Increased Retention of Gadolinium in the Inflamed Brain After Repeated Administration of Gadopentetate Dimeglumine: A Proof-of-Concept Study in Mice Combining ICP-MS and Micro– and Nano–SR-XRF Objectives The aim of this study was to determine in vivo if brain inflammation leads to increased gadolinium (Gd) retention in brain tissue after repeated applications of Gd-based contrast agents (GBCAs). Materials and Methods Experimental autoimmune encephalomyelitis (EAE) was induced in female SJL/J mice (n = 6). Experimental autoimmune encephalomyelitis and healthy control mice (n = 4) received 2.5 mmol/kg Gd-DTPA over 10 days (8 injections, cumulated dose of 20 mmol/kg), starting at day 14 post immunization when EAE mice reached the maximal clinical disability. In a group of mice, T1-weighted 2-dimensional RARE images were acquired before the first GBCA injection and 1 day after the last injection. Mice were killed either 1 day or 10 days after the last Gd application. From each single animal, a brain hemisphere was used for Gd detection using inductively coupled plasma mass spectrometry, whereas the other hemisphere was processed for histology and synchrotron x-ray fluorescence spectroscopy (SR-XRF) analysis. Results Gadolinium deposition in inflamed brains was mapped by SR-XRF 1 day after the last Gd-DTPA injections, although only mild signal hyperintensity was found on unenhanced T1-weighted images. In addition, using inductively coupled plasma mass spectrometry, we detected and quantified Gd in both healthy and EAE brains up to 10 days after the last injections. However, EAE mouse brains showed higher levels of Gd (mean ± SD, 5.3 ± 1.8 μg/g; range, 4.45–8.03 μg/g) with respect to healthy controls (mean ± SD, 2.4 ± 0.6 μg/g; range, 1.8–3.2 μg/g). By means of micro–SR-XRF, we identified submicrometric Gd hotspots in all investigated samples containing up to 5893 μg Gd/g tissue. Nano–SR-XRF further indicated that Gd small hotspots had an average size of ~160 nm diameter and were located in areas of high inflammatory activity. Conclusions After repeated administrations of Gd-DTPA, ongoing inflammation may facilitate the retention of Gd in the brain tissue. Thus, neuroinflammation should be considered as a risk factor in the recommendation on use of linear GBCA-enhanced MRI.

Computer-Aided Diagnosis of Pulmonary Fibrosis Using Deep Learning and CT Images Objectives The objective of this study is to assess the performance of a computer-aided diagnosis (CAD) system (INTACT system) for the automatic classification of high-resolution computed tomography images into 4 radiological diagnostic categories and to compare this with the performance of radiologists on the same task. Materials and Methods For the comparison, a total of 105 cases of pulmonary fibrosis were studied (54 cases of nonspecific interstitial pneumonia and 51 cases of usual interstitial pneumonia). All diagnoses were interstitial lung disease board consensus diagnoses (radiologically or histologically proven cases) and were retrospectively selected from our database. Two subspecialized chest radiologists made a consensual ground truth radiological diagnosis, according to the Fleischner Society recommendations. A comparison analysis was performed between the INTACT system and 2 other radiologists with different years of experience (readers 1 and 2). The INTACT system consists of a sequential pipeline in which first the anatomical structures of the lung are segmented, then the various types of pathological lung tissue are identified and characterized, and this information is then fed to a random forest classifier able to recommend a radiological diagnosis. Results Reader 1, reader 2, and INTACT achieved similar accuracy for classifying pulmonary fibrosis into the original 4 categories: 0.6, 0.54, and 0.56, respectively, with P > 0.45. The INTACT system achieved an F-score (harmonic mean for precision and recall) of 0.56, whereas the 2 readers, on average, achieved 0.57 (P = 0.991). For the pooled classification (2 groups, with and without the need for biopsy), reader 1, reader 2, and CAD had similar accuracies of 0.81, 0.70, and 0.81, respectively. The F-score was again similar for the CAD system and the radiologists. The CAD system and the average reader reached F-scores of 0.80 and 0.79 (P = 0.898). Conclusions We found that a computer-aided detection algorithm based on machine learning was able to classify idiopathic pulmonary fibrosis with similar accuracy to a human reader.

