CT Reconstruction Research Articles

Prevalence and Clinical Implications of Pulmonary Vein Stenosis in Bronchiectasis: A 3D Reconstruction CT Study

Background: Recent studies on bronchiectasis have revealed significant structural abnormalities and pathophysiological changes. However, there is limited research focused on pulmonary venous variability and congenital variation. Through our surgical observations, we noted that coarctation of pulmonary veins and atrophied lung volume are relatively common in bronchiectasis patients. Therefore, we conducted a retrospective study to explore pulmonary venous variation and secondary manifestations in bronchiectasis cases, utilizing 3D reconstruction software (Mimics Innovation Suite 21.0, Materialise Dental, Leuven, Belgium) to draw conclusions supported by statistical evidence. Method: This retrospective study included patients with bronchiectasis and healthy individuals who underwent CT examinations at Beijing Chao-Yang Hospital between January 2017 and July 2023. Chest CT data were reconstructed using Materialise Mimics. Pulmonary veins and lung lobes were segmented from surrounding tissue based on an appropriate threshold determined by local grey values and image gradients. Subsequently, venous cross-sectional areas and lung volumes were measured for statistical analysis. Result: CT data from 174 inpatients with bronchiectasis and 75 cases from the health examination center were included. Three-dimensional reconstruction data revealed a significant reduction in cross-sectional areas of pulmonary veins in the left lower lobe (p < 0.001), the right lower lobe (p = 0.030), and the right middle lobe (p = 0.009) of bronchiectasis patients. Subgroup analyses indicated that approximately 73.5% of localized cases of the left lower lobe exhibited pulmonary vein stenosis, while in the diffuse group, this proportion was only 52.6%. Furthermore, the cross-sectional area of pulmonary veins had a gradually decreasing trend, based on a small sample. Lung function tests showed significant reductions in FEV1, FVC, and FEV1% in bronchiectasis patients, attributed to the loss of lung volume in the left lower lobe, which accounted for 60.9% of the included sample. Conclusions: Our recent findings suggest that pulmonary venous stenosis is a common variation in bronchiectasis and is often observed concurrently with reduced lung volume, particularly affecting the left lower lobe. Moreover, localized cases are more likely to suffer from pulmonary venous stenosis, with an ambiguous downtrend as the disease progresses. In conclusion, increased attention to pulmonary venous variation in bronchiectasis is warranted, and exploring new therapies to intervene in the early stages or alleviate obstruction may be beneficial.

Influence analysis of glenohumeral bone structure on anterior shoulder instability

To investigate the effect of glenohumeral bone structure on anterior shoulder instability by three-dimensional CT reconstruction. The clinical data of 48 patients with unilateral anterior shoulder dislocation (instability group) and 46 patients without shoulder joint disease (control group) admitted between February 2012 and January 2024 were retrospectively analyzed. There was no significant difference in gender and side between the two groups ( P>0.05). The patients were significantly younger in the instability group than in the control group ( P<0.05). The glenoid joint morphological parameters such as glenoid height, glenoid width, ratio of glenoid height to width, glenoid inclination, the humeral containing angle, and glenoid version were measured on three-dimensional CT reconstruction of the glenoid. The differences of the above indexes between the two groups were compared, and the differences of the above indexes between the two groups were compared respectively in the male and the female. Random forest model was used to analyze the influencing factors of anterior shoulder instability. The comparison between the two groups and the comparison between the two groups in the male and the female showed that the ratio of of the instability group glenoid height to width was larger than that of the control group, the glenoid width and humeral containing angle were smaller than those of the control group, and the differences were significant ( P<0.05); there was no significant difference in glenoid height, glenoid inclination, and glenoid version between the two groups ( P>0.05). The accuracy of the random forest model was 0.84. The results showed that the top four influencing factors of anterior shoulder instability were ratio of glenoid height to width, the humeral containing angle, age, and glenoid width. Ratio of glenoid height to width and the humeral containing angle are important influencing factors of anterior shoulder instability.

