European Radiology Experimental最新文献

Background: This retrospective study aims to evaluate the impact of a content-based image retrieval (CBIR) application on diagnostic accuracy and confidence in interstitial lung disease (ILD) assessment using high-resolution computed tomography CT (HRCT).

Methods: Twenty-eight patients with verified pattern-based ILD diagnoses were split into two equal datasets (1 and 2). The images were assessed by two radiology residents (3rd and 5th year) and one expert radiologist in four sessions. Dataset 1 was used for sessions A and C, assessing diagnostic accuracy and confidence with mandatory and without CBIR software. Dataset 2 was used for sessions B and D with optional CBIR use, assessing time spending and frequency of CBIR usage. Accuracy was assessed on the CT pattern level, comparing readers' diagnoses with reference diagnoses and CBIR results with region-of-interest (ROI) patterns.

Results: Diagnostic accuracy and confidence of readers showed an increasing trend with CBIR use compared to no CBIR use (53.6% versus 35.7% and 50.0% versus 32.2%, respectively). Time for reading significantly decreased in both datasets (A versus C: 104 s versus 54 s, p < 0.001; B versus D: 88.5 s versus 70 s, p = 0.009), whereas time for research increased with CBIR software use (A versus C: 31 s versus 81 s, p = 0.040). CBIR results showed a high pattern-based accuracy of overall 73.4%. Comparison between readers indicates a slightly higher accuracy of CBIR results when more than one ROI was used as input (77.7% versus 70.1%).

Conclusion: CBIR software improves in-training radiologist diagnostic accuracy and confidence while reducing interpretation time in ILD assessment.

Relevance statement: Content-based image retrieval software improves the assessment of interstitial lung diseases (ILD) in high-resolution CT, especially for radiology residents, by increasing diagnostic accuracy and confidence while reducing interpretation time. This can provide educational benefits and more time-efficient management of complex cases.

Key points: A content-based image retrieval (CBIR) software improves diagnostic accuracy and confidence for in-training radiologists for interstitial lung disease (ILD) assessment on computed tomography (CT). A CBIR application provides condensed information about similar HRCT cases reducing time for ILD assessment. CBIR algorithms benefit from the input of multiple regions of interest per ILD case.

Background: We examined chronic gadolinium retention impact on gene expression in the mouse central nervous system (CNS) after injection of linear or macrocyclic gadolinium-based contrast agents (GBCAs).

Methods: From 05/2022 to 07/2023, 36 female mice underwent weekly intraperitoneal injections of gadodiamide (2.5 mmol/kg, linear), gadobutrol (2.5 mmol/kg, macrocyclic), or saline. Mice were sacrificed on day 29 or 391 after a 1-year washout. Assessments included magnetic resonance imaging (MRI), mechanical hyperalgesia tests, and inductively coupled plasma mass spectrometry to measure gadolinium levels. Ribonucleic acid (RNA) sequencing and bioinformatic analyses identified differentially expressed genes (DEGs), with validation by quantitative reverse transcription polymerase chain reaction (qRT-PCR) and western blot (WB).

Results: Post-gadodiamide, MRI showed increased signal intensity in the deep cerebellar nuclei (pre, 0.997 ± 0.006 versus post, 1.086 ± 0.013, p < 0.001). Mechanical hyperalgesia tests indicated transient sensory changes. After 1-year, gadolinium retention was noted in the brain (5.92 ± 0.32 nmol/kg) and spinal cord (1.23 ± 0.66 nmol/kg) with gadodiamide, compared to saline controls (0.06 ± 0.02 nmol/kg in brains and 0.28 ± 0.06 nmol/kg in spinal cords). RNA sequencing identified 17 shared DEGs between brain and spinal cord in the gadodiamide group on day 391, with altered Hmgb2 and Sgk1 expression confirmed by qRT-PCR and WB. Reactome pathway analysis showed enrichment in neuroinflammation pathways. No DEGs were detected in brains on day 29.

Conclusion: Chronic gadolinium deposition from repeated linear GBCA but not macrocyclic administration causes significant gene expression alterations in the mouse CNS, particularly affecting neuroinflammation pathways.

Relevance statement: This study examined the long-term impact of chronic gadolinium retention on gene expression in the mouse CNS, uncovering significant changes associated with neuroinflammation pathways after repeated administration of linear GBCA, but not with macrocyclic GBCA. These findings highlight the importance of further research on the long-term safety of linear GBCA in medical imaging.

