CT Image Quality Research Articles

Evaluation of Abdominal CT Obtained Using a Deep Learning-Based Image Reconstruction Engine Compared with CT Using Adaptive Statistical Iterative Reconstruction.

Purpose:To compare the image quality of CT obtained using a deep learning-based image reconstruction (DLIR) engine with images with adaptive statistical iterative reconstruction-V (AV).Materials and Methods:Using a phantom, the noise power spectrum (NPS) and task-based transfer function (TTF) were measured in images with different reconstructions (filtered back projection [FBP], AV30, 50, 100, DLIR-L, M, H) at multiple doses. One hundred and twenty abdominal CTs with 30% dose reduction were processed using AV30, AV50, DLIR-L, M, H. Objective and subjective analyses were performed.Results:The NPS peak of DLIR was lower than that of AV30 or AV50. Compared with AV30, the NPS average spatial frequencies were higher with DLIR-L or DLIR-M. For lower contrast objects, TTF in images with DLIR were higher than those with AV. The standard deviation in DLIR-H and DLIR-M was significantly lower than AV30 and AV50. The overall image quality was the best for DLIR-M (p < 0.001).Conclusions:DLIR showed improved image quality and decreased noise under a decreased radiation dose.

CT image quality enhancement via a dual-channel neural network with jointing denoising and super-resolution

In recent years, computed tomography (CT) has been widely used in various clinical diagnosis. Given potential health risks bring by the X-ray radiation, the major objective of the current research is to achieve high-quality CT imaging while reducing X-ray radiation. However, most existing studies on low-dose CT image super-resolution reconstruction do not focus on the interaction between the denoising task and the super-resolution task. In this paper, we propose a dual-channel joint learning framework to accurately reconstruct high-resolution CT images from low-resolution CT images. Unlike the previous cascaded models which directly combine the denoising network and the super-resolution network, our method can process the denoising reconstruction and the super-resolution reconstruction in parallel. Additionally, we design a filter gate module that can filter features from the denoising branch and highlight important features which can benefit the super-resolution task. We evaluate the performance of our method in medical image enhancement by testing on the 2016 Low-Dose CT Grand Challenge dataset and the piglet dataset. The experimental results show that the proposed network is superior to other state-of-the-art methods in terms of both peak signal-to-noise ratio (PSNR) and structural similarity index (SSIM). We also demonstrate that our method can better remove noise and recover details. Furthermore, the method achieves competitive results not only for super-resolution reconstruction of low-dose CT, but also for super-resolution reconstruction of sparse-view CT.

Improvement of Quantitative Single-Photon Emission Computed Tomography Image Quality by the New Step-and-Shoot Scan Mode

The step-and-shoot (SS) mode and continuous mode are currently used for single-photon emission computed tomography (SPECT) scan mode, and a new scan mode that combines both modes, step-and-shoot plus continuous (SSC) mode, was developed. It is expected to allow a shorter scan time and lower injected dose because the SSC mode is more sensitive than the SS mode. We confirmed the image quality of this scan mode, including various quantitative correction methods for scatter (SC), attenuation (AC), and resolution recovery (RR) in a phantom study and clinical case study. Image quality was evaluated by the count, contrast-to-noise ratio (CNR), and percent of the coefficient of variation (%CV). Independent of the correction methods, the count, CNR, and %CV of the SSC mode were superior to those of the SS mode. The ACSCRR was the best method, with a maximum increased rate of 66.4% in counts and 57.8% in CNR for the 13-mm sphere and 19.6% in CNR for other sphere sizes. The %CV for the SSC mode was the best for AC and ACRR, which was at 15.1%. With regards to attaining short bone SPECT scan time, the combination of the SSC mode and ACRR or ACSCRR demonstrated the best physical performance.