A prediction method of thermal deformation of near-zero warping sandwich structure under non-uniform temperature field

The thermal deformation of structure in non-uniform thermal environment in space is a critical factor affecting the imaging accuracy of remote sensing satellite. In this study, the thermal deformation theory of near-zero warping sandwich structure under non-uniform temperature field is established. A high-fidelity prediction method based on CT scanning and reconstruction of structural defects is proposed. Thermal deformation experiments show that this method can accurately characterize the ultra-low thermal deformation behavior of the near-zero warping sandwich metastructure, which is significant for the real-time correction of satellite imaging in orbit.

State-of-the-Art Deep Learning CT Reconstruction Algorithms in Abdominal Imaging.

The implementation of deep neural networks has spurred the creation of deep learning reconstruction (DLR) CT algorithms. DLR CT techniques encompass a spectrum of deep learning-based methodologies that operate during the different steps of the image creation, prior to or after the traditional image formation process (eg, filtered backprojection [FBP] or iterative reconstruction [IR]), or alternatively by fully replacing FBP or IR techniques. DLR algorithms effectively facilitate the reduction of image noise associated with low photon counts from reduced radiation dose protocols. DLR methods have emerged as an effective solution to ameliorate limitations observed with prior CT image reconstruction algorithms, including FBP and IR algorithms, which are not able to preserve image texture and diagnostic performance at low radiation dose levels. An additional advantage of DLR algorithms is their high reconstruction speed, hence targeting the ideal triad of features for a CT image reconstruction (ie, the ability to consistently provide diagnostic-quality images and achieve radiation dose imaging levels as low as reasonably possible, with high reconstruction speed). An accumulated body of evidence supports the clinical use of DLR algorithms in abdominal imaging across multiple CT imaging tasks. The authors explore the technical aspects of DLR CT algorithms and examine various approaches to image synthesis in DLR creation. The clinical applications of DLR algorithms are highlighted across various abdominal CT imaging domains, with emphasis on the supporting evidence for diverse clinical tasks. An overview of the current limitations of and outlook for DLR algorithms for CT is provided. ©RSNA, 2024.

Image-based finite element model stiffness and vBMD by single and dual energy CT reconstruction kernel

Single-energy quantitative computed tomography (SEQCT) provides volumetric bone mineral density (vBMD) measures for bone analysis and input to image-based finite element models (FEMs). Dual-energy CT (DECT) improves vBMD by accounting for voxel-specific material variations utilizing scans at multiple x-ray energies. vBMD is also altered by reconstruction kernel that cannot be accounted for using calibration phantoms. This study compared vBMD and FEM stiffness derived from SEQCT and DECT images reconstructed with two common kernels. SEQCT and DECT images of cadaveric shoulders (n = 10) were collected using standard (STD) and boneplus (BONE) kernels. Hounsfield Units were converted to vBMD using specimen-specific calibrations. DECT STD and BONE images were generated using an established material decomposition method with 40 and 90 keV simulated monochromatic images. A proximal humerus bone section below the anatomic neck was used for vBMD analysis and FEM generation. FEMs were loaded to 1 % apparent strain for stiffness measurements.Between STD and BONE kernel images, average vBMD differed 0.9 mgK2HPO4/cc and 4.1 mg K2HPO4/cc, in SEQCT and DECT images, respectively. Significant differences occurred in DECT images (p = 0.001). BONE reconstructed images produced higher vBMD measures across both SEQCT and DECT images. The difference between STD and BONE in both SEQCT- and DECT-based FEMs persisted, with larger estimated stiffness in BONE models. For six of the models DECT-based had higher stiffness than SEQCT-based models using the same kernel, although these models differed between STD and BONE kernels. Differences in stiffness between STD and BONE derived models were similar across image types (DECT: 17.5 kN/mm; SEQCT: 19.0 kN/mm). Stiffness values were significantly different within SECT kernels and between SEQCT BONE and DECT STD models. This study shows important differences in vBMD and FEM stiffness that occur due to CT-based imaging parameters alone. These results indicate that consistent imaging parameters should be used for vBMD analysis and FEM input to avoid systematic measurement errors.

Dual-energy CT of acute bowel ischemia-influence on diagnostic accuracy and reader confidence.

