Coronary CT Angiography with Photon-counting CT: First-In-Human Results.

Background Spatial resolution, soft-tissue contrast, and dose-efficient capabilities of photon-counting CT (PCCT) potentially allow a better quality and diagnostic confidence of coronary CT angiography (CCTA) in comparison to conventional CT. Purpose To compare the quality of CCTA scans obtained with a clinical prototype PCCT system and an energy-integrating detector (EID) dual-layer CT (DLCT) system. Materials and Methods In this prospective board-approved study with informed consent, participants with coronary artery disease underwent retrospective electrocardiographically gated CCTA with both systems after injection of 65-75 mL of 400 mg/mL iodinated contrast agent at 5 mL/sec. A prior phantom task-based quality assessment of the detectability index of coronary lesions was performed. Ultra-high-resolution parameters were used for PCCT (1024 matrix, 0.25-mm section thickness) and EID DLCT (512 matrix, 0.67-mm section thickness). Three cardiac radiologists independently performed a blinded analysis using a five-point quality score (1 = insufficient, 5 = excellent) for overall image quality, diagnostic confidence, and diagnostic quality of calcifications, stents, and noncalcified plaques. A logistic regression model, adjusted for radiologists, was used to evaluate the proportion of improvement in scores with the best method. Results Fourteen consecutive participants (12 men; mean age, 61 years ± 17) were enrolled. Scores of overall quality and diagnostic confidence were higher with PCCT images with a median of 5 (interquartile range [IQR], 2) and 5 (IQR, 1) versus 4 (IQR, 1) and 4 (IQR, 3) with EID DLCT images, using a mean tube current of 255 mAs ± 0 versus 349 mAs ± 111 for EID DLCT images (P < .01). Proportions of improvement with PCCT images for quality of calcification, stent, and noncalcified plaque were 100%, 92% (95% CI: 71, 98), and 45% (95% CI: 28, 63), respectively. In the phantom study, detectability indexes were 2.3-fold higher for lumen and 2.9-fold higher for noncalcified plaques with PCCT images. Conclusion Coronary CT angiography with a photon-counting CT system demonstrated in humans an improved image quality and diagnostic confidence compared with an energy-integrating dual-layer CT. © RSNA, 2022 Online supplemental material is available for this article. See also the editorial by Sandfort and Bluemke in this issue.

In Vivo Molecular K-Edge Imaging of Atherosclerotic Plaque Using Photon-counting CT.

Background Macrophage burden is a major factor in the risk of atherosclerotic plaque rupture, and its evaluation remains challenging with molecular noninvasive imaging approaches. Photon-counting CT (PCCT) with k-edge imaging aims to allow for the specific detection of macrophages using gold nanoparticles. Purpose To perform k-edge imaging in combination with gold nanoparticles to detect and quantify the macrophage burden within the atherosclerotic aortas of rabbits. Materials and Methods Atherosclerotic and control New Zealand white rabbits were imaged before and at several time points up to 2 days after intravenous injection of gold nanoparticles (3.5 mL/kg, 65 mg gold per milliliter). Aortic CT angiography was performed at the end of the follow-up using an intravenous injection of an iodinated contrast material. Gold k-edge and conventional CT images were reconstructed for qualitative and quantitative assessment of the macrophage burden. PCCT imaging results were compared with findings at histologic examination, quantitative histomorphometry, transmission electron microscopy, and quantitative inductively coupled plasma optical emission spectrometry. Pearson correlations between the macrophage area measured in immunostained sections and the concentration of gold and attenuation measured in the corresponding PCCT sections were calculated. Results Seven rabbits with atherosclerosis and four control rabbits without atherosclerosis were analyzed. In atherosclerotic rabbits, calcifications were observed along the aortic wall before injection. At 2 days after injection of gold nanoparticles, only gold k-edge images allowed for the distinction of plaque enhancement within calcifications and for lumen enhancement during angiography. A good correlation was observed between the gold concentration measured within the wall and the macrophage area in 35 plaques (five per rabbit) (r = 0.82; 95% CI: 0.67, 0.91; P < .001), which was higher than that observed on conventional CT images (r = 0.41; 95% CI: 0.09, 0.65; P = .01). Transmission electron microscopy and inductively coupled plasma optical emission spectrometry analyses confirmed the gold k-edge imaging findings. Conclusion Photon-counting CT with gold nanoparticles allowed for the noninvasive evaluation of both molecular and anatomic information in vivo in rabbits with atherosclerotic plaques. Published under a CC BY 4.0 license. Online supplemental material is available for this article. See also the editorial by Leiner in this issue.

