Deep neural networks (DNNs) have achieved state-of-the-art performance in many important domains, including medical diagnosis, security, and autonomous driving. In domains where safety is highly ...critical, an erroneous decision can result in serious consequences. While a perfect prediction accuracy is not always achievable, recent work on Bayesian deep networks shows that it is possible to know when DNNs are more likely to make mistakes. Knowing what DNNs do not know is desirable to increase the safety of deep learning technology in sensitive applications; Bayesian neural networks attempt to address this challenge. Traditional approaches are computationally intractable and do not scale well to large, complex neural network architectures. In this paper, we develop a theoretical framework to approximate Bayesian inference for DNNs by imposing a Bernoulli distribution on the model weights. This method called Monte Carlo DropConnect (MC-DropConnect) gives us a tool to represent the model uncertainty with little change in the overall model structure or computational cost. We extensively validate the proposed algorithm on multiple network architectures and datasets for classification and semantic segmentation tasks. We also propose new metrics to quantify uncertainty estimates. This enables an objective comparison between MC-DropConnect and prior approaches. Our empirical results demonstrate that the proposed framework yields significant improvement in both prediction accuracy and uncertainty estimation quality compared to the state of the art.
The coronavirus disease 2019 (COVID-19) pandemic is a global health care emergency. Although reverse-transcription polymerase chain reaction testing is the reference standard method to identify ...patients with COVID-19 infection, chest radiography and CT play a vital role in the detection and management of these patients. Prediction models for COVID-19 imaging are rapidly being developed to support medical decision making. However, inadequate availability of a diverse annotated data set has limited the performance and generalizability of existing models. To address this unmet need, the RSNA and Society of Thoracic Radiology collaborated to develop the RSNA International COVID-19 Open Radiology Database (RICORD). This database is the first multi-institutional, multinational, expert-annotated COVID-19 imaging data set. It is made freely available to the machine learning community as a research and educational resource for COVID-19 chest imaging. Pixel-level volumetric segmentation with clinical annotations was performed by thoracic radiology subspecialists for all COVID-19-positive thoracic CT scans. The labeling schema was coordinated with other international consensus panels and COVID-19 data annotation efforts, the European Society of Medical Imaging Informatics, the American College of Radiology, and the American Association of Physicists in Medicine. Study-level COVID-19 classification labels for chest radiographs were annotated by three radiologists, with majority vote adjudication by board-certified radiologists. RICORD consists of 240 thoracic CT scans and 1000 chest radiographs contributed from four international sites. It is anticipated that RICORD will ideally lead to prediction models that can demonstrate sustained performance across populations and health care systems.
Structured abstract Background The National Lung Screening Trial (NLST) showed low-dose screening chest CT reduced lung cancer mortality rate up to 20% in high risk patients in the United States. We ...aim to investigate the impact of applying the NLST eligibility criteria to the population in Taiwan, and to identify additional risk factors to select subjects at risk of lung cancer. Patients and Methods We retrospectively review the medical record of 1763 asymptomatic healthy subjects (40∼80 year old) who voluntarily underwent low-dose chest CT (1029 male, 734 female) from August 2013 to August 2014. Clinical information and nodule characteristics were recorded. Results of subsequent follow-up and outcome were also recorded. Results 8.4% (148/1763) of subjects would have been eligible for lung cancer screening based on the NLST criteria. However, only one of these eligible subjects would have a lung cancer detected at baseline. Among the 1615 subjects who did not meet the NLST criteria, the detection rates of lung cancer was 2.6% in women and 0.56% in men. Logistic regression showed that female gender and a family history of lung cancer were the two most important predictor of lung cancer in Taiwan (odds ratio of 6.367, P value = 0.003; odds ratio of 3.017, P value = 0.016, respectively). Conclusions In conclusion, NLST eligibility criteria may not be effective in screening lung cancer in Taiwan. Risk-based prediction model based on the family history of lung cancer and female gender can potentially improve efficiency of lung cancer screening programs in Taiwan.
The epidemiology of lung cancer de Groot, Patricia M.; Wu, Carol C.; Carter, Brett W. ...
