Search

Background MRI is frequently used for early diagnosis of axial spondyloarthritis (axSpA). However, evaluation is time-consuming and requires profound expertise because noninflammatory degenerative changes can mimic axSpA, and early signs may therefore be missed. Deep neural networks could function as assistance for axSpA detection. Purpose To create a deep neural network to detect MRI changes in sacroiliac joints indicative of axSpA. Materials and Methods This retrospective multicenter study included MRI examinations of five cohorts of patients with clinical suspicion of axSpA collected at university and community hospitals between January 2006 and September 2020. Data from four cohorts were used as the training set, and data from one cohort as the external test set. Each MRI examination in the training and test sets was scored by six and seven raters, respectively, for inflammatory changes (bone marrow edema, enthesitis) and structural changes (erosions, sclerosis). A deep learning tool to detect changes indicative of axSpA was developed. First, a neural network to homogenize the images, then a classification network were trained. Performance was evaluated with use of area under the receiver operating characteristic curve (AUC), sensitivity, and specificity.

An abdominal aortic aneurysm (AAA) is a permanent dilatation of the abdominal aorta, usually accompanied by thrombus formation. The current clinical imaging modalities cannot reliably visualize the thrombus composition. Remodeling of the extracellular matrix (ECM) during AAA development leads to stiffness changes, providing a potential imaging marker. 14 apolipoprotein E-deficient mice underwent surgery for angiotensin II-loaded osmotic minipump implantation. 4 weeks post-op, 5 animals developed an AAA. The aneurysm was imaged ex vivo by microscopic multifrequency magnetic resonance elastography (µMMRE) with an in-plane resolution of 40 microns. Experiments were performed on a 7-Tesla preclinical magnetic resonance imaging scanner with drive frequencies between 1000 Hz and 1400 Hz. Shear wave speed (SWS) maps indicating stiffness were computed based on tomoelastography multifrequency inversion. As control, the aortas of 5 C57BL/6J mice were examined with the same imaging protocol. The regional variation of SWS in the thrombus ranging from 0.44±0.07 to 1.20±0.31 m/s was correlated fairly strong with regional histology-quantified ECM accumulation (R2=0.79). Our results suggest that stiffness changes in aneurysmal thrombus reflect ECM remodeling, which is critical for AAA risk assessment. In the future, µMMRE could be used for a mechanics-based clinical characterization of AAAs in patients. STATEMENT OF SIGNIFICANCE: : To our knowledge, this is the first study mapping the stiffness of abdominal aortic aneurysms with microscopic resolution of 40 µm. Our work revealed that stiffness critically changes due to extracellular matrix (ECM) remodeling in the aneurysmal thrombus. We were able to image various levels of ECM remodeling in the aneurysm reflected in distinct shear wave speed patterns with a strong correlation to regional histology-quantified ECM accumulation. The generated results are significant for the application of microscopic multifrequency magnetic resonance elastography for quantification of pathological remodeling of the ECM and may be of great interest for detailed characterization of AAAs in patients.

Ferumoxytol is an ultrasmall iron oxide nanoparticle, originally approved in 2009 by the FDA for IV treatment of iron deficiency in adults with chronic kidney disease. Subsequently, its off-label use as an MRI contrast agent has increased in clinical practice, particularly in pediatric patients in North America. Unlike conventional MRI contrast agents that are based on the rare earth metal gadolinium [gadolinium-based contrast agents (GBCAs)], ferumoxytol is biodegradable and carries no potential risk of nephrogenic systemic fibrosis. At FDA-approved doses, ferumoxytol demonstrates no long-term tissue retention in patients with intact iron metabolism. Ferumoxytol provides unique MRI properties including long-lasting vascular retention (facilitating high-quality vascular imaging) and retention in reticuloendothelial system tissues, thereby supporting a variety of applications beyond those possible with GBCAs. This Clinical Perspective describes clinical and early translational applications of ferumoxytol-enhanced MRI in children and young adults through off-label use for a variety of settings, including vascular, cardiac, and cancer imaging, drawing on the authors' institutional experience. In addition, we describe current preclinical and clinical research advances using ferumoxytol with regard to cellular and molecular imaging, and also as a novel potential cancer therapeutic agent.

UNSTRUCTURED Generative models, such as DALL-E 2 (OpenAI), could represent promising future tools for image generation, augmentation, and manipulation for artificial intelligence research in radiology, provided that these models have sufficient medical domain knowledge. Herein, we show that DALL-E 2 has learned relevant representations of x-ray images, with promising capabilities in terms of zero-shot text-to-image generation of new images, the continuation of an image beyond its original boundaries, and the removal of elements; however, its capabilities for the generation of images with pathological abnormalities (eg, tumors, fractures, and inflammation) or computed tomography, magnetic resonance imaging, or ultrasound images are still limited. The use of generative models for augmenting and generating radiological data thus seems feasible, even if the further fine-tuning and adaptation of these models to their respective domains are required first.