Harmonization of technical image quality in computed tomography: comparison between different reconstruction algorithms and kernels from six scanners

Purpose. The radiology department faces a large number of reconstruction algorithms and kernels during their computed tomography (CT) optimization process. These reconstruction methods are proprietary and ensuring consistent image quality between scanners is becoming increasingly difficult. This study contributes to solving this challenge in CT image quality harmonization by modifying and evaluating a reconstruction algorithm and kernel matching scheme. Methods. The Catphan 600 phantom was scanned with six different CT scanners from four vendors. The phantom was scanned with volumetric CT dose indices (CTDIvols) of 10 mGy and 40 mGy, and the data were reconstructed using 1 mm and 5 mm slices with each combination of reconstruction algorithm, body region kernel, and iterative and deep learning reconstruction strength. A matching scheme developed in previous research, which utilizes the noise power spectrum (NPS) and modulation transfer function (MTF), was modified based on our organization’s needs and used to identify the matching reconstruction algorithms and kernels between different scanners. Results. The matching paradigm produced good matching results, and the mean ± standard deviation (median) matching function values for the different acquisition settings were (a value of 1 indicates a perfect match): CTDIvol 10 mGy, 1 mm slice: 0.78 ± 0.31 (0.94); CTDIvol 10 mGy, 5 mm slice: 0.75 ± 0.33 (0.93); CTDIvol 40 mGy, 1 mm slice: 0.81 ± 0.28 (0.95); CTDIvol 40 mGy, 5 mm slice: 0.75 ± 0.33 (0.93). In general, soft reconstruction kernels, i.e., noise-reducing kernels that reduce sharpness, of one vendor were matched with the soft kernels of another vendor, and vice versa for sharper kernels. Conclusions. Combined quantitative assessment of NPS and MTF allows effective strategy for harmonization of technical image quality between different CT scanners. A software was also shared to support CT image quality harmonization in other institutions.

Quantitative assessment of motion effects in dual-source dual-energy CT and dual-source photon-counting detector CT.

Conventional dual-source CT scanners can be used to either provide better temporal resolution or dual-energy imaging, but not both at the same time. This presents a dilemma in cardiac CT as both high temporal resolution and multi-energy imaging are desirable. The current study evaluated a dual-source photon-counting-detector (DS-PCD) CT which can acquire multi-energy images at high temporal resolution. A cardiac motion phantom with a 3-mm diameter iodinated rod, mimicking the right coronary artery, was scanned 25 times using a DS-PCD CT (66 ms resolution) and a dual-source dual-energy (DS-DE, 125 ms resolution) CT. Low/high energy images and iodine maps were reconstructed at 40% and 75% cardiac phases. To quantify the impact of motion on image quality, dice similarity coefficient was computed between the low/high energy images while the circularity and effective diameter of the iodinated rod were computed on the iodine maps. The dice coefficients were higher for DS-PCD with a mean of 0.89 and 0.91 at the 40% and 70% phases, while DS-DE had a lower mean of 0.20 and 0.78, respectively. The circularity was excellent for DS-PCD with a mean of 0.97 and 0.98 at the 40% and 75% phases, while DS-DE had a mean of 0.71 and 0.98, respectively. The effective diameter was accurate for DS-PCD with a mean of 2.9 mm (true size of 3 mm) at both phases, while DS-DE had a mean of 4.0 mm and 3.2 mm at the 40% and 75% phases, respectively. These results indicate that DS-PCD CT enables simultaneous high temporal resolution and multi-energy cardiac imaging with minimal motion artifacts.

Control of Variance and Bias in CT Image Processing with Variational Training of Deep Neural Networks.

Optimization of CT image quality typically involves balancing noise and bias. In filtered back-projection, this trade-off is controlled by the particular filter and cutoff frequency. In penalized-likelihood iterative reconstruction, the penalty weight serves the same function. Deep neural networks typically do not provide this tuneable control over output image properties. Models are often trained to minimize mean squared error which penalizes both variance and bias in image outputs but does not offer any control over the trade-off between the two. In this work, we propose a method for controlling the output image properties of neural networks with a new loss function called weighted covariance and bias (WCB). Our proposed method includes separate weighting parameters to control the relative importance of noise or bias reduction. Moreover, we show that tuning these weights enables targeted penalization of specific image features (e.g. spatial frequencies). To evaluate our method, we present a simulation study using digital anthropormorphic phantoms, physical simulation of non-ideal CT data, and image formation with various algorithms. We show that WCB offers a greater degree of control over trade-offs between variance and bias whereas MSE has only one configuration. We also show that WCB can be used to control specific image properties including variance, bias, spatial resolution, and the noise correlation of neural network outputs. Finally, we present a method to optimize the proposed weights for stimulus detectability. Our results demonstrate the potential for this new capability to control the image properties of DNN outputs and optimize image quality for the task-specific applications.

