New method of evaluating edge-preserving adaptive filters for computed tomography (CT): digital phantom method

The evaluation of the reduction of radiation dose via deep learning-based reconstruction for cadaveric human lung CT images

To compare the quality of CT images of the lung reconstructed using deep learning-based reconstruction (True Fidelity Image: TFI ™; GE Healthcare) to filtered back projection (FBP), and to determine the minimum tube current-time product in TFI without compromising image quality. Four cadaveric human lungs were scanned on CT at 120 kVp and different tube current-time products (10, 25, 50, 75, 100, and 175 mAs) and reconstructed with TFI and FBP. Two image evaluations were performed by three independent radiologists. In the first experiment, using the same tube current-time product, a side-by-side TFI and FBP comparison was performed. Images were evaluated with regard to noise, streak artifacts, and overall image quality. Overall image quality was evaluated in view of whole image quality. In the second experiment, CT images reconstructed using TFI and FBP with five different tube current-time products were displayed in random order, which were evaluated with reference to the 175 mAs-FBP image. Images were scored with regard to normal structure, abnormal findings, noise, streak artifacts, and overall image quality. Median scores from three radiologists were statistically analyzed. Quantitative evaluation of noise was performed by setting regions of interest (ROIs) in air. In first experiment, overall image quality was improved, and noise was decreased in images of TFI compared to that of FBP for all tube current-time products. In second experiment, scores of all evaluation items except for small vessels in images of 25 mAs-TFI were almost the same as that of 175 mAs-FBP (all p > 0.31). Using TFI instead of FBP, at least 85% radiation dose reduction could be possible without any degradation in the image quality.

A phantom study on usefulness of modifying image parameters to reduce radiation exposure and maintain image quality in chest HRCT

PURPOSE

To investigate the feasibility of reducing radiation dose by modifying tube voltage, window settings, and algorithm while maintaining image quality, based on the qualitative evaluation of its quality and the radiation dose, using raw data acquired in chest high-resolution computed tomography (HRCT).

METHODS

Radiation exposure was measured using a Fluke dosimeter while modifying the tube voltage to 80 and 100 from 120 kVp in a 64-slice multi-detector computed tomography for comparison and analysis. Changes in image quality as a result of the different tube voltage settings, 3 different window settings (-550, -600, and -700), and 2 algorithms (standard and edge) were analyzed using ImageJ.

RESULTS

Relative to 120kVp, the dose decreased by approximately 67.8% and 36.9% at 80 and 100 kVp, respectively. Image quality assessment showed that changing the window setting to -700 (window level) after scanning with the tube voltage set at 100 kVp and applying the edge algorithm reduced the radiation dose while maintaining the image quality.

CONCLUSIONS

The findings are significant with respect to the reduction of scan dose in that they demonstrate how radiation exposure can be reduced in a clinical scenario by altering the settings on an existing HRCT apparatus. Additional clinical trials and image assessments should be conducted on human participants to confirm the feasibility of altering HRCT settings for reducing scan doses.

Usefulness of Ultralow-Dose (Submillisievert) Chest CT Using Iterative Reconstruction for Initial Evaluation of Sharp Fish Bone Esophageal Foreign Body

OBJECTIVE

The purpose of this article was to evaluate the usefulness of ultralow-dose chest CT as an initial imaging study for evaluation of sharp fish bone esophageal foreign body (FB).

MATERIALS AND METHODS

A total of 57 subjects who underwent ultralow-dose chest CT were included in this retrospective study. All subjects had a history of ingestion and symptoms of esophageal FB. All ultralow-dose chest CT data were reconstructed twice, once with filtered back projection (FBP) and once with iterative reconstruction, and three observers reviewed the images independently. ROC analysis was used to evaluate diagnostic performance of ultralow-dose chest CT. Intraclass correlation coefficient (ICC) was calculated for analysis of interobserver agreement.

RESULTS

Among 57 patients, 42 were confirmed as having esophageal FB. Significant objective noise reduction of mediastinum was achieved using an iterative reconstruction technique. Subjective image noise of iterative reconstruction was significantly better than that of FBP. Overall diagnostic performance of ultralow-dose chest CT for esophageal FB of iterative reconstruction (AUC = 0.999) was significantly better than that of FBP (AUC = 0.95) (p = 0.02). Interobserver agreement was greater for iterative reconstruction (ICC = 0.944) than for FBP (ICC = 0.778).

CONCLUSION

Ultralow-dose chest CT using iterative reconstruction provided satisfactory diagnostic image quality for identifying fish bone esophageal FB with reduced radiation dose and high observer accuracy. Therefore, ultralow-dose chest CT would be adequate as a first-line imaging modality for fish bone esophageal FB.