Background and Objectives: Bedside chest radiographs (CXRs) are challenging to interpret but impor-tant for monitoring cardiothoracic disease and invasive therapy devices in critical care and emergency medicine. Taking surrounding anatomy into account is likely to improve the diagnostic accuracy of artifi-cial intelligence and bring its performance closer to that of a radiologist. Therefore, we aimed to develop a deep convolutional neural network for efficient automatic anatomy segmentation of bedside CXRs.Methods: To improve the efficiency of the segmentation process, we introduced a "human-in-the-loop" segmentation workflow with an active learning approach, looking at five major anatomical structures in the chest (heart, lungs, mediastinum, trachea, and clavicles). This allowed us to decrease the time needed for segmentation by 32% and select the most complex cases to utilize human expert annotators effi-ciently. After annotation of 2,0 0 0 CXRs from different Level 1 medical centers at Charite - University Hospital Berlin, there was no relevant improvement in model performance, and the annotation process was stopped. A 5-layer U-ResNet was trained for 150 epochs using a combined soft Dice similarity coeffi-cient (DSC) and cross-entropy as a loss function. DSC, Jaccard index (JI), Hausdorff distance (HD) in mm, and average symmetric surface distance (ASSD) in mm were used to assess model performance. External validation was performed using an independent external test dataset from Aachen University Hospital ( n = 20).Results: The final training, validation, and testing dataset consisted of 1900/50/50 segmentation masks for each anatomical structure. Our model achieved a mean DSC/JI/HD/ASSD of 0.93/0.88/32.1/5.8 for the lung, 0.92/0.86/21.65/4.85 for the mediastinum, 0.91/0.84/11.83/1.35 for the clavicles, 0.9/0.85/9.6/2.19 for the trachea, and 0.88/0.8/31.74/8.73 for the heart. Validation using the external dataset showed an overall robust performance of our algorithm. Conclusions: Using an efficient computer-aided segmentation method with active learning, our anatomy -based model achieves comparable performance to state-of-the-art approaches. Instead of only segmenting the non-overlapping portions of the organs, as previous studies did, a closer approximation to actual anatomy is achieved by segmenting along the natural anatomical borders. This novel anatomy approach could be useful for developing pathology models for accurate and quantifiable diagnosis.(c) 2023 Elsevier B.V. All rights reserved.

Generative models such as DALL-E 2 could represent a promising future tool for image generation, augmentation, and manipulation for artificial intelligence research in radiology provided that these models have sufficient medical domain knowledge. Here we show that DALL-E 2 has learned relevant representations of X-ray images with promising capabilities in terms of zero-shot text-to-image generation of new images, continuation of an image beyond its original boundaries, or removal of elements, while pathology generation or CT, MRI, and ultrasound images are still limited. The use of generative models for augmenting and generating radiological data thus seems feasible, even if further fine-tuning and adaptation of these models to the respective domain is required beforehand., Comment: 4 Figures

Objective This study aimed to improve the accuracy of CT for detection of COVID-19-associated pneumonia and to identify patient subgroups who might benefit most from CT imaging. Methods A total of 269 patients who underwent CT for suspected COVID-19 were included in this retrospective analysis. COVID-19 was confirmed by reverse-transcription-polymerase-chain-reaction. Basic demographics (age and sex) and initial vital parameters (O2-saturation, respiratory rate, and body temperature) were recorded. Generalized mixed models were used to calculate the accuracy of vital parameters for detection of COVID-19 and to evaluate the diagnostic accuracy of CT. A clinical score based on vital parameters, age, and sex was established to estimate the pretest probability of COVID-19 and used to define low, intermediate, and high risk groups. A p-value of, Highlights • A simple score to estimate the pre-test probability of COVID-19 can be calculated using vital parameters, basic vital parameters. • Accuracy of CT for the detection of COVID-19 might be increased by selecting patients with a high-pretest probability. • In patients with mild symptoms, CT examinations are unlikel unlikely to be effective with a very low positive predictive value.

Purpose Emphysema and chronic obstructive lung disease were previously identified as major risk factors for severe disease progression in COVID-19. Computed tomography (CT)-based lung-density analysis offers a fast, reliable, and quantitative assessment of lung density. Therefore, we aimed to assess the benefit of CT-based lung density measurements to predict possible severe disease progression in COVID-19. Material and methods Thirty COVID-19-positive patients were included in this retrospective study. Lung density was quantified based on routinely acquired chest CTs. Presence of COVID-19 was confirmed by reverse transcription polymerase chain reaction (RT-PCR). Wilcoxon test was used to compare two groups of patients. A multivariate regression analysis, adjusted for age and sex, was employed to model the relative increase of risk for severe disease, depending on the measured densities. Results Intensive care unit (ICU) patients or patients requiring mechanical ventilation showed a lower proportion of medium- and low-density lung volume compared to patients on the normal ward, but a significantly larger volume of high-density lung volume (12.26 dl IQR 4.65 dl vs. 7.51 dl vs. IQR 5.39 dl, p = 0.039). In multivariate regression analysis, high-density lung volume was identified as a significant predictor of severe disease. Conclusions The amount of high-density lung tissue showed a significant association with severe COVID-19, with odds ratios of 1.42 (95% CI: 1.09-2.00) and 1.37 (95% CI: 1.03-2.11) for requiring intensive care and mechanical ventilation, respectively. Acknowledging our small sample size as an important limitation; our study might thus suggest that high-density lung tissue could serve as a possible predictor of severe COVID-19.