AI Denoising Significantly Enhances Image Quality and Diagnostic Confidence in Interventional Cone-Beam Computed Tomography.

(1) To investigate whether interventional cone-beam computed tomography (cbCT) could benefit from AI denoising, particularly with respect to patient body mass index (BMI); (2) From 1 January 2016 to 1 January 2022, 100 patients with liver-directed interventions and peri-procedural cbCT were included. The unenhanced mask run and the contrast-enhanced fill run of the cbCT were reconstructed using weighted filtered back projection. Additionally, each dataset was post-processed using a novel denoising software solution. Place-consistent regions of interest measured signal-to-noise ratio (SNR) per dataset. Corrected mixed-effects analysis with BMI subgroup analyses compared objective image quality. Multiple linear regression measured the contribution of “Radiation Dose”, “Body-Mass-Index”, and “Mode” to SNR. Two radiologists independently rated diagnostic confidence. Inter-rater agreement was measured using Spearman correlation (r); (3) SNR was significantly higher in the denoised datasets than in the regular datasets (p < 0.001). Furthermore, BMI subgroup analysis showed significant SNR deteriorations in the regular datasets for higher patient BMI (p < 0.001), but stable results for denoising (p > 0.999). In regression, only denoising contributed positively towards SNR (0.6191; 95%CI 0.6096 to 0.6286; p < 0.001). The denoised datasets received overall significantly higher diagnostic confidence grades (p = 0.010), with good inter-rater agreement (r ≥ 0.795, p < 0.001). In a subgroup analysis, diagnostic confidence deteriorated significantly for higher patient BMI (p < 0.001) in the regular datasets but was stable in the denoised datasets (p ≥ 0.103).; (4) AI denoising can significantly enhance image quality in interventional cone-beam CT and effectively mitigate diagnostic confidence deterioration for rising patient BMI.

Radiation dose optimization potential of deep learning-based reconstruction for multiphase hepatic CT: A clinical and phantom study

PurposeThis clinical and phantom study aimed to evaluate the impact of deep learning-based reconstruction (DLR) on image quality and its radiation dose optimization capability for multiphase hepatic CT relative to hybrid iterative reconstruction (HIR). MethodsTask-based image quality was assessed with a physical evaluation phantom; the high- and low-contrast detectability of HIR and DLR images were computed from the noise power spectrum and task-based transfer function at five different size-specific dose estimate (SSDE) values in the range 5.3 to 18.0-mGy. For the clinical study, images of 73 patients who had undergone multiphase hepatic CT under both standard-dose (STD) and lower-dose (LD) examination protocols within a time interval of about four-months on average, were retrospectively examined. STD images were reconstructed with HIR, while LD with HIR (LD-HIR) and DLR (LD-DLR). SSDE, quantitative image noise, and contrast-to-noise ratio (CNR) were compared between protocols. The noise magnitude, noise texture, streak artifact, image sharpness, interface smoothness, and overall image quality were subjectively rated by two independent radiologists. ResultsIn phantom study, the high- and low-contrast detectability of DLR images obtained at 5.3-mGy and 7.3-mGy, respectively, were slightly higher than those obtained with HIR at the STD protocol dose (18.0-mGy). In clinical study, LD-DLR yielded lower image noise, higher CNR, and higher subjective scores for all evaluation criteria than STD (all, p ≤ 0.05), despite having 52.8% lower SSDE (8.0 ± 2.5 vs. 16.8 ± 3.4-mGy). ConclusionsDLR improved the subjective and objective image quality of multiphase hepatic CT compared with HIR techniques, even at approximately half the radiation dose.

