows lessening of streak artifacts compared with C.

The acceptability of the four sets of images and the results of comparison of pairs of sets are summarized in Tables 3 and 4. There were significant differences in the number of acceptable series between images without and images with filter processing (50 mAs unprocessed vs 50 mAs processed, p < 0.001; 150 mAs unprocessed vs 150 mAs processed, p = 0.01) and between 50-mAs and 150-mAs images (50 mAs unprocessed vs 150 mAs unprocessed, p < 0.001; 50 mAs processed vs 150 mAs processed, p = 0.04). There was no difference, however, between 50-mAs images obtained with a filter and 150-mAs images without a filter (p = 0.72). A case of improvement in image quality after filter application is presented in Figure 2A, 2B, 2C, 2D.

The results of image quality comparison among the lung regions are presented in Table 5. Comparisons between corresponding regions of the right and left lungs showed a significant difference between the right middle lobe and the lingula and between the right and left lower lobes. There was no significant difference between the right and left upper lobes. Comparison between the upper lobes and other lobes (middle lobe, lingula, and lower lobes) showed the quality of images of the upper lobes was significantly worse than that of images of the other lobes. Among the middle lobes, lingula, and lower lobes, image quality was lowest for the lingula, and there was a significant difference between the lingula and the other lobes. Except for the 50-mAs images with filtering, the quality score of images of the left lower lobe was significantly lower that of images of the right middle and left lobes. Except for the 150-mAs images without filtering, there was no significant difference between the right middle lobe and lower lobe.

Quantitative Analysis of Image Quality
The SD of attenuation at the four image levels (apices, aortic arch, left ventricle, and lung bases) and their difference (filtered SD minus unfiltered SD) for low-dose (50 mAs) and standard-dose (150 mAs) images are summarized in Tables 6 and 7. For both 150- and 50-mA images, a significant decrease in SD was observed on filter-processed images in the following regions: central ROIs and right–left ROIs at the levels of the apices and lung bases, right–left ROIs at the level of the aortic arch, and anteroposterior ROIs at the level of the ventricle. For 50-mAs images, anteroposterior ROIs at the lung bases also had a significantly lower SD on filter-processed images compared with unprocessed images.

Scatterplots of the SD of attenuation for the 50-mAs images are shown in Figure 3A, 3B, 3C, 3D, 3E, 3F, 3G, 3H, 3I, 3J, 3K, 3L and for the 150-mAs images in Figure 4A, 4B, 4C, 4D, 4E, 4F, 4G, 4H, 4I, 4J, 4K, 4L. The graphs show there were two types of ROIs: ROIs with decreased SD on filter-processed images and ROIs with an almost unchanged SD on filter-processed images, which predominated at the level of the aortic arch and left ventricle and in the anteroposterior ROIs.

View larger version (13K): [in this window] [in a new window] [as a PowerPoint slide]   Fig. 3A —Scatterplots of SD values in filter-processed and unprocessed low-dose images (50 mAs) show three groups of regions of interest (ROIs) at levels of lung apices (A–C), aortic arch (D–F), left ventricle (G–I), and lung bases at diaphragmatic dome (J–L). Central ROI at level of lung apices.

View larger version (11K): [in this window] [in a new window] [as a PowerPoint slide]   Fig. 3B —Scatterplots of SD values in filter-processed and unprocessed low-dose images (50 mAs) show three groups of regions of interest (ROIs) at levels of lung apices (A–C), aortic arch (D–F), left ventricle (G–I), and lung bases at diaphragmatic dome (J–L). Anteroposterior ROI at level of lung apices.

View larger version (14K): [in this window] [in a new window] [as a PowerPoint slide]   Fig. 3C —Scatterplots of SD values in filter-processed and unprocessed low-dose images (50 mAs) show three groups of regions of interest (ROIs) at levels of lung apices (A–C), aortic arch (D–F), left ventricle (G–I), and lung bases at diaphragmatic dome (J–L). Right–left ROI at level of lung apices.

View larger version (12K): [in this window] [in a new window] [as a PowerPoint slide]   Fig. 3D —Scatterplots of SD values in filter-processed and unprocessed low-dose images (50 mAs) show three groups of regions of interest (ROIs) at levels of lung apices (A–C), aortic arch (D–F), left ventricle (G–I), and lung bases at diaphragmatic dome (J–L). Central ROI at level of aortic arch.