Regional muscle changes in adult dysfunctional hip conditions of femoroacetabular impingement and hip dysplasia.

OBJECTIVE: To analyze regional muscle CT density and bulk in femoroacetabular impingement (FAI) and hip dysplasia (HD) versus controls. MATERIALS AND METHODS: Patients who obtained perioperative CT imaging for FAI and HD before surgery were retrospectively studied. Asymptomatic controls included for comparison. Two readers independently evaluated regional hip muscle [iliopsoas (IP), rectus femoris (RF), gluteus minimus (Gm), and medius (GM)] density, muscle area, and muscle circumference. Inter-observer reliability calculated using intra-class correlation coefficient (ICC). RESULTS: A consecutive series of 25 FAI patients, 16 HD patients, and 38 controls were recruited in the study. FAI patients had significantly greater Gm and GM circumferences as well as greater RF and IP areas on the normal side compared to the asymptomatic control group (p values 0.004, 0.032, 0.033, and 0.028, respectively). In addition, Gm and RF circumferences and RF area were significantly larger (p values 0.029, 0.036, and 0.014, respectively) in FAI patients on the affected side compared to the control group. HD patients had significantly smaller Gm and GM circumferences on the affected side than normal side measurements in FAI group (p values 0.043 and 0.003, respectively). Normal side GM circumference was also smaller in HD patients than normal side FAI hips (p value 0.02). There was no significant difference between the measurements on normal and abnormal sides in each disease group. No significant difference was found between measurements of HD compared to controls (p > 0.05). No muscle density differences were seen among different groups. There was moderate to excellent inter-reader reliability for all measurements except Gm muscle density. CONCLUSIONS: Muscle analysis was able to quantify differences among patients with FAI, HD, and asymptomatic controls. These changes could indicate either a muscle imbalance contributing to the pathology or disuse atrophy, which may have implications for specific muscle-strengthening therapies and rehabilitation procedures in such patients.

Cam-type femoroacetabular impingement-correlations between alpha angle versus volumetric measurements and surgical findings.

AIM: Determine correlations of 3DCT cam-type femoroacetabular impingement (FAI) measurements with surgical findings of labral tear and cartilage loss. METHODS: Digital search of symptomatic cam-type FAI from July 2013 to August 2016 yielded 43 patients. Two readers calculated volumes of femoral head, bump, and alpha angles on 3DCT images. Correlations between CT and surgical findings, inter-, and intra-reader reliabilities were assessed using Spearman rank correlation and intraclass correlation coefficients (ICC). RESULTS: Thirteen men and 14 women aged 37 ± 10 (mean ± SD) years were included. Most common clinical finding was positive flexion-adduction-internal rotation (70.4%). Twenty-seven labral tears and 20 cartilage defects were surgically detected. Significant correlations existed between femoral bump, head volumes, and extent of the labral tear (p = 0.008 and 0.003). No significant correlations were found between the alpha angles at 12 to 3 o'clock and the extent of labral tear (p = 0.2, 0.8, 0.9, and 0.09) or any measurement with the cartilage loss (p values for alpha 12 to 3, bump, and head volumes = 0.7, 0.3, 0.9, 0.9, 0.07, and 0.2). Inter- and intra-reader reliabilities were excellent to moderate for femoral head and bump volumes (ICC = 0.85, 0.52, and 0.8, 0.5) and moderate to poor for alpha angles (ICC = 0.48, 0.40, 0.05, 0.25 and 0.3, 0.24, 0.29, 0.49). CONCLUSION: Three dimensional volumetric measurements of cam-type FAI significantly correlate with the extent of intraoperative labral tears. Superior inter- and intra-reader reliability to that of alpha angles renders it a more clinically relevant measurement for quantifying cam morphology. KEY POINTS: • The 3DCT bump volume and femoral head volume showed significant correlations with the extent of labral tear (p values = 0.008 and 0.003). • No significant correlations were seen between alpha angles and the extent of labral tear (p values > 0.05). • Inter- and intra-reader reliability was excellent to moderate (ICC = 0.85 and 0.52, 0.8, and 0.5) for femoral head and bump volumes while inter- and intra-reader reliability was fair to poor (ICC = 0.48, 0.40, 0.05, 0.25 and 0.3, 0.24, 0.29, 0.49) for alpha angles.

