Lung Ultrasound Interpretation Research Articles

Clinical benefit of AI-assisted lung ultrasound in a resource-limited intensive care unit

BackgroundInterpreting point-of-care lung ultrasound (LUS) images from intensive care unit (ICU) patients can be challenging, especially in low- and middle- income countries (LMICs) where there is limited training available. Despite recent advances in the use of Artificial Intelligence (AI) to automate many ultrasound imaging analysis tasks, no AI-enabled LUS solutions have been proven to be clinically useful in ICUs, and specifically in LMICs. Therefore, we developed an AI solution that assists LUS practitioners and assessed its usefulness in a low resource ICU.MethodsThis was a three-phase prospective study. In the first phase, the performance of four different clinical user groups in interpreting LUS clips was assessed. In the second phase, the performance of 57 non-expert clinicians with and without the aid of a bespoke AI tool for LUS interpretation was assessed in retrospective offline clips. In the third phase, we conducted a prospective study in the ICU where 14 clinicians were asked to carry out LUS examinations in 7 patients with and without our AI tool and we interviewed the clinicians regarding the usability of the AI tool.ResultsThe average accuracy of beginners’ LUS interpretation was 68.7% [95% CI 66.8–70.7%] compared to 72.2% [95% CI 70.0–75.6%] in intermediate, and 73.4% [95% CI 62.2–87.8%] in advanced users. Experts had an average accuracy of 95.0% [95% CI 88.2–100.0%], which was significantly better than beginners, intermediate and advanced users (p < 0.001). When supported by our AI tool for interpreting retrospectively acquired clips, the non-expert clinicians improved their performance from an average of 68.9% [95% CI 65.6–73.9%] to 82.9% [95% CI 79.1–86.7%], (p < 0.001). In prospective real-time testing, non-expert clinicians improved their baseline performance from 68.1% [95% CI 57.9–78.2%] to 93.4% [95% CI 89.0–97.8%], (p < 0.001) when using our AI tool. The time-to-interpret clips improved from a median of 12.1 s (IQR 8.5–20.6) to 5.0 s (IQR 3.5–8.8), (p < 0.001) and clinicians’ median confidence level improved from 3 out of 4 to 4 out of 4 when using our AI tool.ConclusionsAI-assisted LUS can help non-expert clinicians in an LMIC ICU improve their performance in interpreting LUS features more accurately, more quickly and more confidently.

Comparison of pulmonary congestion severity using artificial intelligence-assisted scoring versus clinical experts: a secondary analysis of BLUSHED-AHF.

Acute decompensated heart failure (ADHF) is the leading cause of cardiovascular hospitalizations in the United States. Detecting B-lines through lung ultrasound (LUS) can enhance clinicians' prognostic and diagnostic capabilities. Artificial intelligence/machine learning (AI/ML)-based automated guidance systems may allow novice users to apply LUS to clinical care. We investigated whether an AI/ML automated LUS congestion score correlates with expert's interpretations of B-line quantification from an external patient dataset. This was a secondary analysis from the BLUSHED-AHF study which investigated the effect of LUS-guided therapy on patients with ADHF. In BLUSHED-AHF, LUS was performed and B-lines were quantified by ultrasound operators. Two experts then separately quantified the number of B-lines per ultrasound video clip recorded. Here, an AI/ML-based lung congestion score (LCS) was calculated for all LUS clips from BLUSHED-AHF. Spearman correlation was computed between LCS and counts from each of the original three raters. A total of 3858 LUS clips were analysed on 130 patients. The LCS demonstrated good agreement with the two experts' B-line quantification score (r = 0.894, 0.882). Both experts' B-line quantification scores had significantly better agreement with the LCS than they did with the ultrasound operator's score (p < 0.005, p < 0.001). Artificial intelligence/machine learning-based LCS correlated with expert-level B-line quantification. Future studies are needed to determine whether automated tools may assist novice users in LUS interpretation.

Lung ultrasound: A comparison of image interpretation accuracy between curvilinear and phased array transducers.