Effect of 3D Printing Material and Thickness on CT Image Quality and Shielding

In order to proceed with an accurate CT examination and to prevent unnecessary exposure during CT examination, the need for an appropriate fixing device and shielding device is required. In this study, 3D printing technology, a next-generation technology, is applied to radiation science to investigate the effect on CT images and shielding effects through filaments that can be easily obtained using a general 3D printer of the melt-stacking modeling method. Using a 3D printer, a phantom made of four materials: PLA (Poly Lactic Acid), Bronze, Carbon, and eSteel was produced cylindrical shape. Statistical analysis was performed using a non-parametric test method, and the significance level (α) was set to 0.05 (5%). As a result of the measurement, the PLA filament material had the least effect on the CT image quality. After placing the dosimeter inside the phantom, when the shielding rate was measured for each material through scanning, the eSteel material had the highest shielding rate and showed a shielding rate equivalent to or higher than that of the lead shielding device currently used in clinical practice.

Performance analysis of improved hybrid median filter applied to X-ray computed tomography images obtained with high-resolution photon-counting CZT detector: A pilot study

We evaluated the performance of an improved hybrid median filter (IHMF) applied to X-ray computed tomography (CT) images obtained using a high-resolution photon-counting cadmium zinc telluride (CZT) detector. To study how the proposed approach improves the image quality, we measured the noise levels and the overall CT-image quality. We established a CZT imaging system with a detector length of 5.12 cm and thickness of 0.3 cm and acquired phantom images. To evaluate the efficacy of the proposed filter, we first modeled two conventional median filters. Subsequently, we were able to achieve a normalized noise power spectrum result of ∼10−8 mm2, and furthermore, the proposed method improved the contrast-to-noise ratio by a factor of ∼1.51 and the coefficient of variation by 1.55 relative to the counterpart values of the no-filter image. In addition, the IHMF exhibited the best performance among the three filters considered as regards the peak signal-to-noise ratio and no-reference-based image-quality evaluation parameters. Thus, our results demonstrate that the IHMF approach provides a superior image performance over conventional median filtering methods when applied to actual CZT X-ray CT images.

The Image Quality and Diagnostic Performance of CT with Low-Concentration Iodine Contrast (240 mg Iodine/mL) for the Abdominal Organs.

Purpose: To evaluate the difference between CT examinations using 240 mgI/mL contrast material (CM) and 320 mgI/mL CM in the contrast enhancement of the abdominal organs and the diagnostic performance for focal hepatic lesions. Materials and methods: This retrospective study included 422 CT examinations, using 240 mgI/mL iohexol (Group A, 206 examinations) and 320 mgI/mL ioversol (Group B, 216 examinations), performed between April 2019 and May 2020. Two CT scanners (single-source CT (machine A) and dual-source CT (machine B)) were used to obtain CT images. Two radiologists independently drew regions of interest (ROIs) in the liver, pancreas, spleen, kidney, aorta, portal vein, and paraspinal muscle. The signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) were calculated for each organ. They evaluated the degree of subjective enhancement of the organs and detected/differentiated focal hepatic lesions. Results: The SNR, CNR, and subjective enhancement of most organs were significantly higher in Group B than in Group A (p < 0.05). The sensitivity and specificity for cysts and malignancy were higher than 85.0% in both groups. The sensitivity for hemangioma was lower in Group B (<75%) than in Group A. In Group A, the SNR and CNR were significantly higher in most organs with machine B than with machine A. Conclusion: Although the SNR and CNR of the abdominal organs were lower with 240 mgI/mL CM than with 320 mgI/mL CM, 240 mgI/mL CM was feasible for evaluating the liver. A CT scanner with more advanced specifications may be beneficial for examinations with 240 mgI/mL CM by using lower tube voltage.

Impact of pixel value truncation on image quality of low dose chest CT.

