scholarly journals B-Line Detection and Localization in Lung Ultrasound Videos Using Spatiotemporal Attention

2021 ◽  
Vol 11 (24) ◽  
pp. 11697
Author(s):  
Hamideh Kerdegari ◽  
Nhat Tran Huy Phung ◽  
Angela McBride ◽  
Luigi Pisani ◽  
Hao Van Nguyen ◽  
...  
Keyword(s):  
Detection Rate ◽  
Lung Ultrasound ◽  
Saliency Map ◽  
Severe Dengue ◽  
Line Detection ◽  
Resource Limited ◽  
Lung Abnormalities ◽  
Weakly Supervised ◽  
Detection And Localization ◽  
B Lines

The presence of B-line artefacts, the main artefact reflecting lung abnormalities in dengue patients, is often assessed using lung ultrasound (LUS) imaging. Inspired by human visual attention that enables us to process videos efficiently by paying attention to where and when it is required, we propose a spatiotemporal attention mechanism for B-line detection in LUS videos. The spatial attention allows the model to focus on the most task relevant parts of the image by learning a saliency map. The temporal attention generates an attention score for each attended frame to identify the most relevant frames from an input video. Our model not only identifies videos where B-lines show, but also localizes, within those videos, B-line related features both spatially and temporally, despite being trained in a weakly-supervised manner. We evaluate our approach on a LUS video dataset collected from severe dengue patients in a resource-limited hospital, assessing the B-line detection rate and the model’s ability to localize discriminative B-line regions spatially and B-line frames temporally. Experimental results demonstrate the efficacy of our approach for classifying B-line videos with an F1 score of up to 83.2% and localizing the most salient B-line regions both spatially and temporally with a correlation coefficient of 0.67 and an IoU of 69.7%, respectively.

scholarly journals Lung Ultrasound for Initial Diagnosis and Subsequent Monitoring of Aspiration Pneumonia in Elderly in Home Medical Care Setting

2019 ◽  
Vol 5 ◽  
pp. 233372141985844
Author(s):  
Hirofumi Namiki ◽  
Tadashi Kobayashi
Keyword(s):  
Aspiration Pneumonia ◽  
Care Setting ◽  
Lung Ultrasound ◽  
Initial Diagnosis ◽  
Resource Limited Settings ◽  
X Ray ◽  
Resource Limited ◽  
X Ray Imaging ◽  
Chest X Ray ◽  
B Lines

The number of aspiration pneumonia cases has increased in recent times. A definitive diagnosis of aspiration pneumonia is difficult in resource-limited settings where radiological equipment is unavailable. We report the initial diagnosis and subsequent monitoring of aspiration pneumonia in a home medical care setting. An 88-year-old Japanese male presented an acute onset of dyspnea, fever, and productive cough. At home, lung ultrasound displayed pleural effusion along with B-lines and subpleural consolidations. Upon admission, tests revealed increased total leucocyte counts with left-shifted neutrophils, elevated C-reactive protein levels, and positive sputum Gram stain. Chest X-ray imaging and computed tomography (CT) showed bibasilar infiltrates and wall thickening in the left S10 bronchi. The patient was diagnosed with aspiration pneumonia and treated with an antibiotic. After a 10-day hospitalization, lung ultrasound showed some remaining B-lines and disappearance of pleural effusion and subpleural consolidation. Chest X-ray image was normal, and CT revealed pleural abnormality and disappearance of bibasilar infiltrates, consistent with the ultrasound findings. Aspiration pneumonia develops with various clinical signs. However, diagnosis using chest X-ray imaging or CT in resource-limited settings is difficult. Ultrasound might allow physicians to make more accurate judgments, particularly while monitoring aspiration pneumonia following initial diagnosis in resource-limited settings.

scholarly journals LUNG ULTRASOUND FINDINGS IN MEXICAN PATIENTS WITH SARS COV2 INFECTION