Point vs. traditional method evaluation of hallux valgus: interreader reliability and intermethod performance using X-ray and MRI.

BACKGROUND: The two most widely used measurements for diagnosing and assessing the severity of hallux valgus are the hallux valgus angle (HVA) and the intermetatarsal angle (IMA). Traditionally, these have been measured by using the midaxial lines approximating the axis of each bone. A new simpler point method has been recently suggested for measuring these angles by connecting points along the medial corners of each bone. Interreader reliability of these measurements on X-ray and MRI as well as intermethod and intermodality differences have not been assessed. METHODS: A series of 56 consecutive patients between 18 and 100 years old with no history of foot trauma or orthopedic hardware in their feet were included. All had AP and lateral X-rays and MRI performed on the same foot between April 27, 2015 and March 9, 2016. Two readers measured HVA and IMA using both the traditional midaxial and new point methods. ICC correlations were obtained. RESULTS: The interreader reliability for HVA was similar on point method (0.92) and traditional method (0.94). For the IMA, the ICC was 0.77 on point method versus 0.76 on traditional method. The intermodality agreement (between X-ray and MRI) was higher for HVA (ICC = 0.85, 0.88) as compared to IMA (0.58, 0.74), respectively on both methods. The mean difference between the methods was larger on traditional method = 5.5 for HVA and 2.5° for IMA. CONCLUSIONS: HVA is more reliable than IMA on both methods and modalities and a significant difference exists between the magnitudes of values obtained using the two methods. LEVEL OF CLINICAL EVIDENCE: 3.

Subacromial impingement anatomy and its association with rotator cuff pathology in women: radiograph and MRI correlation, a retrospective evaluation.

OBJECTIVE: To evaluate the relationships between acromial anatomy and developmental alterations with rotator cuff tears in female patients and compare these parameters on radiographs and corresponding MRIs along with inter-reader performance. MATERIALS AND METHODS: Patient demographics, symptoms, and acromial characteristics on radiograph (acromial index, lateral acromion angle, subacromial space on AP and Y- views, acromial anterior and lateral downsloping) and MRI (shape, slope, spur, osteoarthrosis, os acromiale) were recorded. Radiographic and MRI findings were compared and correlated with rotator cuff pathology on MRI. Inter-reader analysis was performed. RESULTS: A total of 140 MRIs from 137 female patients were included. No significant correlation (p > 0.05) existed between acromial parameters and rotator cuff tears, except for a smaller subacromial space on the Y view and spurs correlated with subscapularis tendon tear (p = 0.02, p = 0.04). The presence of lateral downsloping on MRI correlated with a smaller lateral acromion angle (p = 0.0002) and the presence of lateral downsloping on radiography (p = 0.0015). Inter-reader agreements were good to excellent (ICC: 0.65-0.89). CONCLUSION: Subacromial impingement anatomy characteristics have no significant associations with supraspinatus or infraspinatus tears in symptomatic women. Among different measures, supine MRI can be reliably used to identify lateral downsloping of the acromion.

Sciatic neurosteatosis: Relationship with age, gender, obesity and height.

AIM: To evaluate inter-reader performance for cross-sectional area and fat quantification of bilateral sciatic nerves on MRI and assess correlations with anthropometrics. METHODS: In this IRB-approved, HIPPA-compliant study, three readers performed a cross-sectional analysis of 3T lumbosacral plexus MRIs over an 18-month period. Image slices were evaluated at two levels (A and B). The sciatic nerve was outlined using a free hand region of interest tool on PACS. Proton-density fat fraction (FF) and cross-sectional areas were recorded. Inter-reader agreement was assessed using intra-class correlation coefficient (ICC). Spearman correlation coefficients were used for correlations with age, BMI and height and Wilcoxon rank sum test was used to assess gender differences. RESULTS: A total of 67 patients were included in this study with male to female ratio of 1:1. Inter-reader agreement was good to excellent for FF measurements at both levels (ICC=0.71-0.90) and poor for sciatic nerve areas (ICC=0.08-0.27). Positive correlations of sciatic FF and area were seen with age (p value<0.05). Males had significantly higher sciatic intraneural fat than females (p<0.05). CONCLUSION: Fat quantification MRI is highly reproducible with significant positive correlations of sciatic FF and area with age, which may have implications for MRI diagnosis of sciatic neuropathy. KEY POINTS: • MR proton density fat fraction is highly reproducible at multiple levels. • Sciatic intraneural fat is positively correlated with increasing age (p < 0.05). • Positive correlations exist between bilateral sciatic nerve areas and age (p < 0.05). • Males had significantly higher sciatic intraneural fat than females (p < 0.05).

