Search

Rationale and Objectives High-resolution computed tomography (HRCT) is paramount in the assessment of interstitial lung disease (ILD). Yet, HRCT interpretation of ILDs may be hampered by inter- and intra-observer variability. Recently, artificial intelligence (AI) has revolutionized medical image analysis. This technology has the potential to advance patient care in ILD. We aimed to systematically evaluate the application of AI for the analysis of ILD in HRCT. Materials and Methods We searched MEDLINE/PubMed databases for original publications of deep learning for ILD analysis on chest CT. The search included studies published up to March 1, 2021. The risk of bias evaluation included tailored Quality Assessment of Diagnostic Accuracy Studies and the modified Joanna Briggs Institute Critical Appraisal checklist. Results Data was extracted from 19 retrospective studies. Deep learning techniques included detection, segmentation, and classification of ILD on HRCT. Most studies focused on the classification of ILD into different morphological patterns. Accuracies of 78%-91% were achieved. Two studies demonstrated near-expert performance for the diagnosis of idiopathic pulmonary fibrosis (IPF). The Quality Assessment of Diagnostic Accuracy Studies tool identified a high risk of bias in 15/19 (78.9%) of the studies. Conclusion AI has the potential to contribute to the radiologic diagnosis and classification of ILD. However, the accuracy performance is still not satisfactory, and research is limited by a small number of retrospective studies. Hence, the existing published data may not be sufficiently reliable. Only well-designed prospective controlled studies can accurately assess the value of existing AI tools for ILD evaluation.

What is the effect of prior experience on sensorimotor behavior? We studied the following intriguing behavior: monkeys fixating a small target direct their gaze above the target if the background is dark. Fixating a target once on a bright background, then on a dark background, yields 2 gaze directions, typically one above the other; hence the name, ‘dark-background-contingent upshift of gaze', which is abbreviated as ‘upshift’. Is the upshift only an attempt to avoid using the fovea in the dark? If it is, we might expect to also observe a downshift and a sideshift. We studied gaze direction in a large group of 10 rhesus monkeys from Tubingen, to which we added published data from 4 cynomolgus monkeys from Rehovot. The upshift was ubiquitous, and there was no systematic sideshift. What is the function of the upshift? Is it related to vision in the dark? Here, we concentrate on the effect of the monkeys’ previous training. Seven of the 14 monkeys were accustomed to working in the dark (‘dark-habituated’), while the other 7 had worked in bright ambient light (‘bright-habituated’). The main result of this study is that the dark-habituated monkeys had a much larger upshift: the mean upshift was 2.2° in the dark-habituated monkeys, versus 0.5° in the bright-habituated. Thus, the upshift depends on habituation; the size of the upshift reflects months-long cumulative experience. These findings suggest that the function of the upshift is indeed related to seeing in the dark.

BACKGROUND: Right ventricular failure is an underrecognized consequence of COVID-19 pneumonia. Those with severe disease are treated with extracorporeal membrane oxygenation (ECMO) but with poor outcomes. Concomitant right ventricular assist device (RVAD) may be beneficial. METHODS: A retrospective analysis of intensive care unit patients admitted with COVID-19 ARDS (Acute Respiratory Distress Syndrome) was performed. Nonintubated patients, those with acute kidney injury, and age > 75 were excluded. Patients who underwent RVAD/ECMO support were compared with those managed via invasive mechanical ventilation (IMV) alone. The primary outcome was in-hospital mortality. Secondary outcomes included 30-d mortality, acute kidney injury, length of ICU stay, and duration of mechanical ventilation. RESULTS: A total of 145 patients were admitted to the ICU with COVID-19. Thirty-nine patients met inclusion criteria. Of these, 21 received IMV, and 18 received RVAD/ECMO. In-hospital (52.4 versus 11.1%, P = 0.008) and 30-d mortality (42.9 versus 5.6%, P= 0.011) were significantly lower in patients treated with RVAD/ECMO. Acute kidney injury occurred in 15 (71.4%) patients in the IMV group and zero RVAD/ECMO patients (P< 0.001). ICU (11.5 versus 21 d, P= 0.067) and hospital (14 versus 25.5 d, P = 0.054) length of stay were not significantly different. There were no RVAD/ECMO device complications. The duration of mechanical ventilation was not significantly different (10 versus 5 d, P = 0.44). CONCLUSIONS: RVAD support at the time of ECMO initiation resulted in the no secondary end-organ damage and higher in-hospital and 30-d survival versus IMV in specially selected patients with severe COVID-19 ARDS. Management of severe COVID-19 ARDS should prioritize right ventricular support.

Purpose To test the primary hypothesis that a CPR video will reduce ICU patients' surrogates' anxiety when deciding code status, as measured by the Hamilton Anxiety Rating (HAM-A) Scale, as compared to the no video group. Materials and methods This is a prospective randomized control trial. Twenty-seven ICU patients' surrogates were enrolled in the study after receiving an ICU team-led code status discussion. After the enrollment, twelve surrogates were randomized to the video group and fifteen to the no video group. The primary outcome of anxiety was quantified using the HAM-A Scale. Demographic information, clinical data, and patients' provenance information (Home vs. Not Home) were collected. The patients' severity of illness was calculated using the Sequential Organ Failure Assessment (SOFA) Score. Results The HAM-A score in the video group was 5.65 points lower than in the no video group ([β = −5.65, 95% CI −11.12 −0.18] P = 0.04). The statistically significant difference was maintained when adjusting for patients' SOFA Score and patients' provenance (P = 0.03). Conclusion CPR video used to supplement ICU team-led code status discussions reduced surrogates' anxiety, as compared to no video. Trial Registration: ClinicalTrials.gov Identifier NCT03630965 .