Key points: Chronic gadolinium retention alters gene expression in the mouse central nervous system. Significant neuroinflammatory pathway changes were observed after linear gadodiamide exposure. MRI showed increased signal intensity in deep cerebellar nuclei after gadodiamide injection.

Background: We evaluated the accuracy of magnetic resonance imaging (MRI) computed tomography (CT)-like sequences compared to normal-resolution CT (NR-CT) and super-high-resolution CT (SHR-CT) for planning of cochlear implantation.

Methods: Six cadaveric temporal bone specimens were used. 3-T MRI scans were performed using radial volumetric interpolated breath-hold (STARVIBE), pointwise-encoding time reduction with radial acquisition (PETRA), and ultrashort time of echo (UTE) sequences. CT scans were performed on two scanners for SHR-CT and NR-CT acquisitions. Two radiologists evaluated accuracy based on preimplantation metrics and the ability to identify various anatomical structures, particularly the facial recess and round window. Wilcoxon rank-sum test and intraclass correlation coefficient (ICC) were used.

Results: The facial nerve was always clearly visible (score ≥ 2) in the MRI, NR-CT, and SHR-CT scans (p ≥ 0.621). However, the chorda tympani nerve (CTN) was clearly visualized in UTE, STARVIBE, and PETRA sequences in only 33% (2/6 specimens, p = 0.016), 50% (3/6 specimens, p = 0.038), and 83% (5/6 specimens, p = 0.017) of cases, respectively, whereas it was always clearly visualized in SHR and NR-CT (p = 0.426). The round window (RW) was never visualized in MRI sequences (p ≤ 0.010), whereas it was identified in all cases in SHR and NR-CT (p = 1.000). There was a strong correlation between measurements obtained from MRI and CT modalities (ICC ≥ 0.837).

Conclusion: MRI CT-like sequences assessed the facial nerve in all cases and the CTN in up to 87% of cases. However, the detection of the RW was insufficient for surgical planning. CT and MRI measurements were in agreement.

Relevance statement: CT-like MRI sequences can image the anatomy of the facial recess and the length of the basal turn of the cochlea with similar accuracy as conventional CT, although they cannot image the round window.

Key points: CT-like MRI sequences are not widely used in preoperative cochlear implantation imaging. CT-like sequences can image the facial recess as well as conventional CT. CT-like sequences can image the basal turn length of the cochlea as well as conventional CT. Round window depiction is not possible with CT-like MRI sequences.

Background: Photon-counting detector (PCD) technology has the potential to reduce noise in computed tomography (CT). This study aimed to carry out a voxelwise noise characterization for a clinical PCD-CT scanner with a model-based iterative reconstruction algorithm (QIR).

Methods: Forty repeated axial acquisitions (tube voltage 120 kV, tube load 200 mAs, slice thickness 0.4 mm) of a homogeneous water phantom and CTP404 module (Catphan-504) were performed. Water phantom acquisitions were also performed on a conventional energy-integrating detector (EID) scanner with a sinogram/image-based iterative reconstruction algorithm, using similar acquisition/reconstruction parameters. For smooth/sharp kernels, filtered back projection (FBP)- and iterative-reconstructed images were obtained. Noise maps, non-uniformity index (NUI) of noise maps, image noise histograms, and noise power spectrum (NPS) curves were computed.

Results: For FBP-reconstructed images of water phantom, mean noise was (smooth/sharp kernel) 11.7 HU/51.1 HU and 18.3 HU/80.1 HU for PCD-scanner and EID-scanner, respectively, with NUI values for PCD-scanner less than half those for EID-scanner. Percentage noise reduction increased with increasing iterative power, up to (smooth/sharp kernel) 57.7%/72.5% and 56.3%/70.1% for PCD-scanner and EID-scanner, respectively. For PCD-scanner, FBP- and QIR-reconstructed images featured an almost Gaussian distribution of noise values, whose shape did not appreciably vary with iterative power. Noise maps of CTP404 module showed increased NUI values with increasing iterative power, up to (smooth/sharp kernel) 15.7%/9.2%. QIR-reconstructed images showed limited low-frequency shift of NPS peak frequency.

Conclusion: PCD-CT allowed appreciably reducing image noise while improving its spatial uniformity. QIR algorithm decreases image noise without modifying its histogram distribution shape, and partly preserving noise texture.

Relevance statement: This phantom study corroborates the capability of photon-counting detector technology in appreciably reducing CT imaging noise and improving spatial uniformity of noise values, yielding a potential reduction of radiation exposure, though this needs to be assessed in more detail.