View larger version (12K): [in this window] [in a new window] [as a PowerPoint slide]   Fig. 3E —Scatterplots of SD values in filter-processed and unprocessed low-dose images (50 mAs) show three groups of regions of interest (ROIs) at levels of lung apices (A–C), aortic arch (D–F), left ventricle (G–I), and lung bases at diaphragmatic dome (J–L). Anteroposterior ROI at level of aortic arch.

View larger version (12K): [in this window] [in a new window] [as a PowerPoint slide]   Fig. 3F —Scatterplots of SD values in filter-processed and unprocessed low-dose images (50 mAs) show three groups of regions of interest (ROIs) at levels of lung apices (A–C), aortic arch (D–F), left ventricle (G–I), and lung bases at diaphragmatic dome (J–L). Right–left ROI at level of aortic arch.

View larger version (12K): [in this window] [in a new window] [as a PowerPoint slide]   Fig. 3G —Scatterplots of SD values in filter-processed and unprocessed low-dose images (50 mAs) show three groups of regions of interest (ROIs) at levels of lung apices (A–C), aortic arch (D–F), left ventricle (G–I), and lung bases at diaphragmatic dome (J–L). Central ROI at level of left ventricle.

View larger version (12K): [in this window] [in a new window] [as a PowerPoint slide]   Fig. 3H —Scatterplots of SD values in filter-processed and unprocessed low-dose images (50 mAs) show three groups of regions of interest (ROIs) at levels of lung apices (A–C), aortic arch (D–F), left ventricle (G–I), and lung bases at diaphragmatic dome (J–L). Anteroposterior ROI at level of left ventricle.

View larger version (12K): [in this window] [in a new window] [as a PowerPoint slide]   Fig. 3I —Scatterplots of SD values in filter-processed and unprocessed low-dose images (50 mAs) show three groups of regions of interest (ROIs) at levels of lung apices (A–C), aortic arch (D–F), left ventricle (G–I), and lung bases at diaphragmatic dome (J–L). Right–left ROI at level of left ventricle.

View larger version (12K): [in this window] [in a new window] [as a PowerPoint slide]   Fig. 3J —Scatterplots of SD values in filter-processed and unprocessed low-dose images (50 mAs) show three groups of regions of interest (ROIs) at levels of lung apices (A–C), aortic arch (D–F), left ventricle (G–I), and lung bases at diaphragmatic dome (J–L). Central ROI at level of lung bases at diaphragmatic dome.

View larger version (11K): [in this window] [in a new window] [as a PowerPoint slide]   Fig. 3K —Scatterplots of SD values in filter-processed and unprocessed low-dose images (50 mAs) show three groups of regions of interest (ROIs) at levels of lung apices (A–C), aortic arch (D–F), left ventricle (G–I), and lung bases at diaphragmatic dome (J–L). Anteroposterior ROI at level of bases at diaphragmatic dome.

View larger version (13K): [in this window] [in a new window] [as a PowerPoint slide]   Fig. 3L —Scatterplots of SD values in filter-processed and unprocessed low-dose images (50 mAs) show three groups of regions of interest (ROIs) at levels of lung apices (A–C), aortic arch (D–F), left ventricle (G–I), and lung bases at diaphragmatic dome (J–L). Right–left ROI at level of bases at diaphragmatic dome.

View larger version (11K): [in this window] [in a new window] [as a PowerPoint slide]   Fig. 4A —Scatterplots of SD values in filter-processed and unprocessed standard-dose images (150 mAs) show three groups of regions of interest (ROIs) at levels of lung apices (A–C), aortic arch (D–F), left ventricle (G–I), and lung bases at diaphragmatic dome (J–L). Central ROI at level of lung apices.

View larger version (10K): [in this window] [in a new window] [as a PowerPoint slide]   Fig. 4B —Scatterplots of SD values in filter-processed and unprocessed standard-dose images (150 mAs) show three groups of regions of interest (ROIs) at levels of lung apices (A–C), aortic arch (D–F), left ventricle (G–I), and lung bases at diaphragmatic dome (J–L). Anteroposterior ROI at level of lung apices.

View larger version (12K): [in this window] [in a new window] [as a PowerPoint slide]   Fig. 4C —Scatterplots of SD values in filter-processed and unprocessed standard-dose images (150 mAs) show three groups of regions of interest (ROIs) at levels of lung apices (A–C), aortic arch (D–F), left ventricle (G–I), and lung bases at diaphragmatic dome (J–L). Right–left ROI at level of lung apices.

View larger version (10K): [in this window] [in a new window] [as a PowerPoint slide]   Fig. 4D —Scatterplots of SD values in filter-processed and unprocessed standard-dose images (150 mAs) show three groups of regions of interest (ROIs) at levels of lung apices (A–C), aortic arch (D–F), left ventricle (G–I), and lung bases at diaphragmatic dome (J–L). Central ROI at level of aortic arch.

