Journal of Computer Assisted Tomography最新文献

Objective: To investigate the performance of various convolutional neural networks (CNNs) in identifying clear renal cell carcinoma (ccRCC) on MRI and to compare them with radiologists using the clear cell likelihood score (ccLS) algorithm.

Methods: A total of 480 CNN models were retrospectively trained using 1 or 3 (fusion) different types of MR images obtained from 310 patients with pathologically confirmed renal masses (cT1, ≤7 cm). MR images included T2-weighted (T2w), T1-weighted opposed-phase/in-phase (T1wOPIP), and corticomedullary phase (CMphase) contrast-enhanced MR acquisitions. To increase the robustness of classification, a 5-fold cross-validation was performed, and the averaged area under curve (AUC) values were compared among the CNN models. The best performing CNN models were compared with radiologists' performance using the ccLS algorithm.

Results: The performance of CNN models in ccRCC classification was significantly influenced by the choice of models, the type of input images, and image size. The best CNN models for the diagnosis of ccRCC using T2w and all 3 types of images (fusion CNN models) achieved an AUC of 0.79 and 0.8, respectively. Combining these models using logistic regression produced a slightly higher AUC value of 0.83. Radiologists using the ccLS achieved an AUC value of 0.85, which was not statistically different than the T2w+fusion model (P>0.05). Combining the T2w+fusion model and the ccLS achieved an AUC value of 0.86, which was not different than the AUC value using ccLS alone (P>0.05).

Conclusions: A CNN model integrating three different types of MR images demonstrated performance comparable to that of radiologists in the diagnosis of ccRCC.

Objective: To describe the imaging appearances and treatment response patterns of the pancreatic neuroendocrine tumors (panNETs) following radiofrequency ablation (RFA).

Methods: From an internal database, 17 patients (8 male; mean age: 67±14 y) with 18 pathology-proven, localized, nonfunctioning panNETs <3 cm who underwent EUS-RFA for curative intent were included. A total of 32 preablation and 33 postablation scans were included (CT, MRI, or 68Ga-DOTATATE PET). Lesion size and enhancement on CT/MRI were independently assessed by 2 readers, while SUVmax was extracted from the original PET reports by a separate reviewer. The Wilcoxon signed-rank and McNemar tests were performed. Treatment response is defined as a complete response (loss of enhancement and SUVmax), a partial response (decrease in size, enhancement, or SUVmax), or no response (no change).

Results: Mean lesion size decreased from 1.4​​​​​​±0.5​ cm preablation to 0.3±0.5 cm postablation (P<0.0001). Mean SUVmax declined from 17.3±11.2 to 3.1±6.0 (P<0.001). Hyperenhancement was present in 15/18 (83.3%) lesions preablation versus 5/18 (27.8%) postablation (P<0.01). Of these 15 hyperenhancing lesions, 11 were solid, 3 were cystic, and 1 was mixed cystic and solid. Complete response occurred in 12/18 (66.7%) lesions, with either complete disappearance 5/12 (41.4%) or bland cavity formation 7/12 (58.5%). Partial response occurred in 5/18 (27.8%) lesions; 4/5 decreased in size (mean±SD: 1.4±0.5 cm preablation vs. 0.6±0.7 cm postablation), and 3/5 demonstrated decreased SUVmax. One patient with partial response underwent 2 repeat ablations with an ultimate decrease in SUVmax from 34.1 to 5.9. One solid, hyperenhancing pancreatic body lesion demonstrated no response (1.3 cm); preablation SUVmax was 10.8, but they did not undergo postablation DOTATATE PET. One patient developed postablation pancreatitis. Mean clinical follow-up was 650 days (423).

Conclusion: RFA is an emerging treatment for small, nonfunctioning panNET. Postablation imaging findings most commonly included complete resolution of the tumor, decreased enhancement, decreased SUVmax, and formation of a bland cavity. As interest in this technique continues to grow, radiologists' familiarity with expected post-treatment imaging appearances and their associated response patterns is essential for accurate assessment.

Chest CTs done for unrelated reasons have the potential to reveal unsuspected abnormalities and provide complementary information to help direct appropriate management. By thoroughly understanding the diagnostic role and limitations of CT in assessing ventricular size and morphology as well as myocardial attenuation, enhancement, and thickness, radiologists will be better equipped to differentiate between normal and abnormal findings and provide precise, clinically appropriate recommendations.

Objective: To evaluate the diagnostic value of ultra-high-resolution computed tomography (UHCT) in assessing small airway disease (SAD) in patients with emphysema and analyse its correlation with quantitative expiratory CT parameters, while comparing the performance of different CT systems.