Key points: First voxelwise characterization of noise for a clinical CT scanner with photon-counting detector technology. Photon-counting detector technology has the capability to appreciably reduce CT imaging noise and improve spatial uniformity of noise values. In photon-counting CT, a model-based iterative reconstruction algorithm (QIR) allows decreasing effectively image noise. This is done without modifying noise histogram distribution shape, while limiting the low-frequency shift of noise power spectrum peak frequency.

Background: Several factors are frequently considered for outcome prediction rin stroke patients. We assessed the value of digital subtraction angiography (DSA)-based brain perfusion measurements after mechanical thrombectomy (MT) for outcome prediction in acute ischaemic stroke.

Methods: From DSA image data (n = 90; 38 females; age 73.3 ± 13.1 years [mean ± standard deviation]), time-contrast agent (CA) concentration curves were acquired, and maximum slope (MS), time to peak (TTP), and maximum CA concentration (CA_max) were calculated using an arterial input function. This data was used to predict neurological deficits at 24 h and upon discharge by using multiple regression analysis; the predictive capability was compared with the predictive power of the "Thrombolysis in cerebral infarction" (TICI) score. Intraclass correlation coefficients (ICC) of the NIHSS values were analysed.

Results: The comparison of means revealed a linear trend after stratification into TICI classes for CA_max (TICI 0: 0.07 ± 0.02 a.u. to TICI 3: 0.22 ± 0.07 a.u.; p < 0.001), and for MS (TICI 0: 0.04 ± 0.01 a.u./s to TICI 3: 0.12 ± 0.0  a.u./s; p < 0.001). Regression analyses demonstrated equivalent capabilities for estimating neurological deficits after 24 h and at discharge using both the TICI score and DSA-based perfusion parameters (ΔR² ~ 0.03). Compared to the actual NIHSS, the ICC ranged from 0.55 to 0.84 for DSA-based models and from 0.6 to 0.82 for TICI-based models.

Conclusion: Semi-quantitative evaluation of DSA-based perfusion parameters prior to and after MT is feasible and could enhance the objectivity and comparability of MT outcome prediction. This technique may offer novel approaches in acute ischaemic stroke management and data comparability.

Relevance statement: DSA-based brain perfusion measurements following interventional stroke therapy could allow for an experience-independent assessment of reperfusion success. It demonstrates predictive power at least equivalent to the established methods. This could support a future automated DSA-based brain perfusion measurement method.

Key points: Currently, the evaluation of stroke therapy success is based on the treating physician's experience. The present study introduces an objective semi-quantitative evaluation method. In predicting clinical outcomes, the traditional expert-based and semi-quantitative methods are equivalent.

Background: Non-malignant chronic diseases remain a major public health concern. Given the alterations in lipid metabolism and deposition in the lung and its association with fibrotic interstitial lung disease (fILD) and chronic obstructive pulmonary disease (COPD), this study aimed to detect those alterations using computed tomography (CT)-based analysis of pulmonary fat attenuation volume (CTpfav).

Methods: This observational retrospective single-center study involved 716 chest CT scans from three subcohorts: control (n = 279), COPD (n = 283), and fILD (n = 154). Fully automated quantification of CTpfav based on lung segmentation and HU-thresholding. The pulmonary fat index (PFI) was derived by normalizing CTpfav to the CT lung volume. Statistical analyses were conducted using Kruskal-Wallis with Dunn's post hoc tests.

Results: Patients with fILDs demonstrated a significant increase in CTpfav (median 71.0 mL, interquartile range [IQR] 59.7 mL, p < 0.001) and PFI (median 1.9%, IQR 2.4%, p < 0.001) when compared to the control group (CTpfav median 43.6 mL, IQR 16.94 mL; PFI median 0.9%, IQR 0.5%). In contrast, individuals with COPD exhibited significantly reduced CTpfav (median 36.2 mL, IQR 11.4 mL, p < 0.001) and PFI (median 0.5%, IQR 0.2%, p < 0.001).

Conclusion: The study underscores the potential of CTpfav and PFI as imaging biomarkers for detecting changes in lung lipid metabolism and deposition and demonstrates a possibility of tracking these alterations in patients with COPD and ILDs. Further research is needed to validate these findings and explore the clinical relevance of CTpfav and PFI in lung disease management.

Relevance statement: This study introduces a fully automated method for quantifying CTpfav, potentially establishing it as a new imaging biomarker for chronic lung diseases.