(68)Ga-DOTATOC PET/MRI combines the advantages of PET in the acquisition of metabolic–functional information with the high soft-tissue contrast of MRI. SUVs in tumors have been suggested to be a measure of somatostatin receptor expression. A challenge with receptor ligands is that the distribution volume is confined to tissues with tracer uptake, potentially limiting SUV quantification. In this study, various functional 3-dimensional SUV apparent diffusion coefficient (ADC) parameters and arterial tumor enhancement were tested for ability to characterize gastroenteropancreatic (GEP) neuroendocrine tumors (NETs). Methods: For this single-center, cross-sectional study, 22 patients with 24 histologically confirmed GEP NET lesions (15 men and 7 women; median age, 61 y; range, 43–81 y) who underwent hybrid (68)Ga-DOTA PET/MRI at 3 T between January 2017 and July 2019 met the eligibility criteria. SUV, tumor-to-background ratio, total functional tumor volume, and mean and minimum ADC were measured on the basis of volumes of interest and examined with receiver-operating-characteristic analysis to determine cutoffs for differentiation between low- and intermediate-grade GEP NETs. The Spearman rank correlation coefficient was used to assess correlations between functional imaging parameters. Results: The ratio of PET-derived SUV(mean) and diffusion-weighted imaging–derived minimum ADC was introduced as a combined variable to predict tumor grade, outperforming single predictors. On the basis of a threshold ratio of 0.03, tumors could be classified as grade 2 with a sensitivity of 86% and a specificity of 100%. SUV and functional ADCs, as well as arterial contrast enhancement parameters, showed nonsignificant and mostly negligible correlations. Conclusion: Because receptor density and tumor cellularity appear to be independent, potentially complementary phenomena, the combined ratio of PET/MRI and SUV(mean)/ADC(min) may be used as a novel biomarker allowing differentiation between grade 1 and grade 2 GEP NETs.

Constant interactions between tumor cells and the extracellular matrix (ECM) influence the progression of prostate cancer (PCa). One of the key components of the ECM are collagen fibers, since they are responsible for the tissue stiffness, growth, adhesion, proliferation, migration, invasion/metastasis, cell signaling, and immune recruitment of tumor cells. To explore this molecular marker in the content of PCa, we investigated two different tumor volumes (500 mm3 and 1000 mm3) of a xenograft mouse model of PCa with molecular magnetic resonance imaging (MRI) using a collagen-specific probe. For in vivo MRI evaluation, T1-weighted sequences before and after probe administration were analyzed. No significant signal difference between the two tumor volumes could be found. However, we detected a significant difference between the signal intensity of the peripheral tumor area and the central area of the tumor, at both 500 mm3 (p < 0.01, n = 16) and at 1000 mm3 (p < 0.01, n = 16). The results of our histologic analyses confirmed the in vivo studies: There was no significant difference in the amount of collagen between the two tumor volumes (p > 0.05), but within the tumor, higher collagen expression was observed in the peripheral area compared with the central area of the tumor. Laser ablation with inductively coupled plasma mass spectrometry further confirmed these results. The 1000 mm3 tumors contained 2.8 ± 1.0% collagen and the 500 mm3 tumors contained 3.2 ± 1.2% (n = 16). There was a strong correlation between the in vivo MRI data and the ex vivo histological data (y = −0.068x + 1.1; R2 = 0.74) (n = 16). The results of elemental analysis by inductively coupled plasma mass spectrometry supported the MRI data (y = 3.82x + 0.56; R2 = 0.79; n = 7). MRI with the collagen-specific probe in PCa enables differentiation between different tumor areas. This may help to differentiate tumor from healthy tissue, potentially identifying tumor areas with a specific tumor biology.