Methods: In this prospective study, 120 emphysema patients underwent CT scanning using 3 different systems: GE Revolution CT (256-slice), Philips IQon Elite Spectral CT, and GE Discovery CT750 HD (64-slice). Small airway measurements were performed from the third to seventh-generation airways, and quantitative CT parameters were analysed.

Results: The GE Revolution CT (256-slice) enabled visualization of airways up to the seventh generation in 82% of cases, compared with 68% using Philips IQon Elite Spectral CT and 53% using GE Discovery CT750 HD (P<0.001). Fourth-generation airway WA% showed strong negative correlation with FEV1% predicted (r=-0.72, P<0.001). The GE Revolution CT (256-slice) demonstrated superior reproducibility for wall-area measurements and lower radiation doses while maintaining superior image quality.

Conclusion: Ultra-high-resolution CT demonstrates superior capabilities in evaluating SAD in patients with emphysema, providing more precise and reproducible measurements compared with conventional CT systems. The strong correlations between structural and functional parameters validate the clinical relevance of these imaging biomarkers. The improved visualisation and quantification of distal airways, achieved with acceptable radiation exposure, represents an important advancement in SAD assessment.

Objectives: Early prediction of remnant liver growth after portal vein embolization (PVE) would enable earlier surgery in patients with resectable liver malignancy, decreasing the risk of tumor progression. This study aims to demonstrate the feasibility of 4D flow MRI to detect changes in portal blood flow before and after PVE and determine whether 4D flow MRI can predict hypertrophy of nonembolized liver lobes (future liver remnant) in a porcine model.

Methods: Thirteen juvenile pigs were included in this prospective study. In 11 pigs, 2 of the 4 portal vein branches were selectively embolized using polyvinyl alcohol particles. Two pigs served as controls. All animals were imaged before, 1 hour, and 1 and 2 weeks after PVE using 4D flow and T1-weighted 3D spoiled gradient-echo MRI. Flow measurements were correlated with liver volumetry measurements. The predictive value of flow changes immediately after PVE for future liver remnant hypertrophy at 2 weeks was estimated using linear mixed-effect model analysis.

Results: Twelve pigs completed the study. One treatment pig died from complications of anesthesia. Immediately after PVE, relative flow in the 2 embolized portal vein branches decreased from 17.0±2.1% to 3.4±2.1% of total portal flow, respectively. Relative flow increased from 16.6±3.8% to 32.6±3.8% in the remaining 2 nonembolized portal vein branches supplying the future liver remnant. Each 1% increase in relative blood flow directly after the procedure predicted a 4.8±0.7 cm 3 (estimate±SE) increase in volume of the future liver remnant at 2 weeks ( P <0.0001). Conversely, in embolized lobes, each 1% flow decrease predicted a volume reduction of 3.4±1.0 cm 3 ( P =0.003).

Conclusion: 4D flow MRI successfully quantified portal blood flow before and after portal vein embolization in a porcine model. Notably, immediate postembolization changes in portal flow were predictive of lobar volume changes observed at 2 weeks. This porcine model thus provides a valuable platform for investigating the relationship between portal hemodynamics and hypertrophy of the future liver remnant. Furthermore, it will enable systematic evaluation of various embolization agents and techniques, supporting the refinement of interventional strategies in hepatic regenerative research.

Objective: This study aimed to validate a 7-day hematocrit (Hct) window for calculating liver extracellular volume (ECV) from dual-energy CT (DECT) iodine maps could potentially replace the strict 24-hour Hct requirement in cirrhotic patients, and to evaluate its diagnostic accuracy for staging cirrhosis severity, thereby reducing the need for stringent 24-hour blood sampling in routine practice.

Materials and methods: This retrospective study enrolled 46 clinically or pathologically confirmed patients with cirrhosis (case group) and 16 participants without hepatic disease (control group). All participants underwent DECT scanning, with Hct measurements obtained within specified time windows relative to DECT examination: Hct0 (<24 h), Hct1 (24 h to 7 d), Hct2 (8 to 30 d), and Hct3 (>30 d). ECV values corresponding to each time window (ECV0, ECV1, ECV2, and ECV3) were calculated from DECT iodine maps using the respective Hct measurements. Intergroup differences in Hct and ECV values across time windows were analyzed against baseline (Hct0/ECV0) using independent sample t tests or Mann-Whitney U tests, as appropriate. Bland-Altman analysis was used to evaluate the agreement between non-24-hour ECV measurements (ECV1, ECV2, and ECV3) and ECV0. Receiver operating characteristic (ROC) curve analysis was used to assess the discriminative capacity of ECV parameters for Child-Pugh classification in cirrhosis using DeLong test to compare areas under the curve (AUC).