Radiation dose reduction for chest CT with non-linear adaptive filters

BACKGROUND

CT radiation dose reduction results in increased noise or graininess of images which affects the diagnostic information. One of the approaches to lower radiation exposure to patients is to reduce image noise with the use of image processing software in low radiation dose images.

PURPOSE

To assess image quality and accuracy of non-linear adaptive filters (NLAF) at low dose chest CT.

MATERIAL AND METHODS

In an IRB approved prospective study, 24 patients (mean age, 63 ± 7.3 years; M:F ratio, 11:13) gave informed consent for acquisition of four additional chest CT image series at 150, 110, 75, and 40 mAs (baseline image series) on a 64-slice MDCT over an identical 10-cm length. NLAF was used to process three low dose (110, 75, and 40 mAs) image series (postprocessed image series). Two radiologists reviewed baseline and postprocessed images in a blinded manner for image quality. Objective noise, CT attenuation values, patient weight, transverse diameters, CTDIvol, and DLP were recorded. Statistical analysis was performed using parametric and non-parametric tests for comparing postprocessed and baseline images.

RESULTS

No lesions were missed on baseline or postprocessed CT images (n = 80 lesions, 73 lesions <1 cm). At 40 mAs, subjective noise in mediastinal window settings were graded as unacceptable in baseline images and acceptable in postprocessed images. Visibility of smaller structures improved from suboptimal visibility in baseline images at 40 mAs to excellent in postprocessed images at 40 mAs. No major artifacts were seen due to NLAF postprocessing, except for minor beam hardening artifacts not affecting diagnostic decision-making (14/22) in both baseline and postprocessed image series. Diagnostic confidence for chest CT was improved to fully confident in postprocessed images at 40 mAs. Compared to baseline images, postprocessing reduced objective noise by 26% (14.2 ± 4.7/19.2 ± 6.4), 31.5% (15.2 ± 4.7/22.2 ± 5.7), and 41.5% (16.9 ± 6/28.9 ± 10.2) at 110 mAs, 75 mAs, and 40 mAs tube current-time product levels.

CONCLUSION

Applications of NLAF can help reduce tube current down to 40 mAs for chest CT while maintaining lesion conspicuity and image quality.

[Novel method of noise power spectrum measurement for computed tomography images with adaptive iterative reconstruction method]

Adaptive iterative reconstruction techniques (IRs) can decrease image noise in computed tomography (CT) and are expected to contribute to reduction of the radiation dose. To evaluate the performance of IRs, the conventional two-dimensional (2D) noise power spectrum (NPS) is widely used. However, when an IR provides an NPS value drop at all spatial frequency (which is similar to NPS changes by dose increase), the conventional method cannot evaluate the correct noise property because the conventional method does not correspond to the volume data natures of CT images. The purpose of our study was to develop a new method for NPS measurements that can be adapted to IRs. Our method utilized thick multi-planar reconstruction (MPR) images. The thick images are generally made by averaging CT volume data in a direction perpendicular to a MPR plane (e.g. z-direction for axial MPR plane). By using this averaging technique as a cutter for 3D-NPS, we can obtain adequate 2D-extracted NPS (eNPS) from 3D NPS. We applied this method to IR images generated with adaptive iterative dose reduction 3D (AIDR-3D, Toshiba) to investigate the validity of our method. A water phantom with 24 cm-diameters was scanned at 120 kV and 200 mAs with a 320-row CT (Acquilion One, Toshiba). From the results of study, the adequate thickness of MPR images for eNPS was more than 25.0 mm. Our new NPS measurement method utilizing thick MPR images was accurate and effective for evaluating noise reduction effects of IRs.

Radiation dose reduction with application of non-linear adaptive filters for abdominal CT

AIM: To evaluate the effect of non-linear adaptive filters (NLAF) on abdominal computed tomography (CT) images acquired at different radiation dose levels.

METHODS: Nineteen patients (mean age 61.6 ± 7.9 years, M:F = 8:11) gave informed consent for an Institutional Review Board approved prospective study involving acquisition of 4 additional image series (200, 150, 100, 50 mAs and 120 kVp) on a 64 slice multidetector row CT scanner over an identical 10 cm length in the abdomen. The CT images acquired at 150, 100 and 50 mAs were processed with the NLAF. Two radiologists reviewed unprocessed and processed images for image quality in a blinded randomized manner. CT dose index volume, dose length product, patient weight, transverse diameters, objective noise and CT numbers were recorded. Data were analyzed using Analysis of Variance and Wilcoxon signed rank test.