Simple Summary One of the most commonly diagnosed cancers in men is prostate cancer (PCa). Understanding tumor progression can help diagnose and treat the disease at an early stage. Components of the extracellular matrix (ECM) play a key role in the development and progression of PCa. Elastin is an essential component of the ECM and constantly changes during tumor development. This article visualizes and quantifies elastin in magnetic resonance imaging (MRI) using a small molecule probe. Results were correlated with histological examinations. Using an elastin-specific molecular probe, we were able to make predictions about the cellular structure in relation to elastin and thus draw conclusions about the size of the tumor, with smaller tumors having a higher elastin content than larger tumors. Abstract Human prostate cancer (PCa) is a type of malignancy and one of the most frequently diagnosed cancers in men. Elastin is an important component of the extracellular matrix and is involved in the structure and organization of prostate tissue. The present study examined prostate cancer in a xenograft mouse model using an elastin-specific molecular probe for magnetic resonance molecular imaging. Two different tumor sizes (500 mm3 and 1000 mm3) were compared and analyzed by MRI in vivo and histologically and analytically ex vivo. The T1-weighted sequence was used in a clinical 3-T scanner to calculate the relative contrast enhancement before and after probe administration. Our results show that the use of an elastin-specific probe enables better discrimination between tumors and surrounding healthy tissue. Furthermore, specific binding of the probe to elastin fibers was confirmed by histological examination and laser ablation–inductively coupled plasma–mass spectrometry (LA-ICP-MS). Smaller tumors showed significantly higher signal intensity (p > 0.001), which correlates with the higher proportion of elastin fibers in the histological evaluation than in larger tumors. A strong correlation was seen between relative enhancement (RE) and Elastica–van Gieson staining (R2 = 0.88). RE was related to inductively coupled plasma–mass spectrometry data for Gd and showed a correlation (R2 = 0.78). Thus, molecular MRI could become a novel quantitative tool for the early evaluation and detection of PCa.

Atherosclerosis is a progressive inflammatory vascular disease characterized by endothelial dysfunction and plaque burden. Extracellular matrix (ECM)-associated plasma proteins play an important role in disease development. Our magnetic resonance imaging (MRI) study investigates the feasibility of using two different molecular MRI probes for the simultaneous assessment of ECM-associated intraplaque albumin deposits caused by endothelial damage and progressive inflammation in atherosclerosis. Male apolipoprotein E-deficient (ApoE-/-)-mice were fed a high-fat diet (HFD) for 2 or 4 months. Another ApoE-/--group was treated with pravastatin and received a HFD for 4 months. T1- and T2*-weighted MRI was performed before and after albumin-specific MRI probe (gadofosveset) administration and a macrophage-specific contrast agent (ferumoxytol). Thereafter, laser ablation inductively coupled plasma mass spectrometry and histology were performed. With advancing atherosclerosis, albumin-based MRI signal enhancement and ferumoxytol-induced signal loss areas in T2*-weighted MRI increased. Significant correlations between contrast-to-noise-ratio (CNR) post-gadofosveset and albumin stain (R2 = 0.78, p &lt, 0.05), and signal loss areas in T2*-weighted MRI with Perls’ Prussian blue stain (R2 = 0.83, p &lt, 0.05) were observed. No interference of ferumoxytol with gadofosveset enhancement was detectable. Pravastatin led to decreased inflammation and intraplaque albumin. Multi-target MRI combining ferumoxytol and gadofosveset is a promising method to improve diagnosis and treatment monitoring in atherosclerosis.

MR relaxometry increasingly contributes to liver imaging. Studies on native relaxation times mainly describe relation to the presence of fibrosis. The hypothesis was that relaxation times are also influenced by other inherent factors, including changes in liver synthesis function. With the approval of the local ethics committee and written informed consent, data from 94 patients referred for liver MR imaging, of which 20 patients had cirrhosis, were included. Additionally to standard sequences, both native T1 and T2 parametric maps and T1 maps in the hepatobiliary phase of gadoxetate disodium were acquired. Associations with laboratory variables were assessed. Altogether, there was a negative correlation between albumin and all acquired relaxation times in cirrhotic patients. In non-cirrhotic patients, only T1 values exhibited a negative correlation with albumin. In all patients, bilirubin correlated significantly with post-contrast T1 relaxation times, whereas native relaxation times correlated only in cirrhotic patients. Evaluating patients with pathological INR values, post-contrast relaxation times were significantly higher, whereas native relaxation times did not correlate. In conclusion, apart from confirming the value of hepatobiliary phase T1 mapping, our results show a correlation of native T1 with serum albumin even in non-cirrhotic liver parenchyma, suggesting a direct influence of liver’s synthesis capacity on T1 relaxation times.

