Investigative Radiology最新文献

Objectives: Small lesions are the limiting factor for reducing gadolinium-based contrast agents in brain magnetic resonance imaging (MRI). The purpose of this study was to compare the sensitivity and precision in metastasis detection on true contrast-enhanced T1-weighted (T1w) images and artificial images synthesized by a deep learning method using low-dose images.

Materials and methods: In this prospective, multicenter study (5 centers, 12 scanners), 917 participants underwent brain MRI between October 2021 and March 2023 including T1w low-dose (0.033 mmol/kg) and full-dose (0.1 mmol/kg) images. Forty participants with metastases or unremarkable brain findings were evaluated in a reading (mean age ± SD, 54.3 ± 15.1 years; 24 men). True and artificial T1w images were assessed for metastases in random order with 4 weeks between readings by 2 neuroradiologists. A reference reader reviewed all data to confirm metastases. Performances were compared using mid- P McNemar tests for sensitivity and Wilcoxon signed rank tests for false-positive findings.

Results: The reference reader identified 97 metastases. The sensitivity of reader 1 did not differ significantly between sequences (sensitivity [precision]: true, 66.0% [98.5%]; artificial, 61.9% [98.4%]; P = 0.38). With a lower precision than reader 1, reader 2 found significantly more metastases using true images (sensitivity [precision]: true, 78.4% [87.4%]; artificial, 60.8% [80.8%]; P < 0.001). There was no significant difference in sensitivity for metastases ≥5 mm. The number of false-positive findings did not differ significantly between sequences.

Conclusions: One reader showed a significantly higher overall sensitivity using true images. The similar detection performance for metastases ≥5 mm is promising for applying low-dose imaging in less challenging diagnostic tasks than metastasis detection.

Objectives: Artificial intelligence (AI) is thought to improve lesion detection. However, a lack of knowledge about human performance prevents a comparative evaluation of AI and an accurate assessment of its impact on clinical decision-making. The objective of this work is to quantitatively evaluate the ability of humans to detect focal cortical dysplasia (FCD), compare it to state-of-the-art AI, and determine how it may aid diagnostics.

Materials and methods: We prospectively recorded the performance of readers in detecting FCDs using single points and 3-dimensional bounding boxes. We acquired predictions of 3 AI models for the same dataset and compared these to readers. Finally, we analyzed pairwise combinations of readers and models.

Results: Twenty-eight readers, including 20 nonexpert and 5 expert physicians, reviewed 180 cases: 146 subjects with FCD (median age: 25, interquartile range: 18) and 34 healthy control subjects (median age: 43, interquartile range: 19). Nonexpert readers detected 47% (95% confidence interval [CI]: 46, 49) of FCDs, whereas experts detected 68% (95% CI: 65, 71). The 3 AI models detected 32%, 51%, and 72% of FCDs, respectively. The latter, however, also predicted more than 13 false-positive clusters per subject on average. Human performance was improved in the presence of a transmantle sign ( P < 0.001) and cortical thickening ( P < 0.001). In contrast, AI models were sensitive to abnormal gyration ( P < 0.01) or gray-white matter blurring ( P < 0.01). Compared with single experts, expert-expert pairs detected 13% (95% CI: 9, 18) more FCDs ( P < 0.001). All AI models increased expert detection rates by up to 19% (95% CI: 15, 24) ( P < 0.001). Nonexpert+AI pairs could still outperform single experts by up to 13% (95% CI: 10, 17).

Conclusions: This study pioneers the comparative evaluation of humans and AI for FCD lesion detection. It shows that AI and human predictions differ, especially for certain MRI features of FCD, and, thus, how AI may complement the diagnostic workup.

Objectives: 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.

Materials and methods: 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.

Results: 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.

Conclusions: Accurate, sensitive, and specific diagnosis of pulmonary chronic thromboembolism at low radiation dose is possible using iodine maps reconstructed from PCD-CT scans.

Aim: Peripheral nerve scarring is a severe yet common complication following nerve injury or surgery that can lead to impaired nerve function, including chronic pain and sensory or motor deficits. In this study, we aimed to establish high-resolution magnetic resonance neurography (MRN) to accurately visualize and monitor de novo-formed epineural fibrotic adhesions (EFAs) of the sciatic nerve in a rat nerve injury model.