New concepts of radiologic preoperative evaluation of anterior shoulder instability: on-track and off-track lesions.

The shoulder is the most frequently dislocated joint in the body due to a larger range of motion and a small area of articulation between the humeral and glenoid surfaces. Traumatic shoulder dislocations, especially those associated with injury to the labroligamentous or bony stabilizers of the joint, lead to further reduction of articular surface contact with resultant glenohumeral instability and recurrent shoulder dislocations. Imaging plays an increasingly important role in the preoperative evaluation of patients with traumatic shoulder instability by evaluating glenohumeral bone loss (uni- or bipolar), assessing soft tissue injuries and identifying patients at risk of postoperative recurrence. Quantification of bone loss is key to differentiate engaging vs. non-engaging Hill-Sachs lesions, while newer concepts of "on-track" vs. "off-track" lesions are being discussed that can determine the required surgical approaches. In this article, we review the preoperative imaging approaches, traditional treatments, outline the bone loss measurement strategies and review these new tracking concepts with relevant case examples.

Magnetic Resonance Neurography of Traumatic and Nontraumatic Peripheral Trigeminal Neuropathies.

PURPOSE: The clinical neurosensory testing (NST) is currently the reference standard for the diagnosis of traumatic and nontraumatic peripheral trigeminal neuropathies (PTNs), but exhibits both false-positive and false-negative results compared with surgical findings and frequently results in treatment decision delays. We tested the hypothesis that magnetic resonance neurography (MRN) of PTNs can serve as a diagnostic modality by correlating the NST, MRN, and surgical findings. MATERIALS AND METHODS: Sixty patients with traumatic and nontraumatic PTN of varying etiologies and Sunderland classifications underwent NST, followed by MRN using 1.5T and 3.0T scanners. The protocol included 2-dimensional and 3-dimensional (3D) imaging, including diffusion imaging and isotropic 3D PSIF. The MRN findings were read by 2 readers in consensus with the clinical findings but without knowing the side of abnormality. The MRN results were summarized using the Sunderland classification. In 26 patients, surgery was performed, and the Sunderland classification was assigned using the surgical photographs. Agreement between the MRN findings and NST/surgical classification was evaluated using kappa statistics. Pearson's correlation coefficient was used to assess the correlation between continuous measurements of MRN/NST and surgical classification. RESULTS: Of the 60 patients, 19 males and 41 females, mean age 41 years (range 12 to 75), with 54 complaints of altered sensation of the lip, chin, or tongue, including 16 with neuropathic pain and 4 with no neurosensory complaint, were included. Third molar surgery (n = 29) represented the most common cause of traumatic PTN. Assuming 1 nerve abnormality per patient, the lower class was accepted, a kappa of 0.57 was observed between the MRN and NST classification. A kappa of 0.5 was found between MRN and surgical findings with a Pearson correlation coefficient of 0.67. CONCLUSIONS: MRN anatomically maps PTNs and stratifies the nerve injury and neuropathies with moderate to good agreement with NST and surgical findings for clinical use.

Hallux Valgus Evaluation on MRI: Can Measurements Validated on Radiographs Be Used?