This study evaluates the advantages in diagnostic accuracy, confidence, and reading time of additional dual-energy CT-derived reconstructions for assessing acute bowel ischemia. This retrospective study includes 25 patients with surgically proven acute bowel ischemia and 25 gender- and age-matched controls who underwent biphasic abdominal dual-energy CT. Two fellowship-trained abdominal radiologists and two residents evaluated all cases with and without additional dual-energy CT-derived iodine maps and virtual non-contrast images. Diagnostic confidence was rated on a 10-point Likert scale. Reading time was recorded. The inter-reader agreement was assessed using Fleiss' kappa. Sensitivity and specificity were compared using McNemar's test, reader confidence, and reading times with the Wilcoxon signed-rank test. For conventional images alone, the inter-reader agreement was moderate (κ = 0.58), with a sensitivity of 77% (95% CI: 67.5-84.8%) and specificity of 90% (95% CI: 82.4-95.1%). Adding dual-energy CT images, inter-reader agreement increased to substantial (κ = 0.69), sensitivity increased significantly to 89% (95% CI: 81.2-94.4%, p = 0.02), while specificity increased non-significantly to 93% (95% CI: 86.1-97.1%, p = 0.51). Diagnostic confidence increased significantly from 8 (IQR: 6-8) to 9 (IQR: 8-9) (p < 0.01). Equivalent diagnostic accuracy and confidence increases were observed for fellowship-trained and resident radiologists. A non-significant increase in mean reading time per case from 196 s to 201 s was observed (p = 0.30). Additional dual-energy CT reconstructions increase diagnostic accuracy and confidence without increasing reading time when evaluating suspected acute bowel ischemia. Both experienced and resident readers benefit from dual-energy CT images. Question There are too few clinical studies assessing the diagnostic accuracy of dual-energy CT (DECT) to recommend its use for evaluating suspected acute bowel ischemia. Findings Adding DECT-derived iodine maps and virtual-non-contrast images increase reader sensitivity and confidence while maintaining high specificity when evaluating for acute mesenteric ischemia. Clinical relevance Dual-energy CT should be used to investigate suspected acute bowel ischemia. Both diagnostic accuracy and confidence can be increased independent of reader experience without significantly increasing reading time.

Quantification of normal bone and osseous metastases in castration-resistant prostate cancer using SPECT/CT with xSPECT Quant: prospective imaging sub-study of a phase 2 clinical trial investigating the combination of pembrolizumab plus radium-223 compared to radium-223 alone.

The purpose of this study is to demonstrate the consistency and reproducibility of quantitative SPECT/CT by evaluating the maximum SUV (SUVmax) in normal bone, to provide the reference value of metastatic lesions, and to evaluate the clinical implication of SUVmax changes of osseous metastasis during treatment. This prospective imaging sub-study was performed as part of a phase 2 clinical trial of patients with metastatic castration-resistant prostate cancer (mCRPC) randomized to the combination of pembrolizumab plus radium-223 or to radium-223 alone (NCT03093428). The maximum standardized uptake value (SUVmax) and mean Hounsfield Unit (HUmean) of normal bone as well as metastases were measured using a 1.5cm region of interest (ROI) on CT and xSPECT Quant reconstruction on the baseline study (S0) and restaging scans. The most tracer-avid metastatic lesion in each patient on S0 was selected as a target lesion, and changes of SUVmax and HUmean of the target lesion were compared on the first restaging scan (S1). Correlations between the percentage changes of SUVmax of the target lesion with alkaline phosphatase (ALP) and prostate-specific antigen (PSA) were assessed. Twenty-one patients were enrolled on the imaging sub-study of which 15 had paired baseline S0 and S1 data. On S0, the median SUVmax and HUmean of normal bone was 5.85g/mL (0.42-14.98) and 133.03 (range, 28.47-461.91), respectively. The median SUVmax and HUmean of metastasis were 42.2g/mL (range, 17.96-143.36) and 549.58 (177.87-1107.64), respectively. There was significant reduction in SUVmax (- 40.1%, range - 86.2 to + 23.5%), p < 0.001) and increase in HUmean (+ 8.3%, range - 11.3 to + 61.7%, p = 0.0479, Wilcoxon signed-rank test) of target lesions between S0 and S1. Spearman correlation between the percentage changes of SUVmax of a target lesion and both serum PSA (r = 0.33, p = 0.226) and ALP (r = 0.45, p = 0.094) were not statistically significant. Quantitative SPECT/CT provides consistent and objective imaging parameters, which can help monitor tumor burden. The median SUVmax of metastasis at baseline was roughly 7.2-fold higher than normal bone. Quantitative SPECT/CT may help visualize the early osteoblastic treatment response in prostate cancer patients treated with radium-223 alone or combined with pembrolizumab.