Both curvilinear and phased array transducers are commonly used to perform lung ultrasound (LUS). This study seeks to compare LUS interpretation accuracy of images obtained using a curvilinear transducer with those obtained using a phased array transducer. We invited 166 internists and trainees to interpret 16 LUS images/cineloops of eight patients in an online survey: eight curvilinear and eight phased array, performed on the same lung location. Images depicted normal lung, pneumothorax, pleural irregularities, consolidation/hepatisation, pleural effusions and B-lines. Primary outcome for each participant is the difference in image interpretation accuracy scores between the two transducers. A total of 112 (67%) participants completed the survey. The mean paired accuracy score difference between the curvilinear and phased array images was 3.0% (95% CI: 0.6 to 5.4%, P = 0.015). For novices, scores were higher on curvilinear images (mean difference: 5.4%, 95% CI: 0.9 to 9.9%, P = 0.020). For non-novices, there were no differences between the two transducers (mean difference: 1.4%, 95% CI: -1.1 to 3.9%, P = 0.263). For pleural-based findings, the mean of the paired differences between transducers was higher in the novice group (estimated mean difference-in-differences: 9.5%, 95% CI: 0.6 to 18.4%; P = 0.036). No difference in mean accuracies was noted between novices and non-novices for non-pleural-based pathologies (estimated mean difference-in-differences: 0.6%, 95% CI to 5.4-6.6%; P = 0.837). Lung ultrasound images obtained using the curvilinear transducer are associated with higher interpretation accuracy than the phased array transducer. This is especially true for novices interpreting pleural-based pathologies.

Lung Ultrasound in Critical Care and Emergency Medicine: Clinical Review.

Lung ultrasound has become a part of the daily examination of physicians working in intensive, sub-intensive, and general medical wards. The easy access to hand-held ultrasound machines in wards where they were not available in the past facilitated the widespread use of ultrasound, both for clinical examination and as a guide to procedures; among point-of-care ultrasound techniques, the lung ultrasound saw the greatest spread in the last decade. The COVID-19 pandemic has given a boost to the use of ultrasound since it allows to obtain a wide range of clinical information with a bedside, not harmful, repeatable examination that is reliable. This led to the remarkable growth of publications on lung ultrasounds. The first part of this narrative review aims to discuss basic aspects of lung ultrasounds, from the machine setting, probe choice, and standard examination to signs and semiotics for qualitative and quantitative lung ultrasound interpretation. The second part focuses on how to use lung ultrasound to answer specific clinical questions in critical care units and in emergency departments.

Self-Supervised Pretraining Improves Performance and Inference Efficiency in Multiple Lung Ultrasound Interpretation Tasks

Self-Supervised Pretraining Improves Performance and Inference Efficiency in Multiple Lung Ultrasound Interpretation Tasks

Volume sweep imaging lung teleultrasound for detection of COVID-19 in Peru: a multicentre pilot study

ObjectivesPulmonary disease is a significant cause of morbidity and mortality in adults and children, but most of the world lacks diagnostic imaging for its assessment. Lung ultrasound is a portable,...

Ultrasound evaluation of interstitial lung disease in rheumatoid arthritis and autoimmune diseases.

The interpretation of lung ultrasound (US) is the result of the analysis of artifacts, rather than exact representations of anatomical structures, which appear when changes in the physical properties of the lung occur. Its application to the study of interstitial lung disease (ILD) associated with autoimmune diseases has aroused great interest in the last 10 years, as evidenced by a growing number of publications studying its usefulness in the diagnostic process, as a prognostic marker, and as an aid in monitoring of patients. The main elements in lung US interpretation in ILD are the B lines and the changes in the pleural line. B lines are vertical artifacts that are generated when there is a partial decrease in the air content of the lung parenchyma and/or the volume of the interstitial area expands. Pleural line alterations that can be seen are irregularities, thickening, fragmentation, or subpleural nodules. Both the B lines and the changes in the pleural line have shown a significant positive correlation with the evidence on chest computed tomography (high-resolution computed tomography [HRCT]) of ILD associated with autoimmune diseases, with sensitivity and negative predictive values of up to 100%. These results, together with the safety, accessibility, and low cost of lung US, support this imaging technique as a promising screening method for optimizing the indication for HRCT. The role of lung US regarding sensitivity to change needs further investigation with multicenter prospective studies.