2020 ◽  
Author(s):  
Luis Fernando L Paredes ◽  
Ivan I Ilescas
Keyword(s):  
Lung Ultrasound ◽  
Early Recognition ◽  
Point Of Care ◽  
New Disease ◽  
Rt Pcr ◽  
Resource Limited ◽  
New Variant ◽  
Ultrasound Findings ◽  
Short Time ◽  
B Lines

In late December 2019, a new disease reported at the time by an unknown pathogen was reported, which was later found to be a new variant of coronavirus, now called SARS-CoV2. This new disease had a very rapid global spread, causing multiple deaths in a short time, and which led to putting the entire world on health alert. In patients who have this disease, they present bilateral opacities in frosted multilobar glass with peripheral distribution. Some authors have suggested the use of ultrasound at the point of care for its early recognition. In this study, we evaluated the findings of lung ultrasound in 25 patients admitted to the General Hospital Dr. Enrique Cabrera, Mexico, with a diagnosis confirmed by RT-PCR of SARS CoV2. This small retrospective study suggests that artifacts like glass rockets with or without the Birolleau variant (White lung), confluent B-lines, thick irregular pleural lines, and variable size (subpleural) consolidations are typical findings of lung ultrasound in patients with COVID-19 pneumonia. The presence of these findings is useful when evaluating patients with suspected COVID-19. In resource-limited and austere settings where chest radiography, CT, and RT-PCR are not available or the response time is long, lung ultrasound performed by trained personnel can be an aid in the diagnosis of COVID-19.

scholarly journals Automatic Detection of B-lines in Lung Ultrasound Videos from Severe Dengue Patients

2021 ◽  
Author(s):  
Hamideh Kerdegari ◽  
Phung Tran Huy Nhat ◽  
Angela McBride ◽  
Reza Razavi ◽  
Nguyen Van Hao ◽  
...  
Keyword(s):  
Lung Ultrasound ◽  
Automatic Detection ◽  
Severe Dengue ◽  
B Lines

scholarly journals POS0842 LUNG ULTRASOUND TO ASSESS THE SEVERITY OF INTERSTITIAL LUNG DISEASE IN SYSTEMIC SCLEROSIS

2021 ◽  
Vol 80 (Suppl 1) ◽  
pp. 675.1-675
Author(s):  
C. Bruni ◽  
L. Mattolini ◽  
L. Tofani ◽  
L. Gargani ◽  
N. Landini ◽  
...  
Keyword(s):  
Systemic Sclerosis ◽  
Interstitial Lung Disease ◽  
Lung Disease ◽  
Quantitative Methods ◽  
Lung Ultrasound ◽  
Significant Negative Correlation ◽  
High Resolution Computed Tomography ◽  
Quantitative Score ◽  
Quantitative Scoring ◽  
B Lines