In some noisy low dose CT lung cancer screening images, we noticed that the CT density values of air were increased and the visibility of emphysema was distinctly decreased. By examining histograms of these images, we found that the CT density values were truncated at -1024 HU. The purpose of this study was to investigate the effect of pixel value truncation on the visibility of emphysema using mathematical models. Assuming CT noise follows a normal distribution, we derived the relationship between the mean CT density value and the standard deviation (SD) when the pixel values below -1024 HU are truncated and replaced by -1024 HU. To validate our mathematical model, 20 untruncated phantom CT images were truncated by simulation, and the mean CT density values and SD of air in the images were measured and compared with the theoretical values. In addition, the mean CT density values and SD of air were measured in 100 cases of real clinical images obtained by GE, Siemens, and Philips scanners, respectively, and the agreement with the theoretical values was examined. Next, the contrast-to-noise ratio (CNR) between air (-1000 HU) and lung parenchyma (-850 HU) was derived from the mathematical model in the presence and absence of truncation as a measure of the visibility of emphysema. In addition, the radiation dose ratios required to obtain the same CNR in the case with and without truncation were also calculated. The mathematical model revealed that when the pixel values are truncated, the mean CT density values are proportional to the noise magnitude when the magnitude exceeds a certain level. The mean CT density values and SD measured in the images with pixel values truncated by simulation and in the real clinical images acquired by GE and Philips scanners agreed well with the theoretical values from our mathematical model. In the Siemens images, the measured and theoretical values agreed well when a portion of the truncated values were replaced by random values instead of simply replacing by -1024 HU. The CNR of air and lung parenchyma was lowered by truncating CT density values compared to that of no truncation. Furthermore, it was found that higher radiation dose was required to obtain the same CNR with truncation as without. As an example, when the noise SD was 60 HU, the radiation dose required for the GE and Philips truncation method was about 1.2 times higher than that without truncation, and that for the Siemens truncation method was about 1.4 times higher. It was demonstrated mathematically that pixel value truncation causes a brightening of the mean CT density value and decreases the CNR of emphysema. Our results indicate that it is advisable to turn off truncation at -1024 HU, especially when scanning at low and ultra-low radiation doses in the thorax.

Framelet tensor sparsity with block matching for spectral CT reconstruction.

Spectral computed tomography (CT) based on the photon-counting detection system has the capability to produce energy-discriminative attenuation maps of objects with a single scan. However, the insufficiency of photons collected into the narrow energy bins results in high quantum noise levels causing low image quality. This work aims to improve spectral CT image quality by developing a novel regularization based on framelet tensor prior. First, similar patches are extracted from highly correlated interchannel images in spectral and spatial domains, and stacked to form a third-order tensor after vectorization along the energy channels. Second, the framelet tensor nuclear norm (FTNN) is introduced and applied to construct the regularization to exploit the sparsity embedded in nonlocal similarity of spectral images, and thus the reconstruction problem is modeled as a constrained optimization. Third, an iterative algorithm is proposed by utilizing the alternating direction method of multipliers framework in which efficient solvers are developed for each subproblem. Both numerical simulations and real data verifications were performed to evaluate and validate the proposed FTNN-based method. Compared to the analytic, TV-based, and the state-of-the-art tensor-based methods, the proposed method achieves higher numerical accuracy on both reconstructed CT images and decomposed material maps in the mouse data indicating the capability in noise suppression and detail preservation of the proposed method. A framelet tensor sparsity-based iterative algorithm is proposed for spectral reconstruction. The qualitative and quantitative comparisons show a promising improvement of image quality, indicating its promising potential in spectral CT imaging.