Key points: This retrospective observational study employed an open-source, automated algorithm for the quantification of CT pulmonary fat attenuation volume (CTpfav). Patients with fibrotic interstitial lung disease (fILD) showed a significantly higher CTpfav and pulmonary fat index (PFI), i.e., CTpfav/CT lung volume, compared to a control group. Patients with chronic obstructive pulmonary disease (COPD) showed significantly lower CTpfav and PFI compared to the control group. CTpfav and PFI may each serve as imaging biomarkers for various lung diseases and warrant further investigation.

Background: Whole lung transpulmonary chemoembolization using a combination of doxorubicin (DXO) and degradable starch microspheres (DSM-TPCE) might be a promising treatment option in soft tissue sarcoma. To pave the way for clinical studies, this study aimed to evaluate the short-term effects of DSM-TPCE with DXO using an ex vivo isolated lung perfusion (ILP) model.

Methods: Nine lung specimens retrieved from patients undergoing lobectomy underwent ex vivo ILP. In groups of three, lung specimens were either treated with sole DXO, sole DSM, or combined substances (DSM + DXO). During ex vivo ILP, histological samples were obtained from each lung every 15 min. Quantitative DXO analysis and histopathological grading of possible tissue damage using a five-point Likert scale was performed. Two-way repeated measures ANOVA tested for differences between treatment groups and changes over time.

Results: We created a preclinical ex vivo ILP model to simulate the effects of DSM-TPCE. In histopathological analysis, only two specimens, treated with only DXO, showed an increase in parenchymal damage over time. No significant effect of time (3.3%, p = 0.305) or group (23.3; p = 0.331) was identified. Within the lung tissue, the DXO concentration ranged from 205 to 1,244 ng/g. No significant effects could be detected regarding different treatment groups (4.9% of total variation, p = 0.103).

Conclusion: In an ex vivo ILP model using human lung lobes, the physiological effects of DSM-TPCE with DXO could be tested. Neither increased DXO concentrations in lung tissue nor histopathological changes indicating early lung toxicity were observed.

Relevance statement: An ex vivo ILP model using human lung specimens did not show any signs of early lung toxicity after transpulmonary chemoembolization with DXO. These results support further evaluation of DSM-TPCE in phase I/II trials.

Key points: Transpulmonary chemoembolization can be investigated in an ex vivo ILP model. DSM did not increase DXO in normal lung tissue. DSM did not increase parenchymal toxicity compared to the control groups.

Background: Noninvasive and functional imaging of the focal liver lesion (FLL) vasculature at microscopic scales is clinically challenging. We investigated the feasibility of using super-resolution ultrasound (SR-US) imaging for visualizing and quantifying the microvasculature of intraparenchymal FLLs.

Methods: Patients with FLLs between June 2022 and February 2023 were prospectively screened. Following bolus injection of microbubbles at clinical concentration, SR-US was performed using a high frame rate (350-500 Hz) modified ultrasound scanner and a convex array transducer with a central frequency of 3.1 MHz.

Results: In total, 47 pathologically proven FLLs at a depth of 5.7 ± 1.7 cm (mean ± standard deviation) were included: 30 hepatocellular carcinomas (HCC), 11 liver metastases (LM), and 6 focal nodular hyperplasias (FNH). The smallest detectable vessel size of the hepatic microvasculature was 128.4 ± 18.6 μm (mean ± standard deviation) at a depth of 8 cm. Significant differences were observed among the three types of lesions in terms of pattern categories, vessel density, minimum flow velocity, and perfusion index. We observed higher vessel density for FNH versus liver parenchyma (p < 0.001) as well as fractal dimension and local flow direction entropy value for FNH versus HCC (p = 0.002 and p < 0.001, respectively) and for FNH versus LM (p = 0.006 and p = 0.002, respectively).

Conclusion: Multiparametric SR-US showed that these three pathological types of FLLs have specific microvascular phenotypes. Vessel density, fractal dimension and local flow direction entropy served as valuable parameters in distinguishing between benign and malignant FLLs.

Trial registration: ClinicalTrials.gov (NCT06018142).

Relevance statement: Multiparametric SR-US imaging offers precise morphological and functional assessment of the microvasculature of intraparenchymal focal liver lesions, providing insights into tumor heterogeneity and angiogenesis.

Key points: Super-resolution (SR)-US imaging allowed morphological and functional evaluation of intraparenchymal hepatic lesion microvasculature. Hepatocellular carcinoma, liver metastasis, and focal nodular hyperplasia exhibit distinct microvascular architectures and hemodynamic profiles. Multiparametric microvasculature characterization via SR-US imaging facilitates the differentiation between benign and malignant microvascular phenotypes.