Background:Interstitial lung disease (ILD) is one of the most common complications and one of the main causes of morbidity and mortality in Systemic Sclerosis (SSc). High-resolution computed tomography (HRCT) is the gold standard for the diagnosis of ILD and it allows its quantification. Among semi-quantitative methods, Goh et al proposed a semi-quantitative scoring system to visually quantify ILD extent, with categorical cut-off of 20% to distinguish limited and extensive parenchymal involvement with prognostic implications. More recently, the use of radiomics has allowed the objective quantification of ILD through the use of dedicated software, which calculate different parameters of lung density.Given the exposure to ionizing radiation that the procedure entails, other methods of ILD evaluation are being studied, among which lung ultrasound (LUS) identifies the B-lines as a main feature of ILD. So far, different evidences have proposed the use of LUS for the screening of ILD, even in the early phases of the disease and in subclinical lung involvement.Objectives:the aim of this study is to test the role of LUS in quantifying the severity of SSc-ILD, evaluated with both semi-quantitative visual radiological and quantitative radiomic scores.Methods:Adult SSc patients classified according to the ACR/EULAR 2013 criteria patients were assessed with pulmonary function test (PFTs), lung ultrasound and HRCT over 60 days. CT images were analysed qualitatively (by presence/absence of ILD), semi-quantitatively (categorical Goh score <20% vs> 20% of extent and the continuous extent Goh score made from 5 levels’ assessment– 0 to 100%) and quantitatively [with the densitometric radiomic data obtained through the Horos software - Mean lung attenuation (MLA), Standard Deviation (SD), Kurtosis, Skewness and Lung volume (LV)]. LUS was used to quantify the B-lines detected in each patient by scanning a total of 13 intercostal spaces, on both anterior and posterior chest wall.Results:Among 59 SSc patients (81% women, mean age 48±14 years, 45% anti-Scl70 positive), 23 (39%) presented ILD on HRCT, of which 14 limited and 9 extensive. The mean visual semi-quantitative score was 6%, ranging from 0 to 66%. Our data showed a significantly different number of B-Lines in ILD vs non-ILD patients (median 38 vs 9, p <.005), a result which was further confirmed among non-ILD vs ILD> 20% (median 47 vs 9, p=.001) and ILD <20% (median 36 vs 9, p=.001) patients. Conversely, the number of B-lines was not statistically different between patients with ILD <20% and >20% (median 47 vs 36, p=.78). We observed a significant negative correlation between the number of B-lines and FVC (r=-.472, p<.05) TLC (r=-.436, p=.003), DLco (r=-.515, p<.001), DLCO/VA (r=.-306, p=.03). Finally, the number of B-lines showed a statistically significant correlation with the Goh score on 5 levels (r=.437, p=.001), MLA (r=.571, p<.001), kurtosis (r=-.285, p=.028), skewness (r=-.370, p = .004) and LV (r=-.277, p=.033). All data were confirmed analysing anterior and posterior B-Lines separately.Conclusion:Our study confirms that LUS represents a useful tool for the identification of SSc-ILD. In addition, we showed that LUS may be useful also for the quantification of the severity of SSc-ILD, by correlating with PFT parameters, radiomics parameters and visual radiological evaluation. Together with the PFTs, LUS could be used to increase the accuracy of the screening and, potentially, of the follow-up of SSc-ILD patients.Disclosure of Interests:Cosimo Bruni: None declared, Lavinia Mattolini: None declared, Lorenzo Tofani: None declared, Luna Gargani Consultant of: GE Healthcare, Philips Healthcare and Caption Health, Nicholas Landini: None declared, Gemma Lepri: None declared, Martina Orlandi: None declared, Serena Guiducci: None declared, Silvia Bellando Randone: None declared, Marco Matucci-Cerinic: None declared

scholarly journals Patients in intensive care unit for COVID-19 pneumonia: the lung ultrasound patterns at admission and discharge. An observational pilot study

2021 ◽  
Vol 13 (1) ◽  
Author(s):  
Persona Paolo ◽  
Valeri Ilaria ◽  
Zarantonello Francesco ◽  
Forin Edoardo ◽  
Sella Nicolò ◽  
...  
Keyword(s):  
Intensive Care ◽  
Lung Ultrasound ◽  
Low Cost ◽  
Low Frequency ◽  
Specific Pattern ◽  
University Hospital ◽  
Left Hemithorax ◽  
Significant Difference ◽  
The Right ◽  
B Lines