Background and Aims Capsule endoscopy (CE) is an important modality for Crohn's disease (CD) diagnosis and follow-up. Endoscopic severity of ulcers is of much significance for prediction of CD course. Deep learning has been proven accurate in detecting ulcers on CE. However, endoscopic classification of ulcers by deep learning has not been attempted. The aim of our study was to develop a deep learning algorithm for the automated grading of CD ulcers on CE. Methods We retrospectively collected CE images of CD ulcers from our CE database. In experiment 1, each ulcer severity was graded by 2 capsule readers based on PillCam CD classification (grade 1-3 from mild to severe), and the inter-reader variability was evaluated. In experiment 2, a consensus reading by 3 capsule readers was used to train an ordinal convolutional neural network (CNN) to automatically grade images of ulcers, and the resulting algorithm was tested against consensus reading. A pretraining stage included training the network on images of normal mucosa and ulcerated mucosa. Results Overall, our dataset included 17,640 CE images from 49 patients; 7391 images with mucosal ulcers and 10,249 normal images. A total of 2598 randomly picked pathological images were further graded from 1 to 3 according to the ulcer severity in the 2 different experiments. In experiment 1, the overall inter-reader agreement occurred in 31% of the images (345/1108). When, and for distinction of grade 1 and 3 - 76% (752/989). In experiment 2, the algorithm was trained on 1242 images. It achieved an overall agreement to consensus reading of 67% (166/248), and 91% (158/173) for distinction of grade 1 and 3. When testing for classification accuracy, the algorithm performance was 0.91 (95% CI, 0.867 - 0.954) for grade 1 versus grade 3 ulcers, 0.78 (95% CI, 0.716-0.844) for 2 versus 3 and 0.624 (95% CI, 0.547-0.701) for 1 versus 2. Conclusions CNN achieved high accuracy in detection of severe CD ulcerations. CNN-assisted CE reading in CD patients can potentially facilitate and improve diagnosis and monitoring these patients.

Introduction Kidney and liver cysts in autosomal dominant polycystic kidney disease (ADPKD) can compress the inferior vena cava (IVC), but IVC compression prevalence and its risk factors are unknown. Methods Patients who have ADPKD (n = 216) with abdominal magnetic resonance imaging (MRI) studies and age-/sex-matched controls (n = 216) were evaluated for IVC compression as well as azygous vein diameter (a marker of collateral blood flow) and IVC aspect ratio (left-to-right dimension divided by anterior-to-posterior dimension with a value of 1 corresponding to a circular (high pressure) IVC caudal to compression. Results Severe IVC compression (≥70%) was observed in 33 (15%) ADPKD subjects and mild compression (≥50% to 40 years old, with Mayo Clinic class 1D/E, and in females with ht-LV > 2000 ml/m., Graphical abstract

Rationale and Objectives Generative adversarial networks (GANs) are deep learning models aimed at generating fake realistic looking images. These novel models made a great impact on the computer vision field. Our study aims to review the literature on GANs applications in radiology. Materials and Methods This systematic review followed the PRISMA guidelines. Electronic datasets were searched for studies describing applications of GANs in radiology. We included studies published up-to September 2019. Results Data were extracted from 33 studies published between 2017 and 2019. Eighteen studies focused on CT images generation, ten on MRI, three on PET/MRI and PET/CT, one on ultrasound and one on X-ray. Applications in radiology included image reconstruction and denoising for dose and scan time reduction (fourteen studies), data augmentation (six studies), transfer between modalities (eight studies) and image segmentation (five studies). All studies reported that generated images improved the performance of the developed algorithms. Conclusion GANs are increasingly studied for various radiology applications. They enable the creation of new data, which can be used to improve clinical care, education and research.

Purpose Natural language processing (NLP) enables conversion of free text into structured data. Recent innovations in deep learning technology provide improved NLP performance. We aimed to survey deep learning NLP fundamentals and review radiology-related research. Methods This systematic review was reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines. We searched for deep learning NLP radiology studies published up to September 2019. MEDLINE, Scopus, and Google Scholar were used as search databases. Results Ten relevant studies published between 2018 and 2019 were identified. Deep learning models applied for NLP in radiology are convolutional neural networks, recurrent neural networks, long short-term memory networks, and attention networks. Deep learning NLP applications in radiology include flagging of diagnoses such as pulmonary embolisms and fractures, labeling follow-up recommendations, and automatic selection of imaging protocols. Deep learning NLP models perform as well as or better than traditional NLP models. Conclusion Research and use of deep learning NLP in radiology is increasing. Acquaintance with this technology can help prepare radiologists for the coming changes in their field.