Objective: To explore the value of three-dimensional arterial spin labeling (ASL) and six diffusion magnetic resonance imaging (MRI) models in differentiating solid benign and malignant renal tumors.

Methods: This retrospective study included 89 patients with renal tumors. All patients underwent ASL and ZOOMit diffusion-weighted imaging (DWI) examinations and were divided into three groups: clear cell renal cell carcinoma (ccRCC), non-ccRCC, and benign renal tumors (BRT). The mean and peak renal blood flow (RBFmean and RBFpeak) from ASL and fourteen diffusion parameters from mono-exponential DWI (Mono_DWI), intravoxel incoherent motion (IVIM), diffusion kurtosis imaging (DKI), stretched exponential model (SEM), fractional order calculus (FROC), and continuous-time random-walk (CTRW) model were analyzed. Binary logistic regression was used to determine the optimal parameter combinations. The diagnostic performance of various MRI-derived parameters and their combinations was compared.

Results: Among the six diffusion models, the SEM model achieved the highest performance in differentiating ccRCC from non-ccRCC (area under the receiver operating characteristic curve [AUC] 0.880) and from BRT (AUC 0.891). IVIM model achieved the highest AUC (0.818) in differentiating non-ccRCC from BRT. Among all the MRI-derived parameters, RBFpeak combined with DKI_MK yielded the highest AUC (0.970) in differentiating ccRCC from non-ccRCC, and the combination of RBFpeak, SEM_DDC, and FROC_μ yielded the highest AUC (0.992) for differentiating ccRCC from BRT.

Conclusion: ASL and all diffusion models showed similar diagnostic performance in differentiating ccRCC from non-ccRCC or BRT, while the IVIM model performed better in distinguishing non-ccRCC from BRT. Combining ASL with diffusion models can provide additional value in predicting ccRCC.

Relevance statement: Considering the increasing detection rate of incidental renal masses, accurate discrimination of benign and malignant renal tumors is crucial for decision-making. Combining ASL with diffusion MRI models offers a promising solution to this clinical issue.

Key points: All assessed models were effective for differentiating ccRCC from non-ccRCC or BRT. ASL and all diffusion models showed similar performance in differentiating ccRCC from non-ccRCC or BRT. Combining ASL with diffusion models significantly improved diagnostic efficacy in predicting ccRCC. IVIM model could better differentiate non-ccRCC from BRT.

Surgical repair of abdominal aortic aneurism (AAA) with horseshoe kidney (HK) is challenging because of several accessory renal arteries (RAs), variable in number, branches, and vascular territories, with subsequent variable renal damage. The identification of RAs and vascular territories could contribute to surgical planning. We developed a semiautomatic presurgical computed tomography angiography (CTA)-based model to measure the renal volume of each RA, validated on postsurgical CTA in patients with HK treated for AAA. Renal parenchyma volume was extracted on both CTAs (Vol_Tot_pre and Vol_Tot_post) after labeling RAs ostia and vascular endpoints by two observers using a semiautomatic model by assigning each renal voxel to the closest vascular ending, obtaining volumes for each vascular territory. Number of RAs number was 4.0 ± 1.4 (mean ± standard deviation (SD)), Vol_Tot_pre 360 ± 76.5 cm³; kidney volume loss at surgery (KVLS) (Vol_Tot_pre minus Vol_Tot_post) 51.9 ± 35.4 cm³; percentage of kidney loss 15.2 ± 11.6%. KVLS and predicted kidney volume loss on preoperative CTA (PKVL) were strongly correlated (r = 0.93; p = 0.023). Interobserver agreement was good (mean bias = 0.000001 ± 1.96 SD of 19.1 cm³). Presurgical semiautomatic segmentation of vascular territories in patients with HK and AAA is feasible. RELEVANCE STATEMENT: This software allowed the preoperative calculation of renal volume perfused by each renal artery in the challenging association of the horseshoe kidney and abdominal aortic aneurism. It helps to determine the feasibility of surgical resection of arteries, thereby improving surgical planning and reducing the risk of postoperative renal function deterioration. KEY POINTS: The association between horseshoe kidney and abdominal aortic aneurism is a challenging condition that may require renal vascular resection. A semiautomatic model measures renal volume perfused by each artery on preoperative computed tomography angiography with high accuracy. Customized use of this tool could improve surgical management by determining which arteries can be safely resected during surgery.