Abstract Background During COVID-19 pandemic, optimization of the diagnostic resources is essential. Lung Ultrasound (LUS) is a rapid, easy-to-perform, low cost tool which allows bedside investigation of patients with COVID-19 pneumonia. We aimed to investigate the typical ultrasound patterns of COVID-19 pneumonia and their evolution at different stages of the disease. Methods We performed LUS in twenty-eight consecutive COVID-19 patients at both admission to and discharge from one of the Padua University Hospital Intensive Care Units (ICU). LUS was performed using a low frequency probe on six different areas per each hemithorax. A specific pattern for each area was assigned, depending on the prevalence of A-lines (A), non-coalescent B-lines (B1), coalescent B-lines (B2), consolidations (C). A LUS score (LUSS) was calculated after assigning to each area a defined pattern. Results Out of 28 patients, 18 survived, were stabilized and then referred to other units. The prevalence of C pattern was 58.9% on admission and 61.3% at discharge. Type B2 (19.3%) and B1 (6.5%) patterns were found in 25.8% of the videos recorded on admission and 27.1% (17.3% B2; 9.8% B1) on discharge. The A pattern was prevalent in the anterosuperior regions and was present in 15.2% of videos on admission and 11.6% at discharge. The median LUSS on admission was 27.5 [21–32.25], while on discharge was 31 [17.5–32.75] and 30.5 [27–32.75] in respectively survived and non-survived patients. On admission the median LUSS was equally distributed on the right hemithorax (13; 10.75–16) and the left hemithorax (15; 10.75–17). Conclusions LUS collected in COVID-19 patients with acute respiratory failure at ICU admission and discharge appears to be characterized by predominantly lateral and posterior non-translobar C pattern and B2 pattern. The calculated LUSS remained elevated at discharge without significant difference from admission in both groups of survived and non-survived patients.

scholarly journals 19 Automated Quantification Of B-Lines in Lung Ultrasound On COVID-19 Patients

2021 ◽  
Vol 78 (2) ◽  
pp. S9-S10
Author(s):  
C. Baloescu ◽  
A. Chen ◽  
B. Raju ◽  
N. Evans ◽  
C.L. Moore
Keyword(s):  
Lung Ultrasound ◽  
Automated Quantification ◽  
B Lines

scholarly journals Lung Ultrasound Patterns and Clinical-Laboratory Correlates during COVID-19 Pneumonia: A Retrospective Study from North East Italy

2021 ◽  
Vol 10 (6) ◽  
pp. 1288
Author(s):  
Riccardo Senter ◽  
Federico Capone ◽  
Stefano Pasqualin ◽  
Lorenzo Cerruti ◽  
Leonardo Molinari ◽  
...  
Keyword(s):  
Lung Ultrasound ◽  
Clinical Laboratory ◽  
Imaging Modality ◽  
Case Series ◽  
Severity Index ◽  
Disease Stage ◽  
Retrospective Case ◽  
North East ◽  
High Flow Oxygen ◽  
B Lines

Background and Aim. Lung ultrasound (LUS) is a convenient imaging modality in the setting of coronavirus disease-19 (COVID-19) because it is easily available, can be performed bedside and repeated over time. We herein examined LUS patterns in relation to disease severity and disease stage among patients with COVID-19 pneumonia. Methods. We performed a retrospective case series analysis of patients with confirmed SARS-CoV-2 infection who were admitted to the hospital because of pneumonia. We recorded history, clinical parameters and medications. LUS was performed and scored in a standardized fashion by experienced operators, with evaluation of up to 12 lung fields, reporting especially on B-lines and consolidations. Results. We included 96 patients, 58.3% men, with a mean age of 65.9 years. Patients with a high-risk quick COVID-19 severity index (qCSI) were older and had worse outcomes, especially for the need for high-flow oxygen. B-lines and consolidations were located mainly in the lower posterior lung fields. LUS patterns for B-lines and consolidations were significantly worse in all lung fields among patients with high versus low qCSI. B-lines and consolidations were worse in the intermediate disease stage, from day 7 to 13 after onset of symptoms. While consolidations correlated more with inflammatory biomarkers, B-lines correlated more with end-organ damage, including extrapulmonary involvement. Conclusions. LUS patterns provide a comprehensive evaluation of patients with COVID-19 pneumonia that correlated with severity and dynamically reflect disease stage. LUS patterns may reflect different pathophysiological processes related to inflammation or tissue damage; consolidations may represent a more specific sign of localized disease, whereas B-lines seem to be also dependent upon generalized illness due to SARS-CoV-2 infection.