Diagnostic value of lung ultrasonography in respiratory disorders of neonates

Background Ionizing radiation-dependent modalities are still the mainstay investigations used in the neonatal intensive care unit (NICU). Neonates are particularly vulnerable to biohazards of ionizing radiation, a fact that renders searching for radiation-free modality a priority in the neonatal care setting.Patients and methods Patients who participated in this case–control study were classified into two groups: 127 neonates who presented to the NICU with respiratory distress manifestations were enrolled as the ‘case’ group, and 80 neonates 80 of comparable age and sex were selected and enrolled as the ‘control’ group, as they all had nonrespiratory manifestations. All participants were subjected to full history taking, thorough physical examinations, and routine investigations. Chest radiograph was performed on first presentation and as needed for follow-up as per the unit protocol; concomitantly, lung ultrasound (LUS) was performed with each radiograph. Throughout our study, if LUS interpretation in the context of clinical data reported the same diagnosis as clinical and radiography did, we considered the case as matching (positive) and vice versa.Results A total of 618 chest radiograph and concomitant LUS images were included in the study, as 207 (127 cases and 80 controls) images for initial diagnosis and 411 for follow-up of neonates presented with respiratory distress. The average number of images was 4.2 images/case during the follow-up (range, 1–22). Cases displayed lighter birth weight, with significantly higher prevalence of low birth weight and very low birth weight, than controls. Moreover, most cases (84%) were delivered by cesarean section. Among cases, transient tachypnea of newborn and respiratory distress syndrome were the most common diagnosis, whereas neonatal hyperbilirubinemia and seizures were the most frequent diagnosis among controls. Plain radiograph was the reference modality for diagnosis and follow-up (100% of cases had radiologic findings, whereas all controls had free chest radiograph). Overall, 8/127 (6%) cases of respiratory distress, clinically and radiologically, displayed normal LUS images at the initial presentation (false negative), whereas 3/80 (4%) of controls, who presented with nonrespiratory manifestations and had free chest radiograph, showed numerous B-lines on LUS (false positive). Nonsignificant differences were confirmed between chest radiograph findings and those of LUS in the initial diagnosis of different neonatal respiratory disorders, with sensitivity of 94%, specificity of 96%, positive predictive value of 97.5%, and negative predictive value of 90.5% for LUS as compared with plain chest radiograph. Similar matching between diagnosis concluded by radiographs and concomitant LUS was appreciated during follow-up of different respiratory diseases.Conclusion LUS is a suitable modality in NICU setting as it is a safe, radiation-free, dynamic, and real-time bedside assessment. It has an accepted validity in diagnosis and follow-up of different respiratory disorders in neonates with even higher abilities in detection of transient tachypnea of newborn, pleural effusion, and atelectasis.

Automation of Lung Ultrasound Interpretation via Deep Learning for the Classification of Normal versus Abnormal Lung Parenchyma: A Multicenter Study.

Lung ultrasound (LUS) is an accurate thoracic imaging technique distinguished by its handheld size, low-cost, and lack of radiation. User dependence and poor access to training have limited the impact and dissemination of LUS outside of acute care hospital environments. Automated interpretation of LUS using deep learning can overcome these barriers by increasing accuracy while allowing point-of-care use by non-experts. In this multicenter study, we seek to automate the clinically vital distinction between A line (normal parenchyma) and B line (abnormal parenchyma) on LUS by training a customized neural network using 272,891 labelled LUS images. After external validation on 23,393 frames, pragmatic clinical application at the clip level was performed on 1162 videos. The trained classifier demonstrated an area under the receiver operating curve (AUC) of 0.96 (±0.02) through 10-fold cross-validation on local frames and an AUC of 0.93 on the external validation dataset. Clip-level inference yielded sensitivities and specificities of 90% and 92% (local) and 83% and 82% (external), respectively, for detecting the B line pattern. This study demonstrates accurate deep-learning-enabled LUS interpretation between normal and abnormal lung parenchyma on ultrasound frames while rendering diagnostically important sensitivity and specificity at the video clip level.

The Inter-Rater Reliability of Pediatric Point-of-Care Lung Ultrasound Interpretation in Children With Acute Respiratory Failure.