Hallux valgus (HV) is a common deformity of the great toe affecting >23% of adults in the United States. The severity of the deformity is traditionally analyzed using radiographs to determine measurements such as the HV and intermetatarsal angles. We sought to determine the relationship between the radiographic and magnetic resonance imaging (MRI) measurements because this is not yet known. Two of us analyzed a series of 56 consecutive patients who had had radiographs and MRI performed on the same foot between April 27, 2015 and March 9, 2016 and who satisfied all other inclusion and exclusion criteria (age 18 to 100 years, no history of recent foot trauma, and no metal hardware in the foot). We found excellent interreader reliability (intraclass correlation 0.89 to 0.96) and intermodality agreement (intraclass correlation 0.83 to 0.91). The HV angle measured 15.0° ± 8.8° on the MRI scans and 13.8° ± 8.7° on the radiographs (mean difference -1.15° ± 3.89°), and the intermetatarsal angle was 9.0° ± 3.1° on the MRI scans and 8.8° ± 2.9° on the radiographs (mean difference -0.22° ± 2.10°). The HV measurements were reliable on both radiographs and MRI for the range of values tested. Small intermodality statistically significant differences in HV angle measurements were found; however, these might not be enough to be clinically significant.

Pediatric posterior reversible encephalopathy syndrome: Can MR imaging features predict outcomes in non-oncologic patients?

PURPOSE: Identify MR features predictive of poor outcomes in non-oncologic pediatric PRES. METHOD: A six-year search of all non-oncologic pediatric patients with clinical and MR features of PRES was performed. Modified Rankin scores were used to classify clinical outcomes into good versus poor, then clinical and MR features were compared among groups. Univariate and multivariate analysis was performed to identify MR predictors of poor outcomes for various imaging features, and p-values < 0.05 were considered statistically significant. RESULTS: One hundred and forty-one patients (mean age 10.1 ± 3.0 years, male to female ratio 1:1.1) were included. Clinically, nephrotic syndrome (p = 0.03), focal deficits (p = 0.04), longer hospitalization (p < 0.001), and mechanical ventilation (p < 0.001) were significantly associated with poor outcomes. Univariate analysis revealed that deep grey matter nuclei (OR = 5.29, 95 % CI: 1.6-18.0) and cerebellar edema patterns (OR = 3.49, 95 % CI: 1.3-9.5), cytotoxic edema (OR = 63.6, 95 % CI:16.5-244.2), hemorrhage (OR = 16.58, 95 % CI: 4.3-64.2), and severe PRES patterns (OR = 11.0, 95 % CI: 3.5-34.7) on MR were all significantly associated with poor outcomes (p-values = 0.008 and 0.014, <0.001, <0.001, and < 0.001, respectively). This remained true for cytotoxic edema (OR = 84.26, 95 % CI: 17.3-410.9, p-value < 0.001) and hemorrhage (OR = 44.56, 95 % CI: 6.9-289.7, p-value < 0.001) on multivariate analysis. CONCLUSION: Diffusion restriction and hemorrhage on initial MR scans were the two independent predictors of poor outcomes in non-oncologic pediatric patients.

Insights Into MRI Neuroimaging Patterns of COVID-19 in Children: A Retrospective Comprehensive Analysis.

RATIONALE AND OBJECTIVES: Neurological complications associated with coronavirus disease (COVID-19) have been reported in children; however, data on neuroimaging findings remain limited. This study aimed to comprehensively examine neuroimaging patterns of COVID-19 in children and their relationship with clinical outcomes. MATERIALS AND METHODS: This retrospective cross-sectional study involved reviewing the medical records and MRI scans of 95 children who developed new neurological symptoms within 2-4 weeks of clinical and laboratory confirmation of COVID-19. Patients were categorized into four groups based on guidelines approved by the Centers for Disease Control and Prevention (CDC). Initial brain/spinal MRI was performed. Images were reviewed by three blinded radiologists, and the findings were analyzed and categorized based on the observed patterns in the brain and spinal cord. Follow-up MRI was performed and analyzed to track lesion progression. RESULTS: Encephalopathy was the most common neurological symptom (50.5%). The most common initial MRI involvement patterns were non-confluent multifocal hyperintense white matter (WM) lesions (36.8%) and ischemia (18.9%). Most patients who underwent follow-up MRI (n = 56) showed complete resolution (69.9%); however, some patients developed encephalomalacia and myelomalacia (23.2% and 7.1%, respectively). Non-confluent hyperintense WM lesions were associated with good outcomes (45.9%, P = 0.014), whereas ischemia and hemorrhage were associated with poor outcomes (44.1%, P < 0.001). CONCLUSION: This study revealed diverse neuroimaging patterns in pediatric COVID-19 patients. Non-confluent WM lesions were associated with good outcomes, whereas ischemia and hemorrhage were associated with poorer prognoses. Understanding these patterns is crucial for their early detection, accurate diagnosis, and appropriate management.