Results: The results demonstrated no significant differences in Hct and ECV measurements obtained within 24 hours to 7 days (Hct1/ECV1) compared with 24-hour baseline values (Hct0/ECV0) (all P > 0.05). Bland-Altman analysis revealed the smallest bias (-0.2%) between ECV1 and ECV0, with most data points falling within the limits of agreement. However, Hct and ECV measurements beyond 7 days (Hct2-3/ECV2-3) showed statistically significant deviations from baseline (all P <0.05), exhibiting progressively increasing bias over time. For distinguishing Child A from Child B+C cirrhosis, ECV0 and ECV1 demonstrated comparable diagnostic performance with AUCs of 0.947 and 0.909, respectively. DeLong test confirmed no significant difference in the AUCs (P = 0.148). Similarly, when discriminating controls from Child A patients, ECV1 and ECV0 maintained comparable AUCs of 0.902 and 0.887, respectively. DeLong test confirmed no significant difference in the AUCs (P = 0.514). Notably, ECV measurements beyond 7 days (ECV2/ECV3) showed significantly reduced diagnostic efficacy (AUC range: 0.685 to 0.842, for all, P <0.05).

Conclusion: Hct measurements within a 7-day window of DECT can be reliably used for ECV quantification and Child-Pugh classification assessment in patients with cirrhosis. This approach provides clinicians with a convenient workflow by reducing the necessity for restrictive same-day phlebotomy.

Objective: Image quality evaluation in radiology is most relevant when reflects radiologists' performance. This study assessed how image quality measurement in terms of in vivo-characterized detectability index () for low-contrast liver lesion assessment in CT is correlated with radiologists' performance across 2 different CT reconstructions.

Methods: Fifty-one contrast-enhanced abdominal studies for investigating colorectal liver metastases were prospectively performed using 2 radiation dose exposures and reconstructed with Filtered back projection (FBP) and deep learning image reconstruction (DL) algorithms for a total of 161 noncalcified hypoattenuating lesions for 3 lesion size (D) subsets (<6 mm, 6 to 10 mm, and >10 mm). Images were assessed by expert radiologists for hepatic lesion detection task and likelihood of malignancy across the 2 imaging conditions. All cases were also evaluated automatically in terms of in vivo as a metric of task-based performance, both using a conventional technique and a new formalism of an added frequency term in the internal noise component of to accommodate the nonlinearity of the DL reconstruction (adj).

Results: The study found conventionally defined d' well-reflective of radiologists' evaluation of FBP images but not well-aligned with that of DL images. The new formalism provided more consistent reflection of performance across reconstruction techniques. In particular, in the lesion group D <=6 mm, the difference between radiologists' accuracy in images acquired with DL and images acquired with FBP was -26%, and the related adj difference was -9%, whereas the was 34%. Analogously, for the lesion group 6 mm < D <=10 mm, the differences were -15%, -13%, and 29%, respectively. Lastly, for the lesion group D>10 mm, radiologists showed the same accuracy in both FPB and DL images, difference in adj was -11%, and difference in was 31%.

Conclusion: The new formalism can robustly reflect CT systems clinical performance irrespective of reconstruction algorithm. The methodology can be more readily applied to assess the real-world performance of CT systems.

Objectives: To investigate the feasibility and image quality of artificial intelligence iterative reconstruction (AIIR) for computed tomography angiography (CTA) of the bronchial artery (BA) with a reduced radiation dose and contrast agent dosage.

Materials and methods: A total of 110 hemoptysis patients were prospectively enrolled for bronchial artery CTA (BA-CTA) and were randomly divided into 2 groups. Routine-dose group (group A, n=55) used a routine CTA protocol (tube voltage: 120 kVp; contrast dosage: 80 mL) with hybrid iterative reconstruction, while the low-dose group (group B, n=55) used the low-dose protocol (tube voltage: 100 kVp; contrast dosage: 50 mL) with AIIR. Attenuation values, noise, signal-to-noise ratio (SNR), and contrast-to-noise ratio (CNR) were measured for objective analysis. Subjective image quality was rated by 2 blinded radiologists using 5-point scales.

Results: No significant differences in demographic characteristics were observed between the 2 groups (all P>0.05). The radiation dose in group B was reduced by 73.8%, respectively, compared with group A. The mediastinal segment of BA was shown in both group images, while the hilar segment of BA was higher in group B than in group A (P<0.05). The mean subjective scores between the 2 groups showed no significant difference (all P>0.05), while SNR and CNR of group B were higher than those of group A (all P<0.0001).