Methods: Employing an established model to induce overshooting EFA, the study included 3 experimental groups of animals (n = 6 each): a positive control group (PC), an intervention group (IG), and a sham group. All groups underwent surgical nerve exposure: both PC and IG received an application of 10 μL 2.5% glutaraldehyde to induce EFA, but only IG received an additional preventive wrapping of the nerve with a collagen-containing matrix. Magnetic resonance imaging was performed 6, 8, and 12 weeks postoperatively using a standardized protocol including T2w and T1w without and with contrast media. Motor function and nerve regeneration was assessed using the visual static sciatic index. Histological specimens were obtained 12 weeks postoperatively and correlated with imaging.

Results: On high-resolution MRN, prominently contrast-enhancing epineural sleeves were present in vivo, which corresponded to histologically confirmed EFA (ratio of EFA to nerve area MRN 1.512 ± 0.106 vs histological ratio 1.459 ± 0.208, nonsignificant). As expected, average EFA in IG (0.310 ± 0.118 mm2) was smaller than in PC (0.909 ± 0.212 mm2, P < 0.01). Also, the average EFA in sham (0.386 ± 0.030 mm2) was less pronounced than in PC (P < 0.01). There was no significant difference in the average EFA between IG und sham. The EFA correlated with the functional outcome, which was measured by visual static sciatic index (correlation coefficient -0.59, P < 0.05).

Conclusions: The results of the present study for the first time confirm the clinical observation that epineural thickening on contrast-enhanced T1w imaging following manipulation to a nerve indeed corresponds to overshooting epineural scarring, which may be linked to impaired nerve function. This can be followed noninvasively in vivo over time providing an important basis for clinical decision-making in cases where further invasive therapies may be necessary.

Objectives: This study evaluates the impact of liver steatosis on the discriminative ability for liver fibrosis and inflammation using a novel Dixon water-only fat-corrected Look-Locker T1 mapping sequence, compared with a standard shortened Modified Look-Locker Inversion Recovery (shMOLLI) sequence, with the aim of overcoming the limitation of steatosis-related confounding in liver T1 mapping.

Materials and methods: 3 T magnetic resonance imaging of the liver including the 2 T1 mapping sequences and proton density fat fraction (PDFF) was prospectively performed in 24 healthy volunteers and 38 patients with histologically proven liver fibrosis evaluated within 90 days of liver biopsy. Paired Mann-Whitney test compared sequences between participants with and without significant liver steatosis (PDFF cutoff 10%), and unpaired Kruskal-Wallis test compared healthy volunteers to patients with early (F0-2) and advanced (F3-4) liver fibrosis, as well as low (A0-1) and marked (A2-3) inflammatory activity. Univariate and multivariate logistic regression models assessed the impact of liver steatosis on both sequences.

Results: Dixon_W T1 was higher than shMOLLI T1 in participants without steatosis (median 896 ms vs 890 ms, P = 0.04), but lower in participants with liver steatosis (median 891 ms vs 973 ms, P < 0.001). Both methods accurately differentiated between volunteers and patients with early and advanced fibrosis (Dixon_W 849 ms, 910 ms, 947 ms, P = 0.011; shMOLLI 836 ms, 918 ms, 978 ms, P < 0.001), and those with mild and marked inflammation (Dixon_W 849 ms, 896 ms, 941 ms, P < 0.01; shMOLLI 836 ms, 885 ms, 978 ms, P < 0.001). Univariate logistic regression showed slightly lower performance of the Dixon_W sequence in differentiating fibrosis (0.69 vs 0.73, P < 0.01), compensated by adding liver PDFF in the multivariate model (0.77 vs 0.75, P < 0.01).

Conclusions: Dixon water-only fat-corrected Look-Locker T1 mapping accurately identifies liver fibrosis and inflammation, with less dependency on liver steatosis than the widely adopted shMOLLI T1 mapping technique, which may improve its predictive value for these conditions.

Rationale and objectives: This study investigates the performance of tomosynthesis in the presence of osteosynthetic implants, aiming to overcome superimposition-induced limitations in conventional radiograms.

Materials and methods: After surgical fracture induction and subsequent osteosynthesis, 8 cadaveric fracture models (wrist, metacarpus, ankle, metatarsus) were scanned with the prototypical tomosynthesis mode of a multiuse x-ray system. Tomosynthesis protocols at 60, 80, and 116 kV (sweep angle 10°, 13 FPS) were compared with standard radiograms. Five radiologists independently rated diagnostic assessability based on an equidistant 7-point scale focusing on fracture delineation, intra-articular screw placement, and implant positioning. The intraclass correlation coefficient (ICC) was calculated to analyze interrater agreement.