Reducing Metal Artifacts in Clinical Photon Counting Detector Computed Tomography-A Phantom Study of an Exemplary Total Hip Arthroplasty.

To examine how different photon-counting detector (PCD) CT scanning and reconstruction methods affect the volume of metal artifacts and image quality for a hip prosthesis phantom. A titanium and cobalt-chromium-molybdenum-alloy total hip prosthesis phantom was scanned using a clinical PCD-CT with a constant tube potential (140kV) and Computed-Tomography-Dose- Index (7mGy). Different scan settings were used: with/without tin-filter (Sn), with/without ultra-high resolution (UHR), both individually and combined, resulting in four modes: Quantumplus (Standard), UHR Quantumplus (HighRes), QuantumSn (Standard-Tin) and UHR QuantumSn (HighRes-Tin). Reconstructions included virtual monoenergetic images (VMI) spanning 40-190keV and polychromatic images, with/without iterative metal artifact reduction (MAR). Artifact volumes rendered in a 3D-printing software were quantified in milliliters (ml), and image quality was evaluated using a Likert score. Polychromatic reconstruction: Tin-filter reduced artifact volumes (298 (Standard-Tin) vs. 347ml (Standard) and 310 (HighRes-Tin) vs. 360ml (HighRes)). The smallest artifact volume was measured in HighRes MAR (150ml). VMI reconstruction: The smallest artifact volume was measured in Standard 130keV (150ml) and HighRes 130keV (164ml) and in Standard-Tin 120keV (169ml) and HighRes-Tin 120keV (172ml). MAR further reduced the artifact volumes to 130ml (Standard 150keV MAR) and 140ml (HighRes 160keV MAR). Image quality was rated best for Standard 65keV MAR, polychromatic HighRes MAR, Standard 100keV MAR, polychromatic Standard-tin MAR, HighRes-tin 100keV and polychromatic HighRes-tin. Combining tin-filter, UHR and MAR in VMI or polychromatic images achieve the strongest artifact reduction.

Integrating 3D-printed customized guides with hemicorticotomy bone transport: clinical outcomes in chronic osteomyelitis management

PurposeChronic osteomyelitis with large bone defects remains a challenging condition in orthopedic surgery. This study aimed to evaluate the clinical efficacy of hemicorticotomy bone transport assisted by 3D-printed customized guides in the treatment of chronic osteomyelitis with bone defects.MethodsA retrospective analysis was conducted on 21 patients with chronic osteomyelitis treated with hemicorticotomy bone transport with 3D-printed guide assistance between January 2015 and January 2021. Preoperative planning utilized 3D CT reconstruction for precise guide design and osteotomy placement. The surgical technique involved hemicorticotomy, placement of the 3D-printed guide, and gradual bone transport using a monolateral external fixator. Clinical outcomes were assessed using the Association for the Study and Application of Methods of Ilizarov (ASAMI) scoring system. The external fixation time (EFT), external fixation index (EFI), and complications evaluated using the Paley classification were recorded.ResultsThe mean bone defect size was 7.9 ± 2.0 cm. The average follow-up period was 28.6 ± 6.4 months. The mean external fixation time was 289.4 ± 52.7 days, with an external fixation index of 36.8 ± 4.2 days/cm. According to the ASAMI criteria, the bone results were excellent in 17 patients (81%), good in 3 (14%), and fair in 1 (5%). The functional results were excellent in 15 patients (71%), good in 4 (19%), and fair in 2 (10%). Complete eradication of the infection and union at the docking site were achieved in all patients. The complication rate was 38% and all of which were successfully managed.ConclusionHemicorticotomy bone transport assisted by 3D-printed customized guides is an effective treatment for chronic osteomyelitis with bone defects. This technique offers excellent bone healing, favorable functional results, and a relatively low complication rate. Further studies with larger sample sizes are warranted to validate these findings.

Photon-Counting CT Iodine Maps for Diagnosing Chronic Pulmonary Thromboembolism: A Pilot Study.