Reduction of CT Artifacts Using Polyetheretherketone (PEEK), Polyetherketoneketone (PEKK), Polyphenylsulfone (PPSU), and Polyethylene (PE) Reconstruction Plates in Oral Oncology

Surgical reconstruction after segmental mandibulectomy in patients with oral cancer commonly requires implantation of titanium plates, which cause metallic artifacts in CT examinations detrimental for image quality. Therefore, the aim of the present study was to evaluate CT artifacts and image quality of the polymer plate materials polyetheretherketone (PEEK), polyetherketoneketone (PEKK), polyphenylsulfone (PPSU), and polyethylene (PE) in mandible reconstruction. Five plates (titanium, PEEK, PEKK, PPSU, and PE) were successively surgically inserted in a human cadaveric specimen using eight (1) titanium and (2) zirconium oxide screws. Streak and blooming artifacts were assessed in 1mm CT scans with and without metallic artifact reduction algorithms (MAR) in this cross-sectional study. The different plate materials were analyzed for their effect on Hounsfield unit (HU) values resulting in streak artifacts. Blooming artifacts were measured as virtual growth of plates compared to actual size. Subjective image quality was assessed by 3 independent observers. Statistics were conducted by multifactorial ANOVA and Bonferroni's post hoc test. Titanium (182.42 HU; SD±3.09) was associated with significantly more streak and blooming artifacts compared to the polymers PEEK (86.41 HU; SD±1.72; P<.001), PEKK (85.50 HU; SD±2.75; P<.001), PPSU (85.51 HU; SD±1.97; P<.001), and PE (85.52 HU; SD±3.33; P<.001) implants. No significant differences could be seen between the screw types (titanium vs zirconium oxide) (P=.75) or the activation or inactivation of MAR (P=.85). Image quality was best for PE and worst for titanium implants in combination with zirconium oxide screws. Polymers plate materials PEEK, PEKK, PPSU, and PE significantly reduce CT artifacts and improve image quality facilitating the radiological interpretation of implant surrounding tissues. This effect is beneficial for radiological tumor follow-up and CT-based planning of adjuvant radiotherapy.

Deep learning image reconstruction for improving image quality of contrast-enhanced dual-energy CT in abdomen.

To evaluate the usefulness of deep learning image reconstruction (DLIR) to improve the image quality of dual-energy computed tomography (DECT) of the abdomen, compared to hybrid iterative reconstruction (IR). This study included 40 patients who underwent contrast-enhanced DECT of the abdomen. Virtual monochromatic 40-, 50-, and 70-keV and iodine density images were reconstructed using three reconstruction algorithms, including hybrid IR (ASiR-V50%) and DLIR (TrueFidelity) at medium- and high-strength level (DLIR-M and DLIR-H, respectively). The standard deviation of attenuation in liver parenchyma was measured as image noise. The contrast-to-noise ratio (CNR) for the portal vein on portal venous phase CT was calculated. The vessel conspicuity and overall image quality were graded on a 5-point scale ranging from 1 (poor) to 5 (excellent). The comparative scale of lesion conspicuity in 47 abdominal solid lesions was evaluated on a 5-point scale ranging from 0 (best) to -4 (markedly inferior). The image noise of virtual monochromatic 40-, 50 -, and 70-keV and iodine density images was significantly decreased by DLIR compared to hybrid IR (p < 0.0001). The CNR was significantly higher in DLIR-H and DLIR-M than in hybrid IR (p < 0.0001). The vessel conspicuity and overall image quality scores were also significantly greater in DLIR-H and DLIR-M than in hybrid IR (p < 0.05). The lesion conspicuity scores for DLIR-M and DLIR-H were significantly higher than those for hybrid IR in the virtual monochromatic image of all energy levels (p ≤ 0.001). DLIR improves vessel conspicuity, CNR, and lesion conspicuity of virtual monochromatic and iodine density images in abdominal contrast-enhanced DECT, compared to hybrid IR. • Deep learning image reconstruction (DLIR) is useful for reducing image noise and improving the CNR of visual monochromatic 40-, 50-, and 70-keV images in dual-energy CT. • DLIR can improve lesion conspicuity of abdominal solid lesions on virtual monochromatic images compared to hybrid iterative reconstruction. • DLIR can also be applied to iodine density maps and significantly improves their image quality.

The Effects of Additional Filtration on Image Quality and Radiation Dose in Cone Beam CT: An In Vivo Preliminary Investigation.