View larger version (10K): [in this window] [in a new window] [as a PowerPoint slide]   Fig. 4E —Scatterplots of SD values in filter-processed and unprocessed standard-dose images (150 mAs) show three groups of regions of interest (ROIs) at levels of lung apices (A–C), aortic arch (D–F), left ventricle (G–I), and lung bases at diaphragmatic dome (J–L). Anteroposterior ROI at level of aortic arch.

View larger version (11K): [in this window] [in a new window] [as a PowerPoint slide]   Fig. 4F —Scatterplots of SD values in filter-processed and unprocessed standard-dose images (150 mAs) show three groups of regions of interest (ROIs) at levels of lung apices (A–C), aortic arch (D–F), left ventricle (G–I), and lung bases at diaphragmatic dome (J–L). Right–left ROI at level of aortic arch.

View larger version (10K): [in this window] [in a new window] [as a PowerPoint slide]   Fig. 4G —Scatterplots of SD values in filter-processed and unprocessed standard-dose images (150 mAs) show three groups of regions of interest (ROIs) at levels of lung apices (A–C), aortic arch (D–F), left ventricle (G–I), and lung bases at diaphragmatic dome (J–L). Central ROI at level of left ventricle.

View larger version (11K): [in this window] [in a new window] [as a PowerPoint slide]   Fig. 4H —Scatterplots of SD values in filter-processed and unprocessed standard-dose images (150 mAs) show three groups of regions of interest (ROIs) at levels of lung apices (A–C), aortic arch (D–F), left ventricle (G–I), and lung bases at diaphragmatic dome (J–L). Anteroposterior ROI at level of left ventricle.

View larger version (11K): [in this window] [in a new window] [as a PowerPoint slide]   Fig. 4I —Scatterplots of SD values in filter-processed and unprocessed standard-dose images (150 mAs) show three groups of regions of interest (ROIs) at levels of lung apices (A–C), aortic arch (D–F), left ventricle (G–I), and lung bases at diaphragmatic dome (J–L). Right–left ROI at level of left ventricle.

View larger version (11K): [in this window] [in a new window] [as a PowerPoint slide]   Fig. 4J —Scatterplots of SD values in filter-processed and unprocessed standard-dose images (150 mAs) show three groups of regions of interest (ROIs) at levels of lung apices (A–C), aortic arch (D–F), left ventricle (G–I), and lung bases at diaphragmatic dome (J–L). Central ROI at level of lung bases at diaphragmatic dome.

View larger version (11K): [in this window] [in a new window] [as a PowerPoint slide]   Fig. 4K —Scatterplots of SD values in filter-processed and unprocessed standard-dose images (150 mAs) show three groups of regions of interest (ROIs) at levels of lung apices (A–C), aortic arch (D–F), left ventricle (G–I), and lung bases at diaphragmatic dome (J–L). Anteroposterior ROI at level of bases at diaphragmatic dome.

View larger version (11K): [in this window] [in a new window] [as a PowerPoint slide]   Fig. 4L —Scatterplots of SD values in filter-processed and unprocessed standard-dose images (150 mAs) show three groups of regions of interest (ROIs) at levels of lung apices (A–C), aortic arch (D–F), left ventricle (G–I), and lung bases at diaphragmatic dome (J–L). Right–left ROI at level of bases at diaphragmatic dome.

Top Abstract Introduction Materials and Methods Results Discussion References

The filter used in this study works on raw (projection) data. Raw data produced with MDCT have three dimensions (detector channels, projection angle, and detector location on the z-axis) and can be represented as a series of sinograms. On sinograms, points representing structures of extremely high attenuation can be identified. At these points, few incident photons are received at the detector, and this phenomenon is responsible for streak artifacts on reconstructed images. The filter corrects the data corrupted by insufficient photon input by interpolating neighboring data. The filter does not change data elsewhere, thus it can reduce streak artifacts while spatial resolution elsewhere is preserved. Combining this filter with reduced tube current may make it possible to reduce radiation dose without compromising image quality.

The results of this study showed that image quality at both 50 mAs and 150 mAs improved significantly with application of an adaptive filter. A preliminary study [7] showed that an adaptive raw-data filter reduced streak artifacts on low-dose chest CT (25 mAs, 120 kVp) in which streak artifacts are prominent. The results of our study showed that improvement in quality also is observed on images obtained with higher tube current–time settings (50 mAs and 150 mAs). This finding suggests that the filter can be applied routinely in the processing of chest CT images without regard for the scan parameters used for a particular examination.