Results: Radiation dose in radiography was 0.48 ± 0.26 dGy·cm 2 versus 0.12 ± 0.01, 0.36 ± 0.02, and 1.95 ± 0.11 dGy·cm 2 for tomosynthesis scans at 60, 80, and 116 kV. Delineation of fracture lines was superior for 80/116 kV tomosynthesis compared with radiograms ( P ≤ 0.003). Assessability of intra-articular screw placement was deemed favorable for all tomosynthesis protocols ( P ≤ 0.004), whereas superiority for evaluation of implant positioning could not be ascertained (all P 's ≥ 0.599). Diagnostic confidence was higher for 80/116 kV tomosynthesis versus radiograms and 60 kV tomosynthesis ( P ≤ 0.002). Interrater agreement was good for fracture delineation (ICC, 0.803; 95% confidence interval [CI], 0.598-0.904), intra-articular screw placement (ICC, 0.802; 95% CI, 0.599-0.903), implant positioning (ICC, 0.855; 95% CI, 0.729-0.926), and diagnostic confidence (ICC, 0.842; 95% CI, 0.556-0.934).

Conclusions: In the postoperative workup of extremity fractures, tomosynthesis allows for superior assessment of fracture lines and intra-articular screw positioning with greater diagnostic confidence at radiation doses comparable to conventional radiograms.

Objectives: The aim of this study was to determine whether MRI radiomic features of key cerebral structures differ between women and men, and whether detection of such differences depends on the image resolution.

Materials and methods: Ultrahigh resolution (UHR) 3D MP2RAGE (magnetization-prepared 2 rapid acquisition gradient echo) T1-weighted MR images (voxel size, 0.7 × 0.7 × 0.7 mm 3 ) of the brain of 30 subjects (18 women and 12 men; mean age, 39.0 ± 14.8 years) without abnormal findings on MRI were retrospectively included. MRI was performed on a whole-body 7 T MR system. A convolutional neural network was used to segment the following structures: frontal cortex, frontal white matter, thalamus, putamen, globus pallidus, caudate nucleus, and corpus callosum. Eighty-seven radiomic features were extracted respectively: gray-level histogram (n = 18), co-occurrence matrix (n = 24), run-length matrix (n = 16), size-zone matrix (n = 16), and dependence matrix (n = 13). Feature extraction was performed at UHR and, additionally, also after resampling to 1.4 × 1.4 × 1.4 mm 3 voxel size (standard clinical resolution). Principal components (PCs) of radiomic features were calculated, and independent samples t tests with Cohen d as effect size measure were used to assess differences in PCs between women and men for the different cerebral structures.

Results: At UHR, at least a single PC differed significantly between women and men in 6/7 cerebral structures: frontal cortex ( d = -0.79, P = 0.042 and d = -1.01, P = 0.010), frontal white matter ( d = -0.81, P = 0.039), thalamus ( d = 1.43, P < 0.001), globus pallidus ( d = 0.92, P = 0.020), caudate nucleus ( d = -0.83, P = 0.039), and corpus callosum ( d = -0.97, P = 0.039). At standard clinical resolution, only a single PC extracted from the corpus callosum differed between sexes ( d = 1.05, P = 0.009).

Conclusions: Nonnegligible differences in radiomic features of several key structures of the brain exist between women and men, and need to be accounted for. Very high spatial resolution may be required to uncover and further investigate the sexual dimorphism of brain structures on MRI.

Objectives: Response Evaluation Criteria in Solid Tumors (RECIST) 1.1 is a clinical and research standard for evaluating malignant tumors and lymph node metastasis. However, quantitative analysis of nodal status is limited to measurement of short axis diameter (SAD), and metastatic lymph nodes below 10 mm in SAD are often not detected. The purpose of this study was to evaluate the value of multifrequency magnetic resonance elastography (MRE) when added to RECIST 1.1 for detection of lymph node metastasis.

Materials and methods: Twenty-five benign and 82 metastatic lymph nodes were prospectively examined by multifrequency MRE at 1.5 T using tomoelastography postprocessing at 30, 40, 50, and 60 Hz (total scan time of 4 minutes). Shear wave speed as a surrogate of soft tissue stiffness was provided in m/s. Positron emission tomography-computed tomography was used as reference standard for identification of abdominal lymph node metastasis from histologically confirmed primary tumors. The diagnostic performance of MRE was compared with that of SAD according to RECIST 1.1 and evaluated by receiver operating characteristic curve analysis using generalized linear mixed models and binary logistic mixed models. Sensitivity, specificity, and predictive values were calculated for different cutoffs.