Purpose To evaluate the use of susceptibility-weighted MRI for the differentiation of predominantly osteoblastic and osteolytic spine metastases. Materials and Methods For this prospective study, 53 study participants (mean age, 54.5 years ± 14.3 [range, 22-88 years]; 27 men with a mean age of 55.3 years ± 12.7 [range, 22-72 years] and 26 women with a mean age of 53.8 years ± 15.7 [range, 23-88 years]) with clinically suspected spine metastases underwent imaging with standard MRI sequences, susceptibility-weighted MRI, and CT. Sensitivities and specificities of MRI sequences for the detection of predominantly osteoblastic and osteolytic metastases were determined by using CT as the reference standard. The metastases-to-vertebral body signal intensity ratio (MVR) was calculated to compare modalities. Phantom measurements were obtained to correlate bone densities between MRI sequences and CT. Results A total of 64 metastases (38 predominantly osteoblastic, 26 predominantly osteolytic) were detected. Susceptibility-weighted MRI achieved a sensitivity of 100% (38 of 38) and specificity of 96% (25 of 26) for predominantly osteoblastic metastases and a sensitivity of 96% (25 of 26) and specificity of 100% (38 of 38) for predominantly osteolytic metastases. Standard MRI sequences achieved a sensitivity of 89% (34 of 38) and specificity of 73% (19 of 26) for predominantly osteoblastic metastases and a sensitivity of 73% (19 of 26) and specificity of 92% (35 of 38) for predominantly osteolytic metastases. MVR measurements obtained with susceptibility-weighted MRI demonstrated a strong correlation with those obtained with CT (R2 = 0.75), whereas those obtained with T1-weighted MRI, T2-weighted MRI, and turbo inversion-recovery magnitude MRI showed a weak to moderate correlation (R2 = 0.00, R2 = 0.35, and R2 = 0.39, respectively). Susceptibility-weighted MRI showed a strong correlation with CT with regard to metastases size (R2 = 0.91). In phantom measurements, susceptibility-weighted MRI enabled the reliable differentiation of different degrees of mineralization (R2 = 0.92 compared with CT). Conclusion Susceptibility-weighted MRI enables the reliable differentiation between predominantly osteoblastic and osteolytic spine metastases with a higher accuracy than standard MRI sequences. © RSNA, 2018 Online supplemental material is available for this article. See also the editorial by Schweitzer in this issue.

Objectives The aims of this study were to discriminate among prostate cancers (PCa's) with Gleason scores 6, 7, and ≥8 on biparametric magnetic resonance imaging (bpMRI) of the prostate using radiomics and to evaluate the added value of image augmentation and quantitative T1 mapping. Materials and methods Eighty-five patients with subsequently histologically proven PCa underwent bpMRI at 3 T (T2-weighted imaging, diffusion-weighted imaging) with 66 patients undergoing additional T1 mapping at 3 T. The PCa lesions as well as the peripheral and transition zones were segmented pixel by pixel in multiple slices of the 3D MRI data sets (T2-weighted images, apparent diffusion coefficient, and T1 maps). To increase the size of the data set, images were augmented for contrast, brightness, noise, and perspective multiple times, effectively increasing the sample size 10-fold, and 322 different radiomics features were extracted before and after augmentation. Four different machine learning algorithms, including a random forest (RF), stochastic gradient boosting (SGB), support vector machine (SVM), and k-nearest neighbor, were trained with and without features from T1 maps to differentiate among 3 different Gleason groups (6, 7, and ≥8). Results Support vector machine showed the highest accuracy of 0.92 (95% confidence interval [CI], 0.62-1.00) for classifying the different Gleason scores, followed by RF (0.83; 95% CI, 0.52-0.98), SGB (0.75; 95% CI, 0.43-0.95), and k-nearest neighbor (0.50; 95% CI, 0.21-0.79). Image augmentation resulted in an average increase in accuracy between 0.08 (SGB) and 0.48 (SVM). Removing T1 mapping features led to a decline in accuracy for RF (-0.16) and SGB (-0.25) and a higher generalization error. Conclusions When data are limited, image augmentations and features from quantitative T1 mapping sequences might help to achieve higher accuracy and lower generalization error for classification among different Gleason groups in bpMRI by using radiomics.

An amendment to this paper has been published and can be accessed via a link at the top of the paper.

An amendment to this paper has been published and can be accessed via a link at the top of the paper.

BackgroundWell-informed decisions about how to best treat patients with axial spondyloarthritis (SpA) regularly include an evaluation of the sacroiliac joints (SIJ) on plain radiographs. However, grading radiographic findings correctly has proven to be a considerable challenge to expert readers as well as to state-of-the-art convolutional neural networks (CNNs). A method to reduce image information to the clinically relevant core would undoubtedly lead to more accurate results. We, therefore, trained a CNN only to detect SIJs on radiographs and evaluated its potential as a preprocessing pipeline in the automated classification of SpA.Materials and MethodsWe employed a CNN of the RetinaNet architecture, which was trained on a total of 423 plain radiographs of the sacroiliac joints (SIJs). Images were taken from two completely independent datasets. Training and tuning were performed on image data from the Patients With Axial Spondyloarthritis (PROOF) study and testing was executed using images from the German Spondyloarthritis Inception Cohort (GESPIC). Performance was evaluated by manual review and standard object detection metrics from PASCAL and Microsoft COCO.ResultsThe CNN produced excellent results in detecting SIJs on the tuning (n =106) and on the holdout dataset (n =140). Object detection metrics for the tuning data were AP = 0.996 and mAP = 0.538; values for the independent holdout data were AP = 0.981 and mAP = 0.515. ConclusionsThe developed CNN was highly accurate in detecting SIJs on radiographs. Such a model could increase the reliability of deep learning-based algorithms in detecting and grading SpA.