Use of point-of-care lung ultrasound (POC-LUS) has increased significantly in pediatrics yet it remains under-studied in the pediatric intensive care unit (PICU). No studies explicitly evaluate the reliability of POC-LUS artifact interpretation among critically ill children with acute respiratory failure (ARF) in the PICU. We thus designed this study to determine the inter-rater reliability of POC-LUS interpretation in pediatric ARF among pediatric intensivists trained in POC-LUS and an expert intensivist. We compared the interpretation of lung sliding, pleural line characteristics, ultrasound artifacts, and POC-LUS diagnoses among pediatric intensivists and an expert intensivist in a cohort of children admitted to the PICU for ARF. Kappa statistics (k) adjusted for maximum attainable agreement (k/kmax ) were used to quantify chance-correct agreement between the pediatric intensivist and expert physician. We enrolled 88 patients, evaluating 3 zones per hemithorax (anterior, lateral, and posterior) for lung sliding, pleural line characteristics, ultrasound artifacts, and diagnosis. There was moderate agreement between the PICU intensivist and expert-derived diagnoses with 56% observed agreement (k/kmax = 0.46, 95% confidence interval [CI] 0.31-0.65). Agreement in identification of lung sliding (k= 0.19, 95% CI -0.17 to 0.56) and pleural line characteristics (k= 0.24, 95% CI 0.08-0.40) was slight and fair, respectively, while agreement in the interpretation of ultrasound artifacts ranged from moderate to substantial. Evidence supporting the evaluation of neonatal and adult patients with POC-LUS should not be extrapolated to critically ill pediatric patients. This study adds to the evidence supporting use of POC-LUS in the PICU by demonstrating moderate agreement between PICU intensivist and expert-derived POC-LUS diagnoses.

A regional multicentric survey on variability and accuracy in lung ultrasound interpretation during the COVID-19 pandemic.

A regional multicentric survey on variability and accuracy in lung ultrasound interpretation during the COVID-19 pandemic.

Point-of-care lung ultrasound patterns in late third-trimester gravidas with and without preeclampsia

Transthoracic point-of-care ultrasonography of the lungs has become a standard technique in critical care medicine for the evaluation of patients with respiratory signs or symptoms but has not been well studied in pregnancy. To compare lung ultrasound patterns in third-trimester gravidas with and without preeclampsia and assess interobserver agreement between 3 obstetrical providers and a physician expert in critical care lung ultrasound. This is a prospective observational study of 262 women with singleton pregnancies between 32 0/7 and 41 6/7 weeks' gestation. Lung ultrasound examinations were performed and interpreted by a team of obstetrical care providers and then interpreted by an expert in point-of-care lung ultrasound. The number of B-lines in each of the 4 lung fields, indicating the accumulation of fluid in the interstitial space and the alveoli, was evaluated. The primary outcome was a positive study for pulmonary interstitial edema, defined as an ultrasound study with 3 or more B-lines in 2 or more bilateral lung fields. The secondary outcome was a lung ultrasound study with 1 or 2 B-lines in 1 or more lung fields or 3 B-lines in 1 lung field. Interobserver agreement in lung ultrasound interpretation between obstetrical care providers and an expert in critical care point-of-care ultrasonography of the lung was assessed. Among healthy gravidas, no subject had a lung ultrasound examination positive for pulmonary interstitial edema. Notably, 2 patients with preeclampsia had positive lung ultrasound studies, and both had respiratory symptoms or signs of pulmonary edema. One or 2 B-lines or 3 B-lines in 1 lung field were identified in 11.4% of healthy gravidas and 18.6% of patients with preeclampsia. There was no difference in lung ultrasound patterns between healthy gravidas and those with preeclampsia. The obstetrical care providers and the lung ultrasound expert had a high proportion of agreement regarding the interpretation of negative lung ultrasound examinations. The 2 patients with clinical signs of pulmonary edema were judged to have positive studies by both the obstetrical team and the expert; however, the obstetrical team classified more studies as positive. Lung ultrasound patterns in women with preeclampsia without respiratory symptoms or clinical signs of pulmonary edema are similar to the lung ultrasound patterns of healthy gravidas. Point-of-care lung ultrasound can be used to evaluate third-trimester gravidas with preeclampsia and respiratory complaints or signs concerning for pulmonary edema. Formal training is important before the widespread adoption of point-of-care lung ultrasound by obstetrical healthcare providers.