The aim of this study was to evaluate the feasibility and efficacy of chronic pulmonary thromboembolism assessment using photon-counting detector computed tomography (PCD-CT) iodine maps of the lung parenchyma. This institutional review board-approved retrospective study included 83 subjects (49.4% male, aged 62.4 ± 13.4 years; 50.6% female, aged 59.9 ± 17.1 years) who underwent clinically indicated PCD-CT scan to rule out chronic thromboembolic pulmonary hypertension (CTEPH). Two blinded readers used iodine maps and corresponding sharp-kernel CT reconstructions in the lung window to evaluate perfusion defects and identify patients with chronic pulmonary thromboembolism (CTEPH, CTEPH overlap with other causes of pulmonary hypertension [PH], chronic thromboembolic disease [CTED]). No other clinical or imaging information was given. Discordance was resolved in a subsequent consensus read. The clinical diagnosis was reviewed in an interdisciplinary clinical setting. The accuracy, sensitivity, and specificity of radiologic evaluation and clinical diagnosis were calculated. Of the 83 subjects included, 32 were diagnosed with CTEPH, CTEPH overlap, or CTED, 35 were diagnosed with PH caused by other pathologic mechanisms, 10 had no PH, and 6 had suffered previous acute pulmonary embolism, which resolved. The interreader agreement was good (Cohen κ = 0.74). The consensus reached high accuracy (0.88), sensitivity (0.94), and specificity (0.84), as well as good agreement with interdisciplinary clinical diagnosis (Cohen κ = 0.75). No cases with confirmed CTEPH as the primary cause of PH or CTED were missed. Patients with pulmonary arterial hypertension were most frequently rated false-positive. The mean effective dose (±standard deviation) was 1.3 (±0.76) mSv. Accurate, sensitive, and specific diagnosis of pulmonary chronic thromboembolism at low radiation dose is possible using iodine maps reconstructed from PCD-CT scans.

Editorial Comment: Standardization of Cardiac CT Acquisition, Contrast Media, and Reconstruction Protocols Will Help Improve the Accuracy of PCAT Measurement.

Editorial Comment: Standardization of Cardiac CT Acquisition, Contrast Media, and Reconstruction Protocols Will Help Improve the Accuracy of PCAT Measurement.

Effect of Vessel Attenuation, VMI Level, and Reconstruction Kernel on Pericoronary Adipose Tissue Attenuation for EID CT and PCD CT: An Ex Vivo Porcine Heart Study.

BACKGROUND. Pericoronary adipose tissue (PCAT) attenuation and the fat attenuation index (FAI) may serve as markers of inflammation and the risk of adverse cardiac events. However, standardization of relevant CT acquisition and reconstruction parameters is lacking. OBJECTIVE. The purpose of this study was to investigate the influence of vessel attenuation, the virtual monoenergetic image (VMI) level, and the reconstruction kernel on PCAT attenuation and FAI by use of energy-integrating detector (EID) and photon-counting detector (PCD) CT systems in an ex vivo porcine heart model. METHODS. The right coronary artery (RCA) of a porcine heart was injected with saline or varying contrast medium dilutions to achieve vessel attenuation ranging from 0 to 1000 HU. After each injection, the heart was sequentially scanned by EID CT at 120 kVp and PCD CT at 140 kVp at a constant CTDIvol of 10 mGy. For EID CT, polychromatic images were reconstructed with a Qr40 kernel. For PCD CT, VMIs (obtained at 40-80 keV in 10-keV increments) were reconstructed with Qr40, Bv40, and Bv56 kernels. ROIs were placed to measure RCA and PCAT attenuation. FAI was determined using software; histogram analysis was performed to assess voxel attenuation in the volumes of interest for FAI calculation. RESULTS. Correlations were observed between attenuation in the RCA and attenuation in the adjacent PCAT (r = 0.3-1.0) and between vessel attenuation and the FAI (r = -0.9 to 1.0). For PCAT attenuation and the FAI, these associations became progressively weaker when progressively sharper kernels were used. For increasing vessel attenuation on EID CT and for increasing VMI level on PCD CT, FAI histograms showed right shifts in peak attenuation values; the percentage of histogram voxels that met the threshold range for inclusion in FAI calculation was 9-38% for EID CT and 6-39% for PCD CT at VMI levels of 70-80 keV. For PCD CT, use of sharper kernels was associated with left shifts in peak attenuation values and greater percentages of voxels within the threshold range for inclusion in FAI calculation. CONCLUSION. PCAT attenuation and the FAI are influenced by vessel lumen attenuation, the VMI level, and the reconstruction kernel. A minority of pericoronary voxels contribute to FAI measurements for polychromatic EID CT and for PCD CT at high VMI levels. CLINICAL IMPACT. These findings may help standardize acquisition and reconstruction parameters for PCAT attenuation and FAI measurements.