Preventive Effect of Changing Contrast Media in Patients With A Prior Mild Immediate Hypersensitivity Reaction to Gadolinium-Based Contrast Agent Objectives Currently, the prevention of recurrent immediate hypersensitivity reactions (HSRs) to contrast media (CM) requests premedication and changing the culprit contrast agent. However, strategies for the prevention of immediate HSRs to gadolinium-based magnetic resonance contrast agents (GBCAs) have not yet been established. This study aimed to evaluate the effectiveness of changing the contrast agent and single-dose premedication for HSR recurrence prevention in patients with a history of mild immediate HSR to GBCA. Materials and Methods The outcomes of patients with mild immediate HSR to GBCA who subsequently underwent enhanced magnetic resonance imaging between October 2012 and July 2017 were analyzed. The institutional CM monitoring system was retrospectively reviewed, and data on the application of premedication and choice of CM were obtained. Gadolinium-based magnetic resonance contrast agents were classified into 3 classes according to their molecular structure (macrocyclic ionic, macrocyclic nonionic, and linear ionic). Intravenous chlorpheniramine 4 mg, 30 minutes before the GBCA administration, or intravenous methylprednisolone sodium succinate 40 mg plus chlorpheniramine 4 mg, 1 hour before the GBCA administration, was administrated as premedication regimen. Recurrence rates of immediate HSR were compared according to prevention strategies. Results A total of 185 patients with a history of mild immediate HSR to GBCA were re-exposed to GBCA 397 times during the study period. The overall recurrence rate was 19.6% (78/397). Changing the culprit GBCA significantly reduced the recurrence rate, compared with reusing the culprit GBCA (6.9%, 9/130 and 25.8%, 69/267; P < 0.001). The recurrence rate was lowest when the GBCA was changed to a different molecular structure class from the culprit agent, followed by changing to CM with the same molecular structure and reusing the culprit GBCA (6.2%, 7/113 vs 11.8%, 2/17 vs 25.8%, 69/267; P < 0.001 with χ2 test for trend). Single-dose premedication demonstrated no significant prophylactic effect on recurrence (20.4%, 17/98 vs 17.3%, 61/299 with and without premedication, respectively; P = 0.509). Premedication in addition to changing CM also showed no additional prophylactic effect (7.2%, 7/97 and 6.1%, 2/33, respectively; P = 0.821). Conclusions Changing the CM from the culprit agent could reduce the chance of HSR recurrence in patients with prior mild immediate HSR to GBCA, especially when the CM was changed to one of a different molecular structure class. However, single-dose premedication administration did not show significant HSR recurrence rate difference.

Quantitative Assessment of Bone Metastasis in Prostate Cancer Using Synthetic Magnetic Resonance Imaging Objectives The aims of this study were to evaluate the feasibility of quantitative synthetic magnetic resonance imaging (SyMRI) for characterizing bone lesions in prostate cancer and to discriminate viable progressive osteoblastic bone metastasis from nonviable bone metastases with treatment-induced sclerosis during the treatment course. Materials and Methods This institutional review board–approved prospective study included 96 consecutive prostate cancer patients who underwent whole-body MRI including diffusion-weighted imaging at the time of staging at diagnosis or starting a new line of anticancer treatment. Additional synthetic MRI of the lumbosacral spine, pelvis, and proximal femurs was performed. A region of interest of 1.0 cm in diameter was set in each bone lesion by 2 independent readers who were blinded to bone lesions' diagnosis. Differences in SyMRI variables between the different bone lesions were compared with the Wilcoxon rank sum test, and associations of SyMRI variables with active disease were analyzed with logistic regression analysis. Performance of T1, T2, and proton density (PD) for diagnosing active disease was assessed using the area under the receiver operating characteristic curve. Results Ninety-three bone lesions were eligible for analysis. The PD values of active (viable) bone metastatic lesions were significantly higher than those of inactive (nonviable) bone metastatic lesions without sclerosis and those of red bone marrow (P < 0.001 for both readers). The PD values of inactive bone metastatic lesions with sclerosis were significantly lower than those of inactive bone metastatic lesions without sclerosis and red bone marrow (P < 0.001 for both readers). The PD value proved to be an independent significant indicator (P < 0.001) for differentiating bone lesions. The areas under the curve of T1/T2/PD for identifying active disease were 0.81/0.69/0.93 for reader 1 and 0.78/0.70/0.92 for reader 2, respectively. Conclusions Signal quantification on SyMRI provides objective assessment of bone lesions in the lower trunk. The PD value can be useful to determine the viability of bone metastases in prostate cancer.