Gadolinium K-edge angiography with a spectral photon counting CT in atherosclerotic rabbits.

PURPOSE: The purpose of this study was to investigate the feasibility of gadolinium-K-edge-angiography (angio-Gd-K-edge) with gadolinium-based contrast agents (GBCAs) as obtained with spectral photon counting CT (SPCCT) in atherosclerotic rabbits. MATERIALS AND METHODS: Seven atherosclerotic rabbits underwent angio-SPCCT acquisitions with two GBCAs, with similar intravenous injection protocol. Conventional and angio-Gd-K-edge images were reconstructed with the same parameters. Regions of interest were traced in different locations of the aorta and its branches. Hounsfield unit values, Gd concentrations, signal-to-noise (SNR) and contrast-to-noise (CNR) were calculated and compared. The maximum diameter and the diameter of the aorta in regard to atherosclerotic plaques were measured by two observers. Images were subjectively evaluated regarding vessels' enhancement, artefacts, border sharpness and overall image quality. RESULTS: In the analyzable six rabbits, Gd-K-edge allowed visualization of target vessels and no other structures. HU values and Gd concentrations were greatest in the largest artery (descending aorta, 5.6 ± 0.8 [SD] mm), and lowest in the smallest (renal arteries, 2.1 ± 0.3 mm). While greater for conventional images, CNR and SNR were satisfactory for both images (all P < 0.001). For one observer there were no statistically significant differences in either maximum or plaque-diameters (P = 0.45 and all P > 0.05 in post-hoc analysis, respectively). For the second observer, there were no significant differences for images reconstructed with the same parameters (all P < 0.05). All subjective criteria scored higher for conventional images compared to K-edge (all P < 0.01), with the highest scores for enhancement (4.3-4.4 vs. 3.1-3.4). CONCLUSION: With SPCCT, angio-Gd-K-edge after injection of GBCAs in atherosclerotic rabbits is feasible and allows for angiography-like visualization of small arteries and for the reliable measurement of their diameters.

A conventional-to-spectral CT image translation augmentation workflow for robust contrast injection-independent organ segmentation.

PURPOSE: In computed tomography (CT) cardiovascular imaging, the numerous contrast injection protocols used to enhance structures make it difficult to gather training datasets for deep learning applications supporting diverse protocols. Moreover, creating annotations on noncontrast scans is extremely tedious. Recently, spectral CT's virtual-noncontrast images (VNC) have been used as data augmentation to train segmentation networks performing on enhanced and true-noncontrast (TNC) scans alike, while improving results on protocols absent of their training dataset. However, spectral data are not widely available, making it difficult to gather specific datasets for each task. As a solution, we present a data augmentation workflow based on a trained image translation network, to bring spectral-like augmentation to any conventional CT dataset. METHOD: The conventional CT-to-spectral image translation network (HUSpectNet) was first trained to generate VNC from conventional housnfied units images (HU), using an unannotated spectral dataset of 1830 patients. It was then tested on a second dataset of 300 spectral CT scans by comparing VNC generated through deep learning (VNCDL ) to their true counterparts. To illustrate and compare our workflow's efficiency with true spectral augmentation, HUSpectNet was applied to a third dataset of 112 spectral scans to generate VNCDL along HU and VNC images. Three different three-dimensional (3D) networks (U-Net, X-Net, and U-Net++) were trained for multilabel heart segmentation, following four augmentation strategies. As baselines, trainings were performed on contrasted images without (HUonly) and with conventional gray-values augmentation (HUaug). Then, the same networks were trained using a proportion of contrasted and VNC/VNCDL images (TrueSpec/GenSpec). Each training strategy applied to each architecture was evaluated using Dice coefficients on a fourth multicentric multivendor single-energy CT dataset of 121 patients, including different contrast injection protocols and unenhanced scans. The U-Net++ results were further explored with distance metrics on every label. RESULTS: Tested on 300 full scans, our HUSpectNet translation network shows a mean absolute error of 6.70 ± 2.83 HU between VNCDL and VNC, while peak signal-to-noise ratio reaches 43.89 dB. GenSpec and TrueSpec show very close results regardless of the protocol and used architecture: mean Dice coefficients (DSCmean ) are equal with a margin of 0.006, ranging from 0.879 to 0.938. Their performances significantly increase on TNC scans (p-values < 0.017 for all architectures) compared to HUonly and HUaug, with DSCmean of 0.448/0.770/0.879/0.885 for HUonly/HUaug/TrueSpec/GenSpec using the U-Net++ architecture. Significant improvements are also noted for all architectures on chest-abdominal-pelvic scans (p-values < 0.007) compared to HUonly and for pulmonary embolism scans (p-values < 0.039) compared to HUaug. Using U-Net++, DSCmean reaches 0.892/0.901/0.903 for HUonly/TrueSpec/GenSpec on pulmonary embolism scans and 0.872/0.896/0.896 for HUonly/TrueSpec/GenSpec on chest-abdominal-pelvic scans. CONCLUSION: Using the proposed workflow, we trained versatile heart segmentation networks on a dataset of conventional enhanced CT scans, providing robust predictions on both enhanced scans with different contrast injection protocols and TNC scans. The performances obtained were not significantly inferior to training the model on a genuine spectral CT dataset, regardless of the architecture implemented. Using a general-purpose conventional-to-spectral CT translation network as data augmentation could therefore contribute to reducing data collection and annotation requirements for machine learning-based CT studies, while extending their range of application.