Purpose The aim of this study was to investigate the effect of reduced radiation doses on the image quality of cone-beam computed tomography scans and the suitability of such imaging for orthodontics, oral surgery, dental implantology, periodontology, and endodontology. Materials and Methods Cone-beam computed tomography scans of a live patient were performed using seven attenuation filters with increased thickness to decrease the effective radiation dose from 22.4 to 1.8 μSv, and the effects of different radiation doses on image quality were further analysed. Quantitative image quality was calculated using dedicated measures, such as signal and contrast-to-noise ratio and sharpness. A panel of five certified raters assessed the cone-beam computed tomography scans qualitatively. Nine anatomical structures relevant to dentistry were identified, and the overall acceptance was assessed. Results Linear reduction of the effective radiation dose had a nonlinear effect on image quality. A 5-fold reduction in the effective dose led to acceptable quantitative and qualitative image quality measures, and the identification rate of dental anatomical structures was 80% or greater. The use of less than 40% of the reference dose was unacceptable for all dental specialties. Conclusions The ideal radiation dose for specific diagnostic requirements remains a patient-related and specialty-related decision that must be made on an individual basis. Based on the results of this study, it is possible to reduce exposure in selected patients, and at the same time obtain sufficient quality of images for clinical purposes.

A CT image feature space (CTIS) loss for restoration with deep learning-based methods

Objective. Deep learning-based methods have been widely used in medical imaging field such as detection, segmentation and image restoration. For supervised learning methods in CT image restoration, different loss functions will lead to different image qualities which may affect clinical diagnosis. In this paper, to compare commonly used loss functions and give a better alternative, we studied a widely generalizable framework for loss functions which are defined in the feature space extracted by neural networks. Approach. For the purpose of incorporating prior knowledge, a CT image feature space (CTIS) loss was proposed, which learned the feature space from high quality CT images by an autoencoder. In the absence of high-quality CT images, an alternate loss function, random-weight (RaW) loss in the feature space of images (LoFS) was proposed. For RaW-LoFS, the feature space is defined by neural networks with random weights. Main results. In experimental studies, we used post reconstruction deep learning-based methods in the 2016 AAPM low dose CT grand challenge. Compared with perceptual loss that is widely used, our loss functions performed better both quantitatively and qualitatively. In addition, three senior radiologists were invited for subjective assessments between CTIS loss and RaW-LoFS. According to their judgements, the results of CTIS loss achieved better visual quality. Furtherly, by analyzing each channel of CTIS loss, we also proposed partially constrained CTIS loss. Significance. Our loss functions achieved favorable image quality. This framework can be easily adapted to other tasks and fields.

Evaluation of the impact of a scanner prototype on proton CT and helium CT image quality and dose efficiency with Monte Carlo simulation

Objective. The use of ion computed tomography (CT) promises to yield improved relative stopping power (RSP) estimation as input to particle therapy treatment planning. Recently, proton CT (pCT) has been shown to yield RSP accuracy on par with state-of-the-art x-ray dual energy CT. There are however concerns that the lower spatial resolution of pCT compared to x-ray CT may limit its potential, which has spurred interest in the use of helium ion CT (HeCT). The goal of this study was to investigate image quality of pCT and HeCT in terms of noise, spatial resolution, RSP accuracy and imaging dose using a detailed Monte Carlo (MC) model of an existing ion CT prototype. Approach. Three phantoms were used in simulated pCT and HeCT scans allowing estimation of noise, spatial resolution and the scoring of dose. An additional phantom was used to evaluate RSP accuracy. The imaging dose required to achieve the same image noise in a water and a head phantom was estimated at both native spatial resolution, and in a scenario where the HeCT spatial resolution was reduced and matched to that of pCT using Hann windowing of the reconstruction filter. A variance reconstruction formalism was adapted to account for Hann windowing. Main results. We confirmed that the scanner prototype would produce higher spatial resolution for HeCT than pCT by a factor 1.8 (0.86 lp mm−1 versus 0.48 lp mm−1 at the center of a 20 cm water phantom). At native resolution, HeCT required a factor 2.9 more dose than pCT to achieve the same noise, while at matched resolution, HeCT required only 38% of the pCT dose. Finally, RSP mean absolute percent error (MAPE) was found to be 0.59% for pCT and 0.67% for HeCT. Significance. This work compared the imaging performance of pCT and HeCT when using an existing scanner prototype, with the spatial resolution advantage of HeCT coming at the cost of increased dose. When matching spatial resolution via Hann windowing, HeCT had a substantial dose advantage. Both modalities provided state-of-the-art RSP MAPE. HeCT might therefore help reduce the dose exposure of patients with comparable image noise to pCT, enhanced spatial resolution and acceptable RSP accuracy at the same time.