A dual-domain deep network for high pitch CT reconstruction

In this paper, we propose an end-to-end dual-domain deep network for the high pitch helical CT reconstruction. Our network is composed of a sinogram domain subnetwork and a CT image domain subnetwork, and the two subnetworks are linked by the Katsevich algorithm layer. The sinogram subnetwork is a discrete cosine transform (DCT) 3D U-net and used to complete the missing sinograms due to high pitch while the CT image subnetwork is a standard 3D U-net used to denoise the CT images. By utilizing the periodic properties of parameters of Katsevich algorithm, our network reconstructs CT images rotation turn by rotation turn and so can reduce the GPU memory consumption. To further reduce the GPU memory consumption, we train the sinogram domain subnetwork and the CT image domain subnetwork separately, where the gradient flow through the Katsevich layer is obtained manually. Experiments show that the proposed network trained by our method outperforms the related methods both in subjective and objective evaluations.

Quantitative analysis of genicular nerve block spread and variability: Anatomical correlations and clinical implications

BackgroundDespite advancements in anaesthesiology and regional anaesthesia, the three-dimensional spread of local anaesthetics is still poorly understood. This study investigates the pericapsular spread of radiopaque contrast after ultrasound-guided genicular nerve block in cadaveric knees using CT reconstruction. The aim is to assess the reproducibility, accuracy, and variability of this technique in order to improve safety and effectiveness. MethodsWe used a four-stage methodology that involved performing genicular nerve block on cadaveric knees, acquiring data using high-resolution ultrasound and helical CT imaging, performing segmentation and surface reconstruction of the CT images, and analysing the data quantitively to determine the spread of the infiltrate in each genicular nerve region. Coefficients of variation were calculated to estimate the spread and reproducibility of each genicular nerve block. ResultsOur results revealed significant variability in spread in all genicular nerve blocks, particularly around the recurrent genicular nerve. Despite this variability, the spread—particularly along the medial-lateral and anterior-posterior axes—showed moderate consistency. Furthermore, the regions covered by 80% of the spread closely matched the pericapsular distribution of the genicular nerves of the knee. ConclusionsNotable variability in spread was observed in genicular nerve blocks, particularly around the recurrent genicular nerve. However, the spread showed moderate consistency and aligns closely with the pericapsular nerve distribution of the knee. Future studies should combine quantitative analysis with anatomical dissection to further investigate the involvement of the deep peroneal motor branch.

Technical Note: An optimized protocol for standard unenhanced whole-body post-mortem Photon Counting CT imaging

In recent years, Photon-counting detector CT (PCD CT) has emerged as a new and groundbreaking technology in clinical radiology. While clinical research and practical applications of PCD-CT are constantly evolving, it has not yet been integrated into post-mortem CT (PMCT) imaging. Documented research into the potential applications of PCD CT in the field of post-mortem human forensic pathology and anatomical pathology is scarce in literature. This is despite the fact that PCD CT shows promise in expanding PMCT imaging diagnostic due to features such as ultra-high resolution and intrinsic spectral multienergy imaging. The authors have recently started scanning forensic corpses with a PCD CT to explore its possibilities and limits in the post-mortem field. In this technical note, the PCD CT acquisition and reconstruction parameters that resulted in excellent image quality in terms of noise and contrast for a slice thickness up to 0.2mm in unenhanced whole-body examinations are presented.

Learned Tensor Neural Network Texture Prior for Photon-Counting CT Reconstruction.