First In-Human Results of Computed Tomography Angiography for Coronary Stent Assessment With a Spectral Photon Counting Computed Tomography.

OBJECTIVES: The aim of this study is to compare the image quality of in vivo coronary stents between an energy integrating detectors dual-layer computed tomography (EID-DLCT) and a clinical prototype of spectral photon counting computed tomography (SPCCT). MATERIALS AND METHODS: In January to June 2021, consecutive patients with coronary stents were prospectively enrolled to undergo a coronary computed tomography (CT) with an EID-DLCT (IQon, Philips) and an SPCCT (Philips). The study was approved by the local ethical committee and patients signed an informed consent. A retrospectively electrocardiogram-gated acquisition was performed with optimized matching parameters on the 2 scanners (EID-DLCT: collimation, 64 × 0.625 mm; kVp, 120, automatic exposure control with target current at 255 mAs; rotation time, 0.27 seconds; SPCCT: collimation, 64 × 0.275 mm; kVp, 120; mAs, 255; rotation time, 0.33 seconds). The injection protocol was the same on both scanners: 65 to 75 mL of Iomeron (Bracco) at 5 mL/s. Images were reconstructed with slice thickness of 0.67 mm, 512 matrix, XCB (Xres cardiac standard) and XCD (Xres cardiac detailed) kernel, iDose 3 for EID-DLCT and 0.25-mm slice thickness, 1024 matrix, Detailed 2 and Sharp kernel, and iDose 6 for SPCCT. Two experienced observers measured the proximal and distal external and internal diameters of the stents to quantify blooming artifacts. Regions of interest were drawn in the lumen of the stent and of the upstream coronary artery. The difference (Δ S-C) between the respective attenuation values was calculated as a quantification of stent-induced artifacts on intrastent image quality. For subjective image quality, 3 experienced observers graded with a 4-point scale the image quality of different parameters: coronary wall before the stent, stent lumen, stent structure, calcifications surrounding the stent, and beam-hardening artifacts. RESULTS: Eight patients (age, 68 years [interquartile range, 8]; all men; body mass index, 26.2 kg/m2 [interquartile range, 4.2]) with 16 stents were scanned. Five stents were not evaluable owing to motion artifacts on the SPCCT. Of the remaining, all were drug eluting stents, of which 6 were platinum-chromium, 3 were cobalt-platinum-iridium, and 1 was stainless steel. For 1 stent, no information could be retrieved. Radiation dose was lower with the SPCCT (fixed CT dose index of 25.7 mGy for SPCCT vs median CT dose index of 35.7 [IQ = 13.6] mGy; P = 0.02). For 1 stent, the internal diameter was not assessable on EID-DLCT. External diameters were smaller and internal diameters were larger with SPCCT (all P < 0.05). Consequently, blooming artifacts were reduced on SPCCT (P < 0.05). Whereas Hounsfield unit values within the coronary arteries on the 2 scanners were similar, the Δ S-C was lower for SPCCT-Sharp as compared with EID-DLCT-XCD and SPCCT-Detailed 2 (P < 0.05). The SPCCT received higher subjective scores than EID-DLCT for stent lumen, stent structure, surrounding calcifications and beam-hardening for both Detailed 2 and Sharp (all P ≤ 0.05). The SPCCT-Sharp was judged better for stent structure and beam-hardening assessment as compared with SPCCT-Detailed 2. CONCLUSION: Spectral photon counting CT demonstrated improved objective and subjective image quality as compared with EID-DLCT for the evaluation of coronary stents even with a reduced radiation dose.