RESULTS: Of the 31 lesions detected in abdominal CT images, 28 lesions were less than 1 cm in size. Subjective image noise was graded as unacceptable in unprocessed images at 50 and 100 mAs, and in NLAF processed images at 50 mAs only. In NLAF processed images, objective image noise was decreased by 21% (14.4 ± 4/18.2 ± 4.9) at 150 mAs, 28.3% (15.7 ± 5.6/21.9 ± 4) at 100 mAs and by 39.4% (18.8 ± 9/30.4 ± 9.2) at 50 mAs compared to unprocessed images acquired at respective radiation dose levels. At 100 mAs the visibility of smaller structures improved from suboptimal in unprocessed images to excellent in NLAF processed images, whereas diagnostic confidence was respectively improved from probably confident to fully confident.

CONCLUSION: NLAF lowers image noise, improves the visibility of small structures and maintains lesion conspicuity at down to 100 mAs for abdominal CT.

Development of a noise reduction filter algorithm for pediatric body images in multidetector CT

Recently, several types of post-processing image filter which was designed to reduce noise allowing a corresponding dose reduction in CT images have been proposed and these were reported to be useful for noise reduction of CT images of adult patients. However, these have not been reported on adaptation for pediatric patients. Because they are not very effective with small (<20 cm) display fields of view, they could not be used for pediatric (e.g., premature babies and infants) body CT images. In order to solve this restriction, we have developed a new noise reduction filter algorithm which can be applicable for pediatric body CT images. This algorithm is based on a three-dimensional post processing, in which output pixel values are calculated by multi-directional, one-dimensional median filters on original volumetric datasets. The processed directions were selected except in in-plane (axial plane) direction, and consequently the in-plane spatial resolution was not affected by the filter. Also, in other directions, the spatial resolutions including slice thickness were almost maintained due to a characteristic of non-linear filtering of the median filter. From the results of phantom studies, the proposed algorithm could reduce standard deviation values as a noise index by up to 30% without affecting the spatial resolution of all directions, and therefore, contrast-to-noise ratio was improved by up to 30%. This newly developed filter algorithm will be useful for the diagnosis and radiation dose reduction of pediatric body CT images.

[Radiation dose reduction using a non-linear image filter in MDCT]

The development of MDCT enabled various high-quality 3D imaging and optimized scan timing with contrast injection in a multi-phase dynamic study. Since radiation dose tends to increase to yield such high-quality images, we have to make an effort to reduce radiation dose. A non-linear image filter (Neuro 3D filter: N3D filter) has been developed in order to improve image noise. The purpose of this study was to evaluate the physical performance and effectiveness of this non-linear image filter using phantoms, and show how we can reduce radiation dose in clinical use of this filter. This N3D filter reduced radiation dose by about 50%, with minimum deterioration of spatial reduction in phantom and clinical studies. This filter shows great potential for clinical application.

The reduction of image noise and streak artifact in the thoracic inlet during low dose and ultra-low dose thoracic CT

Increased pixel noise and streak artifact reduce CT image quality and limit the potential for radiation dose reduction during CT of the thoracic inlet. We propose to quantify the pixel noise of mediastinal structures in the thoracic inlet, during low-dose (LDCT) and ultralow-dose (uLDCT) thoracic CT, and assess the utility of new software (quantum denoising system and BOOST3D) in addressing these limitations. Twelve patients had LDCT (120 kV, 25 mAs) and uLDCT (120 kV, 10 mAs) images reconstructed initially using standard mediastinal and lung filters followed by the quantum denoising system (QDS) to reduce pixel noise and BOOST3D (B3D) software to correct photon starvation noise as follows: group 1 no QDS, no B3D; group 2 B3D alone; group 3 QDS alone and group 4 both QDS and B3D. Nine regions of interest (ROIs) were replicated on mediastinal anatomy in the thoracic inlet, for each patient resulting in 3456 data points to calculate pixel noise and attenuation. QDS reduced pixel noise by 18.4% (lung images) and 15.8% (mediastinal images) at 25 mAs. B3D reduced pixel noise by approximately 8% in the posterior thorax and in combination there was a 35.5% reduction in effective radiation dose (E) for LDCT (1.63-1.05 mSv) in lung images and 32.2% (1.55-1.05 mSv) in mediastinal images. The same combination produced 20.7% reduction (0.53-0.42 mSv) in E for uLDCT, for lung images and 17.3% (0.51-0.42) for mediastinal images. This quantitative analysis of image quality confirms the utility of dedicated processing software in targeting image noise and streak artifact in thoracic LDCT and uLDCT images taken in the thoracic inlet. This processing software potentiates substantial reductions in radiation dose during thoracic LDCT and uLDCT.