Conclusions: The simultaneous reconstruction of BA-CTA images using the AIIR algorithm with reduced tube voltage and contrast agent dosage not only substantially reduces the radiation dose of preoperative BA-CTA for BAE but also achieves better image quality than routine-dose images.

Objective: To investigate the association of the pericoronary fat attenuation index (FAI) derived from coronary computed tomography angiography (CCTA) with target vessel revascularization (TVR) in symptomatic postprocedure patients.

Methods: A retrospective analysis was conducted, including 154 patients with 191 lesions scheduled for stenting who underwent invasive coronary angiography (ICA) after preinterventional CCTA. The proximal pericoronary FAI of the 3 major coronary arteries and lesion-specific pericoronary FAI were measured on preprocedure CCTA using semi-automated software. Lesions were randomly allocated to a training set (n=133) and a test set (n=58). Multivariate logistic regression analyses were performed to identify independent variables associated with TVR in the training cohort. Analyses were performed on the patient and vessel levels.

Results: A total of 154 patients (age 60.9±9.5 y, 68.8% male) with 191 lesions scheduled for stenting were included. On the basis of patient-level analysis, patients with TVR showed higher pericoronary FAI compared with patients without TVR. In vessel-level analysis, the regression model incorporating 1 cm 2 mm lesion-specific pericoronary FAI demonstrated superior diagnostic performance in both cohorts (training set AUC 0.814, 95% CI: 0.721-0.907; test set AUC 0.794, 95% CI: 0.659-0.928). The optimal cutoff value of -70.49 HU for the 1 cm 2 mm lesion-specific pericoronary FAI was determined by maximizing Youden's index, achieving a sensitivity of 75.0% and specificity of 63.0% in the test set. The model exhibited excellent calibration and clinical utility as confirmed by calibration curves and decision curve analysis (DCA). Multivariate logistic regression analysis showed that 1 cm 2 mm lesion-specific pericoronary FAI (OR 1.2, 95% CI: 1.01-1.42, P=0.036) was an independent predictor of TVR.

Conclusions: CCTA-derived pericoronary FAI is significantly associated with TVR in postprocedural patients. The 1 cm 2 mm lesion-specific pericoronary FAI, with an optimal cutoff of -70.49 HU, represents an effective tool for TVR risk stratification in this patient population.

Objective: To evaluate the impact of Super-Resolution Deep Learning Reconstruction (SR-DLR) (Canon Medical Systems Corporation) on image quality and myocardial hemodynamic parameters in dynamic myocardial computed tomography (CT) perfusion compared with filtered-back projection (FBP), hybrid iterative reconstruction (IR), and normal-resolution deep learning reconstruction (NR-DLR).

Methods: This prospective single-center study included 25 patients (mean age ± SD, 65 ± 10; 21 men) who underwent dynamic myocardial CT perfusion. For qualitative analysis, image noise, signal-to-noise ratio (SNR), and contrast-to-noise ratio (CNR) were assessed, while qualitative analysis included overall image quality and lesion visibility. Myocardial blood flow (MBF) at rest and stress, as well as coronary flow reserve (CFR) were analyzed. Image quality and hemodynamic parameters were compared across 4 reconstruction methods.

Results: SR-DLR achieved the lowest image noise (20.33 ± 2.45 HU), significantly lower than FBP (145.20 ± 74.81 HU), hybrid IR (47.19 ± 10.02 HU), and NR-DLR (22.92 ± 2.63 HU) (P < 0.001). In rest imaging, SR-DLR showed significantly higher SNR (6.71 ± 1.88) and CNR (15.41 ± 5.48) compared with other reconstruction methods (P < 0.001). Similar improvements were observed in stress imaging, with SR-DLR providing significantly enhanced SNR and CNR compared with all other methods. The mean CFR was 2.75 ± 1.88 for SR-DLR, 2.75 ± 1.99 for NR-DLR, 2.74 ± 2.44 for hybrid IR, and 2.56 ± 3.17 for FBP, with no statistically significant differences observed in any pairwise comparisons. Qualitative analysis showed that SR-DLR achieved the highest overall image quality and lesion visibility, significantly outperforming FBP and comparable to hybrid IR and NR-DLR.

Conclusions: SR-DLR and NR-DLR significantly enhanced image quality by reducing noise and improving SNR and CNR while maintaining hemodynamic quantification.