Results: Metastatic lymph nodes (1.90 ± 0.57 m/s) were stiffer than benign lymph nodes (0.98 ± 0.20 m/s, P < 0.001). An area under the curve of 0.95 for a cutoff of 1.32 m/s was calculated. Using a conservative approach with 1.0 specificity, we found sensitivity (SAD/MRE/MRE + SAD, 0.56/0.84/0.88), negative predictive values (0.41/0.66/0.71), and overall accuracy (0.66/0.88/0.91) to be improved using MRE and even higher for combined MRE and SAD.

Conclusions: Multifrequency MRE improves metastatic abdominal lymph node detection by 25% based on higher tissue stiffness-even for lymph nodes with an SAD ≤10 mm. Stiffness information is quick to obtain and would be a promising supplement to RECIST.

Objectives: The aim of this study was to assess the interreader reliability and per-RCC sensitivity of high-resolution photon-counting computed tomography (PCCT) in the detection and characterization of renal masses in comparison to MRI.

Materials and methods: This prospective study included 24 adult patients (mean age, 52 ± 14 years; 14 females) who underwent PCCT (using an investigational whole-body CT scanner) and abdominal MRI within a 3-month time interval and underwent surgical resection (partial or radical nephrectomy) with histopathology (n = 70 lesions). Of the 24 patients, 17 had a germline mutation and the remainder were sporadic cases. Two radiologists (R1 and R2) assessed the PCCT and corresponding MRI studies with a 3-week washout period between reviews. Readers recorded the number of lesions in each patient and graded each targeted lesion's characteristic features, dimensions, and location. Data were analyzed using a 2-sample t test, Fisher exact test, and weighted kappa.

Results: In patients with von Hippel-Lindau mutation, R1 identified a similar number of lesions suspicious for neoplasm on both modalities (51 vs 50, P = 0.94), whereas R2 identified more suspicious lesions on PCCT scans as compared with MRI studies (80 vs 56, P = 0.12). R1 and R2 characterized more lesions as predominantly solid in MRIs (R1: 58/70 in MRI vs 52/70 in PCCT, P < 0.001; R2: 60/70 in MRI vs 55/70 in PCCT, P < 0.001). R1 and R2 performed similarly in detecting neoplastic lesions on PCCT and MRI studies (R1: 94% vs 90%, P = 0.5; R2: 73% vs 79%, P = 0.13).

Conclusions: The interreader reliability and per-RCC sensitivity of PCCT scans acquired on an investigational whole-body PCCT were comparable to MRI scans in detecting and characterizing renal masses.

Clinical relevance statement: PCCT scans have comparable performance to MRI studies while allowing for improved characterization of the internal composition of lesions due to material decomposition analysis. Future generations of this imaging modality may reveal additional advantages of PCCT over MRI.

Objectives: Dark-blood late gadolinium enhancement (DB-LGE) cardiac magnetic resonance has been proposed as an alternative to standard white-blood LGE (WB-LGE) imaging protocols to enhance scar-to-blood contrast without compromising scar-to-myocardium contrast. In practice, both DB and WB contrasts may have clinical utility, but acquiring both has the drawback of additional acquisition time. The aim of this study was to develop and evaluate a deep learning method to generate synthetic WB-LGE images from DB-LGE, allowing the assessment of both contrasts without additional scan time.

Materials and methods: DB-LGE and WB-LGE data from 215 patients were used to train 2 types of unpaired image-to-image translation deep learning models, cycle-consistent generative adversarial network (CycleGAN) and contrastive unpaired translation, with 5 different loss function hyperparameter settings each. Initially, the best hyperparameter setting was determined for each model type based on the Fréchet inception distance and the visual assessment of expert readers. Then, the CycleGAN and contrastive unpaired translation models with the optimal hyperparameters were directly compared. Finally, with the best model chosen, the quantification of scar based on the synthetic WB-LGE images was compared with the truly acquired WB-LGE.

Results: The CycleGAN architecture for unpaired image-to-image translation was found to provide the most realistic synthetic WB-LGE images from DB-LGE images. The results showed that it was difficult for visual readers to distinguish if an image was true or synthetic (55% correctly classified). In addition, scar burden quantification with the synthetic data was highly correlated with the analysis of the truly acquired images. Bland-Altman analysis found a mean bias in percentage scar burden between the quantification of the real WB and synthetic white-blood images of 0.44% with limits of agreement from -10.85% to 11.74%. The mean image quality of the real WB images (3.53/5) was scored higher than the synthetic white-blood images (3.03), P = 0.009.

Conclusions: This study proposed a CycleGAN model to generate synthetic WB-LGE from DB-LGE images to allow assessment of both image contrasts without additional scan time. This work represents a clinically focused assessment of synthetic medical images generated by artificial intelligence, a topic with significant potential for a multitude of applications. However, further evaluation is warranted before clinical adoption.