Comparison of lung bedside ultrasound and BNP in the diagnosis of acute heart failure in the emergency department

2020 ◽  
Vol 8 (1) ◽  
pp. 50-50
Author(s):  
Seyed Hossein Ojaghi Haghighi ◽  
Neda Hamed ◽  
Shiva Ebrahimi ◽  
Jafar Ghobadi ◽  
Hoorolnesa Ameli
Keyword(s):  
Heart Failure ◽  
Emergency Room ◽  
Pulmonary Edema ◽  
Acute Heart Failure ◽  
Lung Ultrasound ◽  
P Value ◽  
Ultrasound Evaluation ◽  
Ultrasound Findings ◽  
Medical Advances ◽  
B Lines

Introduction: Congestive heart failure is heart muscle failure that causes pulmonary congestion and eventually pulmonary edema, which despite recent medical advances, is still a progressive syndrome with high mortality, the prevalence of which has increased in recent decades. Therefore, in this study we compared lung ultrasound findings in acute heart failure patients with the BNP. Methods: This study was performed in the emergency room of Imam Reza hospital in Tabriz. For patients entering the emergency room after taking a history, both standard gold (BNP) tests and beside ultrasound of the lung were performed. Ultrasound was performed at the same time as obtaining blood sample to ensure that the ultrasound specialist did not know the result of diagnosis. During the ultrasound, if there were multiple B-Lines that were at least 3 mm apart, patient was diagnosed with pulmonary edema due to heart failure. Results: Number of participants in this study was 108 people, 54.6% of whom were men and the rest were women. The correlation coefficient between width and number of kerley lines was 0.79, between NT-pro BNP and width of kerley lines was 0.65 and between NT-pro BNP and number of kerley lines was 0.77, which indicates a significant positive correlation (P value <0.001). Conclusion: The results of present study showed that in patients with acute heart failure, the number and width of kerley lines in pulmonary ultrasound evaluation increase rapidly. There is also a high correlation between number and length of kerley lines with NT-pro BNP serum values.

Optimal Scale of Hierarchical Image Segmentation with Scribbles Guidance for Weakly Supervised Semantic Segmentation

2021 ◽  
pp. 2154026
Author(s):  
Zaid Al-Huda ◽  
Donghai Zhai ◽  
Yan Yang ◽  
Riyadh Nazar Ali Algburi
Keyword(s):  
Image Segmentation ◽  
Graphical Model ◽  
Semantic Segmentation ◽  
Saliency Map ◽  
Training Data ◽  
Deep Convolutional Neural Networks ◽  
High Quality ◽  
Optimal Scale ◽  
Supervised Segmentation ◽  
Weakly Supervised

Deep convolutional neural networks (DCNNs) trained on the pixel-level annotated images have achieved improvements in semantic segmentation. Due to the high cost of labeling training data, their applications may have great limitation. However, weakly supervised segmentation approaches can significantly reduce human labeling efforts. In this paper, we introduce a new framework to generate high-quality initial pixel-level annotations. By using a hierarchical image segmentation algorithm to predict the boundary map, we select the optimal scale of high-quality hierarchies. In the initialization step, scribble annotations and the saliency map are combined to construct a graphic model over the optimal scale segmentation. By solving the minimal cut problem, it can spread information from scribbles to unmarked regions. In the training process, the segmentation network is trained by using the initial pixel-level annotations. To iteratively optimize the segmentation, we use a graphical model to refine segmentation masks and retrain the segmentation network to get more precise pixel-level annotations. The experimental results on Pascal VOC 2012 dataset demonstrate that the proposed framework outperforms most of weakly supervised semantic segmentation methods and achieves the state-of-the-art performance, which is [Formula: see text] mIoU.

Sign in / Sign up

Export Citation Format

Share Document