Segmentation of pulmonary infiltrates can help assess severity of COVID-19, but manual segmentation is labor and time-intensive. Using neural networks to segment pulmonary infiltrates would enable automation of this task. However, training a 3D U-Net from computed tomography (CT) data is time- and resource-intensive. In this work, we therefore developed and tested a solution on how transfer learning can be used to train state-of-the-art segmentation models on limited hardware and in shorter time. We use the recently published RSNA International COVID-19 Open Radiology Database (RICORD) to train a fully three-dimensional U-Net architecture using an 18-layer 3D ResNet, pretrained on the Kinetics-400 dataset as encoder. The generalization of the model was then tested on two openly available datasets of patients with COVID-19, who received chest CTs (Corona Cases and MosMed datasets). Our model performed comparable to previously published 3D U-Net architectures, achieving a mean Dice score of 0.679 on the tuning dataset, 0.648 on the Coronacases dataset and 0.405 on the MosMed dataset. Notably, these results were achieved with shorter training time on a single GPU with less memory available than the GPUs used in previous studies.

Objectives Validation of deep learning models should separately consider bedside chest radiographs (CXRs) as they are the most challenging to interpret, while at the same time the resulting diagnoses are important for managing critically ill patients. Therefore, we aimed to develop and evaluate deep learning models for the identification of clinically relevant abnormalities in bedside CXRs, using reference standards established by computed tomography (CT) and multiple radiologists. Materials and methods In this retrospective study, a dataset consisting of 18,361 bedside CXRs of patients treated at a level 1 medical center between January 2009 and March 2019 was used. All included CXRs occurred within 24 hours before or after a chest CT. A deep learning algorithm was developed to identify 8 findings on bedside CXRs (cardiac congestion, pleural effusion, air-space opacification, pneumothorax, central venous catheter, thoracic drain, gastric tube, and tracheal tube/cannula). For the training dataset, 17,275 combined labels were extracted from the CXR and CT reports by a deep learning natural language processing (NLP) tool. In case of a disagreement between CXR and CT, human-in-the-loop annotations were used. The test dataset consisted of 583 images, evaluated by 4 radiologists. Performance was assessed by area under the receiver operating characteristic curve analysis, sensitivity, specificity, and positive predictive value. Results Areas under the receiver operating characteristic curve for cardiac congestion, pleural effusion, air-space opacification, pneumothorax, central venous catheter, thoracic drain, gastric tube, and tracheal tube/cannula were 0.90 (95% confidence interval [CI], 0.87-0.93; 3 radiologists on the receiver operating characteristic [ROC] curve), 0.95 (95% CI, 0.93-0.96; 3 radiologists on the ROC curve), 0.85 (95% CI, 0.82-0.89; 1 radiologist on the ROC curve), 0.92 (95% CI, 0.89-0.95; 1 radiologist on the ROC curve), 0.99 (95% CI, 0.98-0.99), 0.99 (95% CI, 0.98-0.99), 0.98 (95% CI, 0.97-0.99), and 0.99 (95% CI, 0.98-1.00), respectively. Conclusions A deep learning model used specifically for bedside CXRs showed similar performance to expert radiologists. It could therefore be used to detect clinically relevant findings during after-hours and help emergency and intensive care physicians to focus on patient care.

Background. Currently, there is no reliable nonsurgical treatment for abdominal aortic aneurysm (AAA). This study, therefore, investigates if doxycycline reduces AAA growth and the number of rupture-related deaths in a murine ApoE-/- model of AAA and whether gadofosveset trisodium-based MRI differs between animals with and without doxycycline treatment. Methods. Nine ApoE-/- mice were implanted with osmotic minipumps continuously releasing angiotensin II and treated with doxycycline (30 mg/kg/d) in parallel. After four weeks, MRI was performed at 3T with a clinical dose of the albumin-binding probe gadofosveset (0.03 mmol/kg). Results were compared with previously published wild-type control animals and with previously studied ApoE-/- animals without doxycycline treatment. Differences in mortality were also investigated between these groups. Results. In a previous study, we found that approximately 25% of angiotensin II-infused ApoE-/- mice died, whereas in the present study, only one out of 9 angiotensin II-infused and doxycycline-treated ApoE-/- mice (11.1%) died within 4 weeks. Furthermore, doxycycline-treated ApoE-/- mice showed significantly lower contrast-to-noise (CNR) values (p = 0.017) in MRI compared to ApoE-/- mice without doxycycline treatment. In vivo measurements of relative signal enhancement (CNR) correlated significantly with ex vivo measurements of albumin staining (R-2 = 0.58). In addition, a strong visual colocalization of albumin-positive areas in the fluorescence albumin staining with gadolinium distribution in LA-ICP-MS was shown. However, no significant difference in aneurysm size was observed after doxycycline treatment. Conclusion. The present experimental in vivo study suggests that doxycycline treatment may reduce rupture-related deaths in AAA by slowing endothelial damage without reversing aneurysm growth.