Translational lung cancer research,
6/2018, Letnik:
7, Številka:
3
Journal Article
Odprti dostop
The incidence and mortality from lung cancer is decreasing in the US due to decades of public education and tobacco control policies, but are increasing elsewhere in the world related to the ...commencement of the tobacco epidemic in various countries and populations in the developing world. Individual cigarette smoking is by far the most common risk factor for lung carcinoma; other risks include passive smoke inhalation, residential radon, occupational exposures, infection and genetic susceptibility. The predominant disease burden currently falls on minority populations and socioeconomically disadvantaged people. In the US, the recent legalization of marijuana for recreational use in many states and the rapid growth of commercially available electronic nicotine delivery systems (ENDS) present challenges to public health for which little short term and no long term safety data is available.
The incidental thyroid nodule (ITN) is one of the most common incidental findings on imaging studies that include the neck. An ITN is defined as a nodule not previously detected or suspected ...clinically, but identified by an imaging study. The workup of ITNs has led to increased costs from additional procedures, and in some cases, to increased risk to the patient because physicians are naturally concerned about the risk of malignancy and a delayed cancer diagnosis. However, the majority of ITNs are benign, and small, incidental thyroid malignancies typically have indolent behavior. The ACR formed the Incidental Thyroid Findings Committee to derive a practical approach to managing ITNs on CT, MRI, nuclear medicine, and ultrasound studies. This white paper describes consensus recommendations representing this committee's review of the literature and their practice experience.
As a pandemic, a most-common pattern resembled organizing pneumonia (OP) has been identified by CT findings in novel coronavirus disease (COVID-19). We aimed to delineate the evolution of CT findings ...and outcome in OP of COVID-19.
106 COVID-19 patients with OP based on CT findings were retrospectively included and categorized into non-severe (mild/common) and severe (severe/critical) groups. CT features including lobar distribution, presence of ground glass opacities (GGO), consolidation, linear opacities and total severity CT score were evaluated at three time intervals from symptom-onset to CT scan (day 0-7, day 8-14, day > 14). Discharge or adverse outcome (admission to ICU or death), and pulmonary sequelae (complete absorption or lesion residuals) on CT after discharge were analyzed based on the CT features at different time interval.
79 (74.5%) patients were non-severe and 103 (97.2%) were discharged at median day 25 (range, day 8-50) after symptom-onset. Of 67 patients with revisit CT at 2-4 weeks after discharge, 20 (29.9%) had complete absorption of lesions at median day 38 (range, day 30-53) after symptom-onset. Significant differences between complete absorption and residuals groups were found in percentages of consolidation (1.5% vs. 13.8%, P = 0.010), number of involved lobe > 3 (40.0% vs. 72.5%, P = 0.030), CT score > 4 (20.0% vs. 65.0%, P = 0.010) at day 8-14.
Most OP cases had good prognosis. Approximately one-third of cases had complete absorption of lesions during 1-2 months after symptom-onset while those with increased frequency of consolidation, number of involved lobe > 3, and CT score > 4 at week 2 after symptom-onset may indicate lesion residuals on CT.
The objective of this article is to describe potential complications of percutaneous needle biopsy of the chest, discuss the risk factors associated with the development of complications, and explain ...how to prevent complications and manage complications when they occur.
Pneumothorax and pulmonary hemorrhage are the most common complications of percutaneous needle biopsy of the chest, whereas air embolism and tumor seeding are extremely rare. Attention to biopsy planning and technique and postprocedural care help to prevent or minimize most potential complications.
Purpose
We use central frequency ratio and noise magnitude ratio from noise power spectrum (NPS) to evaluate the noise reduction techniques of ASiR and ASiR‐V of GE, SAFIRE and ADMIRE of Siemens, and ...PixelShine of AlgoMedica. ASiR, ASiR‐V, SAFIRE and ADMIRE use a combination of image and projection data whereas PixelShine uses artificial intelligence neural network for noise reduction.