Our study also showed that use of the 3D adaptive raw-data filter makes radiation dose reduction possible with retention of acceptability by radiologists. There are several methods of image modification for improvement of CT image quality [5–12]. These methods are expected to be useful in reduced-dose CT examinations. Whether and how these techniques can contribute to actual radiation dose reduction has not been investigated, to our knowledge. Our study showed that with use of the filter, the acceptability of low-dose (50 mAs) images was comparable with that of standard-dose (150 mAs) images. The results thus showed imaging filters have a role as an adjunctive method of radiation dose reduction.

Quantitative analysis of image quality in ROIs showed a decrease in SD of attenuation on filter-processed images, emphasizing the improvement in image quality found at qualitative analysis. The result also showed the fundamental characteristic of filter function, that is, the adaptive behavior of the filter. This adaptive filter is supposed to modify images (or pixel attenuation) when streak artifacts are expected to become prominent. The apparent conservative performance of this filter at the level at which streak artifacts are less likely to occur, such as the aortic arch and ventricles, is well explained by this adaptive filter function. In fact, averaging of the images where there are no visible streak artifacts degrades rather than improves image quality because of the smoothing effect on small structures such as the peripheral pulmonary vessels. Selective artifact suppression in conjunction with retention of anatomic detail may explain why this filter was successful in improving the subjective image quality rating by the radiologists.

The results of this study showed a difference in image quality between the upper lobes and other lobes, the quality of images of the upper lobes being significantly lower than that of images of the other lobes. This difference occurred because the upper lobes of the lungs are surrounded by bony structures of the shoulder girdle and therefore are affected by severe streak artifacts. The decline in quality of images of the upper lobes was consistent at both 50 and 150 mAs and with and without filtering. Use of automatic exposure control techniques in conjunction with the 3D adaptive filter should improve the quality of images of the upper lobes. The results also showed that the quality of images of the lingula and left lower lobe was significantly lower than that of images of the right middle and lower lobes. This difference was probably due to motion artifacts, which are most prominent in the lingula, where lung structures are most likely to be blurred by pulsation of cardiovascular structures.

This study had several limitations. Diagnostic quality was not assessed. Reduced-dose CT images can be used for diagnosis if the result of interpretation is the same as that of standard-dose CT images, even though there is a difference in image quality between the two. The results of previous studies [13, 14] of image quality and diagnostic quality showed that degradation of image quality can occur at tube current settings higher than those compromising the capability of depicting abnormalities. Therefore, given the image quality improvement in our study, it can be assumed that the abnormal findings detectable on the unprocessed images are also detectable on filter-processed images, although direct comparison of the diagnoses with both standard-dose and reduced-dose CT would be a better method of comparison than comparison of image quality for this purpose.

In this study, the quality of mediastinal (soft-tissue) images was not evaluated. It has been shown [8] that the adaptive raw-data filter is effective in artifact removal from low-dose soft-tissue images. Effective artifact removal is even more challenging on lung than on mediastinal images. One of the reasons is that an edge-enhancing reconstruction algorithm is used for reconstruction of lung images; as a result, the images are more affected by marked streak artifacts. The other reason is that overly aggressive artifact removal can decrease the definition of small lung structures. Therefore, the filter function should be minimal when streak artifacts are not present. For these reasons, we used lung images for analysis of filter effectiveness in artifact suppression.

A 3D adaptive raw-data filter was found to enhance the quality of both low-dose CT images and standard-dose CT images of the lung parenchyma. Use of the filter significantly improved the acceptability of 50-mAs images, which compared favorably with the acceptability of 150-mAs images. Use of a 3D adaptive raw-data filter can contribute to reduction of the radiation dose in chest CT examinations. The filter can be used routinely for both reduced-dose and standard-dose CT examinations.