Photon-counting computed tomography (PCCT) reconstructs multiple energy-channel images to describe the same object, where there exists a strong correlation among different channel images. In addition, reconstruction of each channel image suffers photon count starving problem. To make full use of the correlation among different channel images to suppress the data noise and enhance the texture details in reconstructing each channel image, this paper proposes a tensor neural network (TNN) architecture to learn a multi-channel texture prior for PCCT reconstruction. Specifically, we first learn a spatial texture prior in each individual channel image by modeling the relationship between the center pixels and its corresponding neighbor pixels using a neural network. Then, we merge the single channel spatial texture prior into multi-channel neural network to learn the spectral local correlation information among different channel images. Since our proposed TNN is trained on a series of unpaired small spatial-spectral cubes which are extracted from one single reference multi-channel image, the local correlation in the spatial-spectral cubes is considered by TNN. To boost the TNN performance, a low-rank representation is also employed to consider the global correlation among different channel images. Finally, we integrate the learned TNN and the low-rank representation as priors into Bayesian reconstruction framework. To evaluate the performance of the proposed method, four references are considered. One is simulated images from ultra-high-resolution CT. One is spectral images from dual-energy CT. The other two are animal tissue and preclinical mouse images from a custom-made PCCT systems. Our TNN prior Bayesian reconstruction demonstrated better performance than other state-of-the-art competing algorithms, in terms of not only preserving texture feature but also suppressing image noise in each channel image.

A simulation study towards sparse-view spectral X-ray CT segmentation using spectrum-aware deep neural networks

ABSTRACT Spectral X-ray computed tomography (SXCT) is able to provide richer spectral information than conventional X-ray CT, thereby improving the capability of discriminating materials in segmentation tasks. However, segmentation algorithms for conventional CT images often lose their performance in material segmentation with SXCT images due to the spectrum-dependent attenuation properties of X-rays. Recently, deep neural networks (DNNs) have demonstrated their effectiveness in segmentation tasks. However, processing SXCT with DNNs requires significant memory resources and may take significant computation time. To address these issues, this paper introduces a simulation study towards realising deep-learning-based SXCT image segmentation. The proposed method employs a convolutional neural network (CNN) that selectively utilises the significant spectrum channels for accurate material segmentation. The proposed method builds upon a prior study on adaptive CT reconstruction, but to further enhance the performance, we introduce a self-attention-based spectrum-selection module. During training, the selection module assigns importance weights to spectrum channels based on the attention scores. Once the network has been trained, the proposed method can obtain satisfactory segmentation results from projection images at one or a combination of several spectrum channels recommended by the selection module. The experimental results demonstrate that the proposed method achieves a significant reduction in computation time while maintaining the segmentation quality.

Significance of Image Reconstruction Parameters for Future Lung Cancer Risk Prediction Using Low-Dose Chest Computed Tomography and the Open-Access Sybil Algorithm.

Sybil is a validated publicly available deep learning-based algorithm that can accurately predict lung cancer risk from a single low-dose computed tomography (LDCT) scan. We aimed to study the effect of image reconstruction parameters and CT scanner manufacturer on Sybil's performance. Using LDCTs of a subset of the National Lung Screening Trial participants, which we previously used for internal validation of the Sybil algorithm (test set), we ran the Sybil algorithm on LDCT series pairs matched on kilovoltage peak, milliampere-seconds, reconstruction interval, reconstruction diameter, and either reconstruction filter or axial slice thickness. We also evaluated the cumulative effect of these parameters by combining the best- and the worst-performing parameters. A subanalysis compared Sybil's performance by CT manufacturer. We considered any LDCT positive if future lung cancer was subsequently confirmed by biopsy or surgical resection. The areas under the curve (AUCs) for each series pair were compared using DeLong's test. There was no difference in Sybil's performance between 1049 pairs of standard versus bone reconstruction filter (AUC at 1 year 0.84 [95% confidence interval (CI): 0.70-0.99] vs 0.86 [95% CI: 0.75-0.98], P = 0.87) and 1961 pairs of standard versus lung reconstruction filter (AUC at 1 year 0.98 [95% CI: 0.97-0.99] vs 0.98 [95% CI: 0.96-0.99], P = 0.81). Similarly, there was no difference in 1288 pairs comparing 2-mm versus 5-mm axial slice thickness (AUC at 1 year 0.98 [95% CI: 0.94-1.00] vs 0.99 [95% CI: 0.97-0.99], P = 0.68). The best-case scenario combining a lung reconstruction filter with 2-mm slice thickness compared with the worst-case scenario combining a bone reconstruction filter with 2.5-mm slice thickness uncovered a significantly different performance at years 2-4 (P = 0.03). Subanalysis showed no significant difference in performance between Siemens and Toshiba scanners. Sybil's predictive performance for future lung cancer risk is robust across different reconstruction filters and axial slice thicknesses, demonstrating its versatility in various imaging settings. Combining favorable reconstruction parameters can significantly enhance predictive ability at years 2-4. The absence of significant differences between Siemens and Toshiba scanners further supports Sybil's versatility.