Feasibility of a method for low contrast CT image quality assessment using difference detail curves for abdominal scans

This work describes a measurement method for assessing dose-related image-quality of CT scans based on the difference detail curve (DDC) method, and showcases its use in a low contrast setting. The method is based on a phantom consisting of elliptical slices of different sizes into which contrast object modules can be inserted. These modules contain contrast objects based on (synthetic) resin mixtures with sucrose (native) or sodium iodine (contrast medium). Mixing ratios are provided to achieve a range of clinically relevant CT-numbers with these materials. The phantom is characterized in terms of contrast accuracy, energy dependency and long-term drift with satisfying results. Contrast accuracy and energy dependency are similar to that of water or soft tissue. Image quality of 655 scans of the phantom acquired at 30 different clinical institutions and with 16 different CT scanner models from 4 manufacturers was assessed by calculating a difference detail curve (DDC) from evaluation of up to 5 human observers using a custom-made software (RadiVates) described in this work. Based on these measurements, inter-observer variability was quantified using a bootstrap method and was shown to be a large contributor to the overall variability. This work demonstrates that assessment of CT image quality is feasible with the aforementioned phantom and DDC method.

Deep learning reconstruction improves radiomics feature stability and discriminative power in abdominal CT imaging: a phantom study

ObjectivesTo compare image quality of deep learning reconstruction (AiCE) for radiomics feature extraction with filtered back projection (FBP), hybrid iterative reconstruction (AIDR 3D), and model-based iterative reconstruction (FIRST).MethodsEffects of image reconstruction on radiomics features were investigated using a phantom that realistically mimicked a 65-year-old patient’s abdomen with hepatic metastases. The phantom was scanned at 18 doses from 0.2 to 4 mGy, with 20 repeated scans per dose. Images were reconstructed with FBP, AIDR 3D, FIRST, and AiCE. Ninety-three radiomics features were extracted from 24 regions of interest, which were evenly distributed across three tissue classes: normal liver, metastatic core, and metastatic rim. Features were analyzed in terms of their consistent characterization of tissues within the same image (intraclass correlation coefficient ≥ 0.75), discriminative power (Kruskal-Wallis test p value < 0.05), and repeatability (overall concordance correlation coefficient ≥ 0.75).ResultsThe median fraction of consistent features across all doses was 6%, 8%, 6%, and 22% with FBP, AIDR 3D, FIRST, and AiCE, respectively. Adequate discriminative power was achieved by 48%, 82%, 84%, and 92% of features, and 52%, 20%, 17%, and 39% of features were repeatable, respectively. Only 5% of features combined consistency, discriminative power, and repeatability with FBP, AIDR 3D, and FIRST versus 13% with AiCE at doses above 1 mGy and 17% at doses ≥ 3 mGy. AiCE was the only reconstruction technique that enabled extraction of higher-order features.ConclusionsAiCE more than doubled the yield of radiomics features at doses typically used clinically. Inconsistent tissue characterization within CT images contributes significantly to the poor stability of radiomics features.Key Points• Image quality of CT images reconstructed with filtered back projection and iterative methods is inadequate for the majority of radiomics features due to inconsistent tissue characterization, low discriminative power, or low repeatability.• Deep learning reconstruction enhances image quality for radiomics and more than doubled the feature yield at doses that are typically used in clinical CT imaging.• Image reconstruction algorithms can optimize image quality for more reliable quantification of tissues in CT images.