[Evaluation of an adaptive filter for CT under low-CNR condition: comparison with linear filter]

The use of an adaptive filter for CT images is becoming a common procedure and is said to reduce image noise while preserving sharpness and helping to reduce the required X-ray dose. Although many reports support this view, the validity of such evaluations is arguable. When the linearity of a system is in question, physical performance indexes should be measured under conditions similar to those of clinical use. Evaluations of diagnosis using clinical images may be fallible because the non-filtered image used as the reference might not have been optimally reconstructed. We have chosen simple, but commonly used, adaptive filters for our evaluation. As a reference for comparing performance, we designed linear filters that best approximate the noise characteristics of the adaptive filters. MTF is measured through observation of the edge-spread function. Clinical abdominal images are used to compare the performance of adaptive filters and linear filters. We conclude that the performance of the type of adaptive filter we have chosen is virtually the same as that of the linear filter, as long as the image quality of soft tissues is our interest. Both the noise SD and MTF are virtually the same if the contrast of the object is not substantially higher than 150 HU. Images of soft tissues obtained with the use of adaptive filters are also virtually the same as those obtained by linear filters. The edge-preservation characteristic of this adaptive filter is not observable for soft tissues.

Effect of edge-preserving adaptive image filter on low-contrast detectability in CT systems: application of ROC analysis

Objective. For the multislice CT (MSCT) systems with a larger number of detector rows, it is essential to employ dose-reduction techniques. As reported in previous studies, edge-preserving adaptive image filters, which selectively eliminate only the noise elements that are increased when the radiation dose is reduced without affecting the sharpness of images, have been developed. In the present study, we employed receiver operating characteristic (ROC) analysis to assess the effects of the quantum denoising system (QDS), which is an edge-preserving adaptive filter that we have developed, on low-contrast resolution, and to evaluate to what degree the radiation dose can be reduced while maintaining acceptable low-contrast resolution. Materials and Methods. The low-contrast phantoms (Catphan 412) were scanned at various tube current settings, and ROC analysis was then performed for the groups of images obtained with/without the use of QDS at each tube current to determine whether or not a target could be identified. The tube current settings for which the area under the ROC curve (Az value) was approximately 0.7 were determined for both groups of images with/without the use of QDS. Then, the radiation dose reduction ratio when QDS was used was calculated by converting the determined tube current to the radiation dose. Results. The use of the QDS edge-preserving adaptive image filter allowed the radiation dose to be reduced by up to 38%. Conclusion. The QDS was found to be useful for reducing the radiation dose without affecting the low-contrast resolution in MSCT studies.

[Evaluation of non-linear adaptive smoothing filter by digital phantom]

As a result of the development of multi-slice CT, diagnoses based on three-dimensional reconstruction images and multi-planar reconstruction have spread. For these applications, which require high z-resolution, thin slice imaging is essential. However, because z-resolution is always based on a trade-off with image noise, thin slice imaging is necessarily accompanied by an increase in noise level. To improve the quality of thin slice images, a non-linear adaptive smoothing filter has been developed, and is being widely applied to clinical use. We developed a digital bar pattern phantom for the purpose of evaluating the effect of this filter and attempted evaluation from an addition image of the bar pattern phantom and the image of the water phantom. The effect of this filter was changed in a complex manner by the contrast and spatial frequency of the original image. We have confirmed the reduced effect of image noise in the low frequency component of the image, but decreased contrast or increased quantity of noise in the image of the high frequency component. This result represents the effect of change in the adaptation of this filter. The digital phantom was useful for this evaluation, but to understand the total effect of filtering, much improvement of the shape of the digital phantom is required.

Improvement in image quality of noncontrast head images in multidetector-row CT by volume helical scanning with a three-dimensional denoising filter

PURPOSE

The aim of this study was to improve the contrast-to-noise ratio on noncontrast head computed tomography (CT) images, which are crucial for assessing patients with acute ischemic stroke. We applied a technique combining volume helical scanning with a three-dimensional (3D) denoising filter.

MATERIALS AND METHODS

We scanned phantoms for low-contrast resolutions and helical/cone-beam artifacts as well as stroke patients using a 16-row multidetector-row CT (MDCT) unit. Volume helical scans with 1-mm collimation and nonhelical scans with 8-mm thickness were performed. From the 1-mm thick volume data, 8-mm thick contiguous images were generated before and after applying a 3D denoising filter.

RESULTS

On images stacked from volume data, the contrast-to-noise ratio was significantly improved by the 3D denoising filter and was nearly the same as that on nonhelical images. On stacked volume images, artifacts due to the cone beam and the helical scan were increased with larger helical pitches, but bone-related streak artifacts in the posterior fossa and underneath the calvarium were reduced when compared with nonhelical images.

CONCLUSION

Volume helical scan with a 3D denoising filter effectively improves image quality in noncontrast head MDCT images.