Acknowledgments
 
The International Multicenter Study for Low-Dose Chest CT Examination and Diagnosis (iLEAD) is dedicated to the investigation of radiation dose reduction in chest CT examinations. Three institutions currently participate this study group: Beth Israel Deaconess Medical Center, Boston, MA; German Cancer Research Center, Heidelberg, Germany; and Kobe University, Kobe, Japan. The iLEAD study group consists of Takeshi Kubo, Boston, MA; Yoshiharu Ohno, Kobe, Japan; Shiva Gautam, Boston, MA; Pei-Jan P. Lin, Boston, MA; Masaya Takahashi, Boston, MA; Mizuki Nishino, Boston, MA; Arkadiusz Sitek, Boston, MA; Daisuke Takenaka, Kobe, Japan; Munenobu Nogami, Kobe, Japan; Hisanobu Koyama, Kobe, Japan; Julia Ley-Zaporozhan, Heidelberg, Germany; Wolfram Stiller, Heidelberg, Germany; Sebastian Ley, Heidelberg, Germany; Hiroyasu Inokawa, Tochigi, Japan; Miwa Okumura, Tochigi, Japan; Yasuko Fujisawa, Tochigi, Japan; Hiroyuki Kura, Tochigi, Japan; Toshihiro Rifu, Tochigi, Japan; Vassilios Raptopoulos, Boston, MA; Kazuro Sugimura, Kobe, Japan; Hans-Ulrich Kauczor, Heidelberg, Germany; and Hiroto Hatabu, Boston, MA.

References

Top Abstract Introduction Materials and Methods Results Discussion References

 

1. Mettler FA Jr, Wiest PW, Locken JA, Kelsey CA. CT scanning: patterns of use and dose. J Radiol Prot2000; 20:353 –359[CrossRef][Medline]
2. Hart D, Wall BF. UK population dose from medical x-ray examinations. Eur J Radiol2004; 50:285 –291[CrossRef][Medline]
3. Federal Office for Radiation Protection. Environmental radioactivity and radiation exposure in the year 2005 (parliamentary report). www.bfs.de/en/bfs/druck/uus/parlamentsbericht03.pdf. Salzgitter, Germany: Federal Office for Radiation Protection, 2005. Accessed June 28, 2008
4. Tack D, Gevenois PA. Radiation dose in computed tomography of the chest. JBR-BTR2004; 87:281 –288
5. Kubo T, Ohno Y, Stiller W, et al. Radiation dose reduction in chest CT: a review. AJR 2008;190 : 335–343[Abstract/Free Full Text]
6. Kalra MK, Wittram C, Maher MM, et al. Can noise reduction filters improve low-radiation-dose chest CT images? Pilot study. Radiology 2003;228 : 257–264[Abstract/Free Full Text]
7. Kachelriess M, Watzke O, Kalender WA. Generalized multi-dimensional adaptive filtering for conventional and spiral single-slice, multi-slice, and cone-beam CT. Med Phys2001; 28:475 –490[CrossRef][Medline]
8. Kubo T, Nishino M, Kino A, et al. 3-Dimensional adaptive raw-data filter: evaluation in low dose chest multidetector-row computed tomography. J Comput Assist Tomogr 2006;30 : 933–938[CrossRef][Medline]
9. Okumura M, Ota T, Tsukagoshi S, Katada K. New method of evaluating edge-preserving adaptive filters for computed tomography (CT): digital phantom method. Nippon Hoshasen Gijutsu Gakkai Zasshi2006; 62:971 –978[Medline]
10. Kalra MK, Maher MM, Sahani DV, et al. Low-dose CT of the abdomen: evaluation of image improvement with use of noise reduction filters—pilot study. Radiology2003; 228:251 –256[Abstract/Free Full Text]
11. Sasaki T, Hanari T, Sasaki M, et al. Reduction of radiation exposure in CT perfusion study using a quantum de-noising filter. Nippon Hoshasen Gijutsu Gakkai Zasshi2004; 60:1688 –1693[Medline]
12. Kazama M, Tsukagoshi S, Okumura M. Image quality improvement and exposure dose reduction with the combined use of X-ray modulation and Boost3D. In: Flynn MJ, Hsieh J, eds. Medical imaging 2006: physics of medical imaging, vol. 6142. Bellingham, WA: SPIE, 2006:847 –855
13. Mayo JR, Hartman TE, Lee KS, Primack SL, Vedal S, Muller NL. CT of the chest: minimal tube current required for good image quality with the least radiation dose. AJR 1995;164 : 603–607[Abstract/Free Full Text]
14. Takahashi M, Maguire WM, Ashtari M, et al. Low-dose spiral computed tomography of the thorax: comparison with the standard-dose technique. Invest Radiol 1998;33 : 68–73[CrossRef][Medline]

CiteULike

Complore

Connotea

Del.icio.us

Digg

Reddit

Technorati

This Article Abstract Figures Only Full Text (PDF) Alert me when this article is cited Alert me if a correction is posted Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via Google Scholar Google Scholar Articles by Kubo, T. Search for Related Content PubMed PubMed Citation Articles by Kubo, T. Social Bookmarking                      What's this? Hotlight (NEW!) What's Hotlight?