Time varying flat field correction of X-ray microtomography with an improved deep-learning method.

In X-ray microtomography, the flat field image is usually needed to normalize the collected sample projections. Owing to the high brightness of the synchrotron radiation facility, dynamic CT imaging of in-situ or in-operando processes is broadly employed for the investigation of three-dimensional microstructure evolution. However, the fast, continuous data acquisition and the heavy, bulky in-situ devices usually prevent the easy collection of accurate flat field images, which means that conventional flat field correction is hard to efficiently correct the artefacts of X-ray microtomography. We report a deep-learning-based artefact correction method for X-ray microtomography, which uses flat field generated from each CT projection by an improved pix2pixHD model. Experimental results demonstrated that the proposed method has a significant advantage over the conventional method and available deep-learning-based flat field correction method for the flat field correction of projection images. The CT imaging results show that the proposed method efficiently reduces the systematic error during the intensity normalization process, and the CT reconstruction is improved significantly. Therefore, the method developed in this paper is applicable for the flat field correction of dynamic CT. Furthermore, experiments with a set of low Z material samples verified the generalization of the deep-learning-based method for a variety of samples never used for network training. In conclusion, the method developed in this paper is practicable for the flat field correction of in-situ CT imaging of dynamic processes and is also applicable to new samples as long as the neural network model is effectively trained.

The pericoronary adipose tissue attenuation in CT strongly depends on kernels and iterative reconstructions.

To investigate the influence of kernels and iterative reconstructions on pericoronary adipose tissue (PCAT) attenuation in coronary CT angiography (CCTA). Twenty otherwise healthy subjects (16 females; median age 52 years) with atypical chest pain, low risk of coronary artery disease (CAD), and without CAD in photon-counting detector CCTA were included. Images were reconstructed with a quantitative smooth (Qr36) and three vascular kernels of increasing sharpness levels (Bv36, Bv44, Bv56). Quantum iterative reconstruction (QIR) was either switched-off (QIRoff) or was used with strength levels 2 and 4. The fat-attenuation-index (FAI) of the PCAT surrounding the right coronary artery was calculated in each dataset. Histograms of FAI measurements were created. Intra- and inter-reader agreements were determined. A CT edge phantom was used to determine the edge spread function (ESF) for the same datasets. Intra- and inter-reader agreement of FAI was excellent (intra-class correlation coefficient = 0.99 and 0.98, respectively). Significant differences in FAI were observed depending on the kernel and iterative reconstruction strength level (each, p < 0.001), with considerable inter-individual variation up to 34 HU and intra-individual variation up to 33 HU, depending on kernels and iterative reconstruction levels. The ESFs showed a reduced range of edge-smoothing with increasing kernel sharpness, causing an FAI decrease. Histogram analyses revealed a narrower peak of PCAT values with increasing iterative reconstruction levels, causing aFAI increase. PCAT attenuation determined with CCTA heavily depends on kernels and iterative reconstruction levels both within and across subjects. Standardization of CT reconstruction parameters is mandatory for FAI studies to enable meaningful interpretations. Question Do kernels and iterative reconstructions influence pericoronary adipose tissue (PCAT) attenuation in coronary CT angiography (CCTA)? Findings Significant differences in fat-attenuation-index (FAI) were observed depending on the kernel and iterative reconstruction strength level with considerable inter- and intra-individual variation. Clinical relevance PCAT attenuation heavily depends on kernels and iterative reconstructions requiring CT reconstruction parameter standardization to enable meaningful interpretations of fat-attenuation differences across subjects.