Background Minimally invasive, battery-powered drilling systems have become the preferred tool for obtaining representative samples from bone lesions. However, the heat generated during battery-powered bone drilling for bone biopsies has not yet been sufficiently investigated. Thermal necrosis can occur if the bone temperature exceeds a critical threshold for a certain period of time. Purpose To investigate heat production as a function of femur temperature during and after battery-powered percutaneous bone drilling in a porcine in vivo model. Methods We performed 16 femur drillings in 13 domestic pigs with an average age of 22 weeks and an average body temperature of 39.7 °C, using a battery-powered drilling system and an intraosseous temperature monitoring device. The standardized duration of the drilling procedure was 20 s. The bone core specimens obtained were embedded in 4% formalin, stained with haematoxylin and eosin (H&E) and sent for pathological analysis of tissue quality and signs of thermal damage. Results No significant changes in the pigs’ local temperature were observed after bone drilling with a battery-powered drill device. Across all measurements, the median change in temperature between the initial measurement and the temperature measured after drilling (at 20 s) was 0.1 °C. Histological examination of the bone core specimens revealed no signs of mechanical or thermal damage. Conclusion Overall, this preliminary study shows that battery-powered, drill-assisted harvesting of bone core specimens does not appear to cause mechanical or thermal damage.

Objective Training a convolutional neural network (CNN) to detect the most common causes of shoulder pain on plain radiographs and to assess its potential value in serving as an assistive device to physicians. Materials and methods We used a CNN of the ResNet-50 architecture which was trained on 2700 shoulder radiographs from clinical practice of multiple institutions. All radiographs were reviewed and labeled for six findings: proximal humeral fractures, joint dislocation, periarticular calcification, osteoarthritis, osteosynthesis, and joint endoprosthesis. The trained model was then evaluated on a separate test dataset, which was previously annotated by three independent expert radiologists. Both the training and the test datasets included radiographs of highly variable image quality to reflect the clinical situation and to foster robustness of the CNN. Performance of the model was evaluated using receiver operating characteristic (ROC) curves, the thereof derived AUC as well as sensitivity and specificity. Results The developed CNN demonstrated a high accuracy with an area under the curve (AUC) of 0.871 for detecting fractures, 0.896 for joint dislocation, 0.945 for osteoarthritis, and 0.800 for periarticular calcifications. It also detected osteosynthesis and endoprosthesis with near perfect accuracy (AUC 0.998 and 1.0, respectively). Sensitivity and specificity were 0.75 and 0.86 for fractures, 0.95 and 0.65 for joint dislocation, 0.90 and 0.86 for osteoarthrosis, and 0.60 and 0.89 for calcification. Conclusion CNNs have the potential to serve as an assistive device by providing clinicians a means to prioritize worklists or providing additional safety in situations of increased workload.

Adjacent to hepatic metastases, liver parenchyma is often histopathologically altered even if its visual appearance on native magnetic resonance (MR) images is blunt. Yet, relaxation properties in MR imaging may show structural changes prior to visual alteration, and therefore, the aim of this study was to investigate whether T1 relaxation times in the perilesional zone differ between metastases and benign lesions. A total of 113 patients referred for MRI were included prospectively. Images were assessed for metastases, solid benign lesions and cysts, and regions-of-interest were drawn on T1 maps including the focal lesion and a close (inner perilesional zone = IPZ) and a larger perilesional zone (outer perilesional zone = OPZ). Simple ratios between these zones, as well as a gradient ratio between the IPZ and the entire perilesional zone (EPZ) were calculated. Within the collective, 44 patients had lesions of one or two entities. For metastases, the simple ratio between IPZ and OPZ as well as the mean EPZ gradient was significantly higher than for both solid benign lesions and cysts. Lesion size was not a significant covariate. We conclude, that native T1 properties of the perilesional zones differ significantly between malignant and both solid and cystic benign lesions.

Objectives To develop and validate an artificial neural network for the detection of definite radiographic sacroiliitis as a manifestation of axial spondyloarthritis. Methods Conventional radiographs of sacroiliac joints from two independent cohorts of patients with axial spondyloarthritis (axSpA) were used. The first cohort consisted of 1669 radiographs and was used for training and validation of a neural network. The second cohort consisted of 525 radiographs, of which 100 radiographs were randomly selected for the test dataset. In both cohorts all radiographs underwent central reading; the final decision on the presence or absence of definite radiographic sacroiliitis was used as a reference. For performance evaluation of the neural network, areas under the receiver operating characteristic curves (AUROC) were calculated. Sensitivity and specificity for the prediction cut-offs were calculated. Cohen’s Kappa and the absolute agreement were used to assess the agreement between the neural network and the human readers. Results The neural network achieved an excellent performance in recognition of definite radiographic sacroiliitis with AUROC of 0.97 and 0.96 for the validation and test datasets, respectively. Sensitivity and specificity for the cut-off weighting both measurements equally were 0.90 and 0.93 for the validation and 0.87 and 0.97 for the test set. The Cohen’s kappa between the neural network and the reference judgements were 0.80 for both validation and test sets, and the absolute agreement on the classification yielded 91% and 90%, respectively. Conclusions Artificial neural networks enable the accurate detection of definite radiographic sacroiliitis relevant for the diagnosis and classification of axSpA.