Can Ex Vivo Magnetic Resonance Imaging of Rectal Cancer Specimens Improve the Mesorectal Lymph Node Yield for Pathological Examination? Purpose The aim of this study was to use 7 T ex vivo magnetic resonance imaging (MRI) scans to determine the size of lymph nodes (LNs) in total mesorectal excision (TME) specimens and to increase the pathological yield of LNs with MR-guided pathology. Materials and Methods Twenty-two fixated TME specimens containing adenocarcinoma were scanned on a 7 T preclinical MRI system with a T1-weighted 3-dimensional gradient echo sequence with frequency-selective lipid excitation (repetition time/echo time, 15/3 milliseconds; resolution, 0.293 mm3) and a water-excited 3-dimensional multigradient echo (repetition time, 30 milliseconds; computed echo time, 6.2 milliseconds; resolution, 0.293 mm3) pulse sequence. The first series of 11 TME specimens (S1) revealed the number and size of LNs on both ex vivo MRI and histopathology. The second series of 11 TME specimens (S2) was used to perform MR-guided pathology. The number, size, and percentages of yielded LNs of S1 and S2 were compared. Results In all specimens (22/22), a median number of 34 LNs (interquartile range, 26–34) was revealed on ex vivo MRI compared with 14 LNs (interquartile range, 7.5–21.5) on histopathology (P = 0.003). Mean size of all LNs did not differ between the 2 series (ex vivo MRI: 2.4 vs 2.5 mm, P = 0.267; pathology: 3.6 vs 3.5 mm, P = 0.653). The median percentages of harvested LNs compared with nodes visible on ex vivo MRI per specimen for both series were not significantly different (40% vs 43%, P = 0.718). By using a size threshold of greater than 2 mm, the percentage improved to 71% (S1) and to 78% (S2, P = 0.895). The median number of harvested LNs per specimen did not increase by performing MR-guided pathology (S1, 14 LNs; S2, 20 LNs; P = 0.532). Conclusions Ex vivo MRI visualizes more LNs than (MR-guided) pathology is able to harvest. Current pathological examination was not further improved by MR guidance. The majority of LNs or LN-like structures visible on ex vivo MRI below 2 mm in size remain unexplained, which warrants a 3-dimensional approach for pathological reconstruction of specimens.

Can Virtual Contrast Enhancement in Brain MRI Replace Gadolinium?: A Feasibility Study Objectives Gadolinium-based contrast agents (GBCAs) have become an integral part in daily clinical decision making in the last 3 decades. However, there is a broad consensus that GBCAs should be exclusively used if no contrast-free magnetic resonance imaging (MRI) technique is available to reduce the amount of applied GBCAs in patients. In the current study, we investigate the possibility of predicting contrast enhancement from noncontrast multiparametric brain MRI scans using a deep-learning (DL) architecture. Materials and Methods A Bayesian DL architecture for the prediction of virtual contrast enhancement was developed using 10-channel multiparametric MRI data acquired before GBCA application. The model was quantitatively and qualitatively evaluated on 116 data sets from glioma patients and healthy subjects by comparing the virtual contrast enhancement maps to the ground truth contrast-enhanced T1-weighted imaging. Subjects were split in 3 different groups: enhancing tumors (n = 47), nonenhancing tumors (n = 39), and patients without pathologic changes (n = 30). The tumor regions were segmented for a detailed analysis of subregions. The influence of the different MRI sequences was determined. Results Quantitative results of the virtual contrast enhancement yielded a sensitivity of 91.8% and a specificity of 91.2%. T2-weighted imaging, followed by diffusion-weighted imaging, was the most influential sequence for the prediction of virtual contrast enhancement. Analysis of the whole brain showed a mean area under the curve of 0.969 ± 0.019, a peak signal-to-noise ratio of 22.967 ± 1.162 dB, and a structural similarity index of 0.872 ± 0.031. Enhancing and nonenhancing tumor subregions performed worse (except for the peak signal-to-noise ratio of the nonenhancing tumors). The qualitative evaluation by 2 raters using a 4-point Likert scale showed good to excellent (3–4) results for 91.5% of the enhancing and 92.3% of the nonenhancing gliomas. However, despite the good scores and ratings, there were visual deviations between the virtual contrast maps and the ground truth, including a more blurry, less nodular-like ring enhancement, few low-contrast false-positive enhancements of nonenhancing gliomas, and a tendency to omit smaller vessels. These “features” were also exploited by 2 trained radiologists when performing a Turing test, allowing them to discriminate between real and virtual contrast-enhanced images in 80% and 90% of the cases, respectively. Conclusions The introduced model for virtual gadolinium enhancement demonstrates a very good quantitative and qualitative performance. Future systematic studies in larger patient collectives with varying neurological disorders need to evaluate if the introduced virtual contrast enhancement might reduce GBCA exposure in clinical practice.