Spectral augmentation for heart chambers segmentation on conventional contrasted and unenhanced CT scans: an in-depth study.

PURPOSE: Recently, machine learning has outperformed established tools for automated segmentation in medical imaging. However, segmentation of cardiac chambers still proves challenging due to the variety of contrast agent injection protocols used in clinical practice, inducing disparities of contrast between cavities. Hence, training a generalist network requires large training datasets representative of these protocols. Furthermore, segmentation on unenhanced CT scans is further hindered by the challenge of obtaining ground truths from these images. Newly available spectral CT scanners allow innovative image reconstructions such as virtual non-contrast (VNC) imaging, mimicking non-contrasted conventional CT studies from a contrasted scan. Recent publications have demonstrated that networks can be trained using VNC to segment contrasted and unenhanced conventional CT scans to reduce annotated data requirements and the need for annotations on unenhanced scans. We propose an extensive evaluation of this statement. METHOD: We undertake multiple trainings of a 3D multi-label heart segmentation network with (HU-VNC) and without (HUonly) VNC as augmentation, using decreasing training dataset sizes (114, 76, 57, 38, 29, 19 patients). At each step, both networks are tested on a multi-vendor, multi-centric dataset of 122 patients, including different protocols: pulmonary embolism (PE), chest-abdomen-pelvis (CAP), heart CT angiography (CTA) and true non-contrast scans (TNC). An in-depth comparison of resulting Dice coefficients and distance metrics is performed for the networks trained on the largest dataset. RESULTS: HU-VNC-trained on 57 patients significantly outperforms HUonly trained on 114 regarding CAP and TNC scans (mean Dice coefficients of 0.881/0.835 and 0.882/0.416, respectively). When trained on the largest dataset, significant improvements in all labels are noted for TNC and CAP scans (mean Dice coefficient of 0.882/0.416 and 0.891/0.835, respectively). CONCLUSION: Adding VNC images as training augmentation allows the network to perform on unenhanced scans and improves segmentations on other imaging protocols, while using a reduced training dataset.

Comparison of free breathing 3D mDIXON with single breath-hold 3D inversion recovery sequences for the assessment of Late Gadolinium Enhancement.

PURPOSE: To evaluate the technical and diagnostic performance of three dimensional (3D) mDIXON versus 3D inversion recovery (3D VIAB) and 3D spectral presaturation with inversion recovery (3D SPIR) late gadolinium enhancement (LGE) sequences. METHODS: A total of 78 patients (50 males and 28 females, age 49 ± 18 years) with 1.5 T CMR examination including three different 3D LGE sequences (3D mDIXON, 3D VIAB, and 3D SPIR) were evaluated for technical and diagnostic performance by two readers. Qualitative scores and quantitative signal and contrast-to-noise ratios were compared among sequences. Qualitative comparisons were made using Friedman and Wilcoxon signed rank tests. Quantitative comparisons were made using one way ANOVA. Reader agreements were tested using Cohen's Kappa. Any p-value <0.05 was significant. RESULTS: 19 out of 78 patients (24 %) were excluded due to poor (grade 4) image quality and 29 patients were excluded due to absence of LGE. For the remaining 30 patients, free breathing 3D mDIXON showed higher confidence in diagnosis of subepicardial LGE (p-value < 0.05). 3D mDIXON outperformed 3D SPIR in both visualization of LGE (p = 0.02) and quality of fat suppression (p = 0.001). Nevertheless, 3D mDIXON showed lower image quality compared to the other two sequences. CONCLUSION: Free breathing 3D mDIXON is a diagnostic problem-solving tool, especially when making a diagnosis of subepicardial enhancement and/or fat suppression is needed, owing to its high spatial resolution and robust fat suppression. Choice of 3D LGE sequence should be based on patient's breath-hold ability, diagnostic needs, and institutional availability considering the strengths and limitations of each sequence.