INTRODUCTION: During the unprecedented health crisis of the COVID-19 pandemic it was suggested that obesity might aggravate severe acute respiratory syndrome coronavirus-2 (SARS CoV-2). Therefore, this study aims to investigate the association between Compute Tomography (CT)-based measurements of visceral and subcutaneous fat as measures of obesity and COVID-19 severity. METHODS: 30 patients with laboratory-confirmed COVID-19 and a mean age of 65.59 plus/minus 13.06 years from a level one medical center in Berlin, Germany, were retrospectively analyzed and included in the present analysis. SARS-CoV-2 was confirmed by polymerase chain reaction from throat swaps or deep nasal swabs on the day of admission. Severe clinical courses of COVID-19 were defined by hospitalization in intensive care unit (ICU) and invasive mechanical ventilation. All patients received low-dose chest CT-based fat measurements at the level of the first lumbar vertebra. RESULTS: An increase in visceral fat area (VFA) by one square decimeter was associated with a 22.53-fold increased risk for ICU treatment and a 16.11-fold increased risk for mechanical ventilation (adjusted for age and sex). For upper abdominal circumference, each additional centimeter of circumference showed a 1.13-fold increased risk for ICU treatment and a 1.25-fold increased risk for mechanical ventilation. There was no significant correlation of subcutaneous fat area (SFA) or body mass index (BMI) with severe clinical courses of COVID-19. CONCLUSIONS: Our results suggest that visceral adipose tissue and upper abdominal circumference specifically increasing the risk of COVID-19 severity. CT-based quantification of visceral adipose tissue and upper abdominal circumference in routinely acquired chest CTs may therefore be a simple tool for risk assessment in SARS-CoV-2-patients.

Chest radiographs are among the most frequently acquired images in radiology and are often the subject of computer vision research. However, most of the models used to classify chest radiographs are derived from openly available deep neural networks, trained on large image datasets. These datasets differ from chest radiographs in that they are mostly color images and have substantially more labels. Therefore, very deep convolutional neural networks (CNN) designed for ImageNet and often representing more complex relationships, might not be required for the comparably simpler task of classifying medical image data. Sixteen different architectures of CNN were compared regarding the classification performance on two openly available datasets, the CheXpert and COVID-19 Image Data Collection. Areas under the receiver operating characteristics curves (AUROC) between 0.83 and 0.89 could be achieved on the CheXpert dataset. On the COVID-19 Image Data Collection, all models showed an excellent ability to detect COVID-19 and non-COVID pneumonia with AUROC values between 0.983 and 0.998. It could be observed, that more shallow networks may achieve results comparable to their deeper and more complex counterparts with shorter training times, enabling classification performances on medical image data close to the state-of-the-art methods even when using limited hardware.

Objectives The aim of this study was to improve the visualization of coagulation necrosis after computed tomography (CT)-guided microwave ablation (MWA) in routine postablational imaging. Materials and methods Ten MWAs were performed in 8 pigs under CT guidance. After each ablation, we obtained contrast-enhanced CT scans in venous phase. Ablations were then resected as a whole, and histologic slices were obtained orthogonally through the ablation center. Subsequently, a vital stain was applied to the sections for visualization of coagulation necrosis. Computed tomography images were reformatted to match the histologic slices. Afterwards, quantitative imaging features were extracted from the subregions of all images, and binary classifiers were used to predict the presence of coagulation necrosis for each subregion. From this, heatmaps could be created, which visually represented the extent of necrosis in each CT image. Two independent observers evaluated the extent of coagulative necrosis between the heat maps and histological sections. Results We applied 4 different classifiers, including a generalized linear mixed model (GLMM), a stochastic gradient boosting classifier, a random forest classifier, and a k-nearest neighbor classifier, out of which the GLMM showed the best performance to display coagulation necrosis. The GLMM resulted in an area under the curve of 0.84 and a Jaccard index of 0.6 between the generated heat map and the histologic reference standard as well as a good interobserver agreement with a Jaccard index of 0.9. Conclusions Subregion radiomics analysis may improve visualization of coagulation necrosis after hepatic MWA in an in vivo porcine model.

Abdominal aortic aneurysm (AAA) is a life-threatening vascular disease with an up to 80% mortality in case of rupture. Current biomarkers fail to account for size-independent risk of rupture. By combining the information of different molecular probes, multi-target molecular MRI holds the potential to enable individual characterization of AAA. In this experimental study, we aimed to examine the feasibility of simultaneous imaging of extracellular collagen and inflammation for size-independent prediction of risk of rupture in murine AAA. The study design consisted of: (1) A outcome-based longitudinal study with imaging performed once after one week with follow-up and death as the end-point for assessment of rupture risk. (2) A week-by-week study for the characterization of AAA development with imaging after 1, 2, 3 and 4 weeks. For both studies, the animals were administered a type 1 collagen-targeted gadolinium-based probe (surrogate marker for extracellular matrix (ECM) remodeling) and an iron oxide-based probe (surrogate marker for inflammatory activity), in one imaging session. In vivo measurements of collagen and iron oxide probes showed a significant correlation with ex vivo histology (p