Computed Tomography for 4-Dimensional Angiography and Perfusion Imaging of the Prostate for Embolization Planning of Benign Prostatic Hyperplasia Objectives The aim of this study was to evaluate the feasibility of a computed tomography (CT) protocol enabling the visualization of the prostatic artery (PA) before prostatic artery embolization (PAE) in benign prostatic hyperplasia, which provides quantitative perfusion information of the prostate gland. Materials and Methods In this institutional review board–approved study, 22 consecutive patients (mean age, 67 ± 7 years) who were planned to undergo PAE underwent a dynamic CT scan of the pelvis (scan range, 22.4 cm; cycle time, 1.5 seconds; scan time, 44 seconds; 25 scan cycles; 70 kVp; 100 mAs) after the administration of 70 mL of iodinated contrast media (flow rate, 6 mL/s; 10 seconds' delay). Image postprocessing consisted of a spatiotemporal, frequency-depending multiband filtering technique with noise reduction, motion correction, resulting in (1) time-resolved, temporal maximum intensity projection (MIP) images from fusion of multiple arterial time points; (2) 4-dimensional (4D) CT angiography images after bone and calcium plaque removal; and (3) parametric perfusion maps of the prostate. Intraprocedural cone-beam CT was performed with a microcatheter in the PA. In both modalities, the contrast-to-noise ratio of the right internal iliac artery or the PA was calculated, respectively. Visibility of the PA was scored using a Likert scale (score 1 = not seen, to score 4 = intraprostatic PA branches seen). Quantitative perfusion analysis of the dynamic pelvic CT included calculation of the blood flow, blood volume, mean transit time, and flow extraction product. Results The average volume CT dose index and dose length product of CT was 35.7 ± 6.8 mGy and 737.4 ± 146.3 mGy·cm, respectively. Contrast-to-noise ratio of the pelvic vessels on temporal MIP images and cone-beam CT were 45 ± 19 and 69 ± 27, respectively (P < 0.01). The mean visibility score of the PA was 3.6 ± 0.6 for 4D-CT angiography and 3.97 ± 0.2 for cone-beam CT (P < 0.001). The PA was visualized in 100% of 4D-CT angiography examinations, with one PA being visible only proximally. Prostate CT perfusion analysis showed blood flow, blood volume, mean transit time, and flow extraction product values of 27.9 ± 12.5 mL/100 mL/min, 2.0 ± 0.8 mL/100 mL, 4.5 ± 0.5 second, and 12.6 ± 5.4 mL/100 mL/min, respectively, for the whole prostate gland. About half the patients showed a pronounced difference between the lobes. Conclusions We introduced a CT protocol for PAE planning providing excellent visualization of the PA on temporal MIP images and 4D-CT angiography at a reasonable dose and low contrast volume. In addition, quantitative perfusion information is available, which might be useful for outcome prediction after embolization.

Multiparametric Quantitative MRI for the Detection of IgA Nephropathy Using Tomoelastography, DWI, and BOLD Imaging Objectives The aim of this study was to noninvasively evaluate changes in renal stiffness, diffusion, and oxygenation in patients with chronic, advanced stage immunoglobulin A nephropathy (IgAN) by multiparametric magnetic resonance imaging using tomoelastography, diffusion-weighted imaging (DWI), and blood oxygen level–dependent (BOLD) imaging. Materials and Methods In this prospective study, 32 subjects (16 patients with biopsy-proven IgAN and 16 age- and sex-matched healthy controls) underwent multifrequency magnetic resonance elastography with tomoelastography postprocessing at 4 frequencies from 40 to 70 Hz to generate shear wave speed (meter per second) maps reflecting tissue stiffness. In addition, DWI and BOLD imaging were performed to determine the apparent diffusion coefficient in square millimeter per second and T2* relaxation time in milliseconds, respectively. Regions including the entire renal parenchyma of both kidneys were analyzed. Areas under the receiver operating characteristic (AUCs) curve were calculated to test diagnostic performance. Clinical parameters such as estimated glomerular filtration rate and protein-to-creatinine ratio were determined and correlated with imaging findings. Results Success rates of tomoelastography, DWI, and BOLD imaging regarding both kidneys were 100%, 91%, and 87%, respectively. Shear wave speed was decreased in IgAN (−21%, P < 0.0001), accompanied by lower apparent diffusion coefficient values (−12%, P = 0.004). BOLD imaging was not sensitive to IgAN (P = 0.12). Tomoelastography detected IgAN with higher diagnostic accuracy than DWI (area under the curve = 0.9 vs 0.8) and positively correlated with estimated glomerular filtration rate (r = 0.66, P = 0.006). Conclusions Chronic, advanced stage IgAN is associated with renal softening and restricted water diffusion. Tomoelastography is superior to DWI and BOLD imaging in detecting IgAN.