Methods and materials
The homogeneous module of the ACR computed tomography (CT) phantom was scanned on a GE Revolution HD 64‐slice CT for ASiR and ASiR‐V, a Siemens Somatom Force for ADMIRE, and a Siemens Definition AS 64‐slice for SAFIRE for NPS calculation. The baseline filtered back‐projection (FBP) reconstructions were derived from the standard kernel on Revolution HD, Hr44f on Force and D40s on Definition AS. The central frequency ratio (CFR) indicates the degree of shift in the central frequency of NPS after noise reduction. A smaller CFR means a larger shift of the NPS curve, or a larger degree of image blurring. The noise magnitude ratio (NMR) indicates the amount of noise removed. A smaller NMR means a larger degree of noise reduction. An ideal noise reduction shall maintain a CFR close to 1 and a NMR close to 0.
Results
The ideal noise reduction by increasing radiation exposure did not shift the central frequency when the image noise was reduced. PixelShine was the closest to the ideal noise reduction in CFR, and was followed by SAFIRE, ASiR‐V, ADMIRE and ASiR, in sequence. Similarly, PixelShine had the smallest NMR, and was followed by SAFIRE, ASiR‐V, ADMIRE and ASiR in sequence. Overall, PixelShine had the least central frequency shift for the same amount of noise reduction or the most noise reduction for the same amount of central frequency shift. For the same CFR, ASiR‐V reduced more noise than ASiR; and SAFIRE reduced more noise than ADMIRE.
Conclusions
We introduced two new parameters of CFR and NMR from NPS to compare the reconstructions from different manufacturers. PixelShine had the least central frequency shift for the same amount of noise reduction or the most noise reduction for the same amount of central frequency shift. For the same central frequency shift, ASiR‐V reduced more noise than ASiR, and SAFIRE reduced more noise than ADMIRE.
The purpose of this study was to analyze the clinical and CT features of pulmonary artery pseudoaneurysms (PAPs).
A database search of chest CT examinations performed from January 1, 2000 to December ...31, 2014 identified 24 patients with findings consistent with PAPs. A CT finding consistent with a PAP was defined as a focal saccular outpouching of a pulmonary artery. Medical records were reviewed to determine clinical presentations, treatments, and outcomes. CT scans were reviewed by two board-certified fellowship-trained chest radiologists.
A total of 35 PAPs were identified in 24 patients. Hemoptysis and shortness of breath were the most common presenting symptoms. The most commonly identified causes of PAPs were infection (33%), neoplasms (13%), and trauma (17%). Of the 35 PAPs, 29 (83%) were located in segmental or subsegmental pulmonary arteries. A solitary PAP was identified in 20 (83%) patients, and multiple PAPs were identified in three patients with endocarditis and one patient with pulmonary metastases. Only three of 35 (9%) PAPs were associated with a ground-glass halo. Endovascular treatment was successfully performed in 12 patients, and only one patient had immediate recurrent hemoptysis after treatment. PAP was clinically suspected by the referring clinicians in only three patients. Sixteen of the 35 (46%) PAPs were not reported on the initial CT studies.
PAPs showed a strong predilection for the peripheral pulmonary arteries. Multiplicity of PAPs can be seen in the settings of endocarditis and pulmonary metastatic disease. Most PAPs were not associated with a ground-glass halo. PAPs can be lethal but were often not suspected clinically and were underreported by radiologists.
During bone remodelling, osteoclasts induce chemotaxis of osteoblasts and yet maintain spatial segregation. We show that osteoclasts express the repulsive guidance factor Semaphorin 4D and induce ...contact inhibition of locomotion (CIL) in osteoblasts through its receptor Plexin-B1. To examine causality and elucidate how localized Plexin-B1 stimulation may spatiotemporally coordinate its downstream targets in guiding cell migration, we develop an optogenetic tool for Plexin-B1 designated optoPlexin. Precise optoPlexin activation at the leading edge of migrating osteoblasts readily induces local retraction and, unexpectedly, distal protrusions to steer cells away. These morphological changes are accompanied by reorganization of Myosin II, PIP
, adhesion and active Cdc42. We attribute the resultant repolarization to RhoA/ROCK-mediated redistribution of β-Pix, which activates Cdc42 and promotes protrusion. Thus, our data demonstrate a causal role of Plexin-B1 for CIL in osteoblasts and reveals a previously unknown effect of Semaphorin signalling on spatial distribution of an activator of cell migration.
PDF