Interobserver Agreement of Lung Ultrasound Findings of COVID-19

Background: Lung ultrasound (LUS) has received considerable interest in the clinical evaluation of patients with COVID-19. Previously described LUS manifestations for COVID-19 include B-lines, consolidations, and pleural thickening. The interrater reliability (IRR) of these findings for COVID-19 is unknown.Methods: This study was conducted at conducted at two academic medical centers between 03/2020-06/2020. Nine physicians (hospitalists: n=4; emergency medicine: n=5) independently evaluated n=20 LUS scans (n=180 independent observations) collected from RT-PCR confirmed COVID-19 patients. These studies were randomly selected from an image database consisting of COVID-19 patients evaluated in the emergency department with portable ultrasound devices. Physicians were blinded to any patient information or previous LUS interpretation. Kappa values (κ) were used to calculate IRR.Findings: There was substantial IRR on the following items: normal LUS scan (κ=0.79 [95% CI: 0.72-0.87]), presence of B-lines (κ=0.79 [95% CI: 0.72-0.87]), >=3 B-lines observed (κ=0.72 [95% CI: 0.64-0.79]). Moderate IRR was observed for the presence of any consolidation (κ=0.57 [95% CI: 0.50-0.64]), subpleural consolidation (κ=0.49 [95% CI: 0.42-0.56]), and presence of effusion (κ=0.49 [95% CI: 0.41-0.56]). Fair IRR was observed for pleural thickening (κ=0.23 [95% CI: 0.15-0.30]).Interpretation: Many LUS manifestations for COVID-19 appear to have moderate to substantial IRR across providers from multiple specialties utilizing differing portable devices.Trial Registration: This analysis was conducted as part of an ongoing study investigating the role of POCUS for COVID-19 that is being conducted at four medical centers in the United States (Clinicaltrials.gov Registration: NCT04384055). Funding Statement: None.Declaration of Interests: Andre Kumar, MD, MEd is a paid consultant for Vave Health, which manufactures one of the ultrasound devices used in this study. His consultant duties include providing feedback on product development. The other authors do not have any items to disclose.Ethics Approval Statement: This study has received institutional review board (IRB) approval at all participating sites.

Impact of bedside lung ultrasound on physician clinical decision-making in an emergency department in Nepal

BackgroundLung ultrasound is an effective tool for the evaluation of undifferentiated dyspnea in the emergency department. Impact of lung ultrasound on clinical decisions for the evaluation of patients with dyspnea in resource-limited settings is not well-known. The objective of this study was to evaluate the impact of lung ultrasound on clinical decision-making for patients presenting with dyspnea to an emergency department in the resource-limited setting of Nepal.MethodsA prospective, cross-sectional study of clinicians working in the Patan Hospital Emergency Department was performed. Clinicians performed lung ultrasounds on patients presenting with dyspnea and submitted ultrasounds with their pre-test diagnosis, lung ultrasound interpretation, post-test diagnosis, and any change in management.ResultsTwenty-two clinicians participated in the study, completing 280 lung ultrasounds. Diagnosis changed in 124 (44.3%) of patients with dyspnea. Clinicians reported a change in management based on the lung ultrasound in 150 cases (53.6%). Of the changes in management, the majority involved treatment (83.3%) followed by disposition (13.3%) and new consults (2.7%).ConclusionsIn an emergency department in Nepal, bedside lung ultrasound had a significant impact on physician clinical decision-making, especially on patient diagnosis and treatment.

Effect of Different Probes and Expertise on the Interpretation Reliability of Point-of-Care Lung Ultrasound

The effect of different probes and operator experience on the reliability of lung ultrasound (LU) interpretation has not been investigated. We studied the effect of probes and operator experience on the interpretation reliability of LU in critically ill neonates. This was a prospective, blind, cohort study enrolling patients with basic patterns ("B," "severe B," consolidation). Patients were scanned with microlinear (15 MHz; L15), phased-array sectorial (6-12 MHz; S7), and microconvex (8 MHz; C8) probes, in random order. Static images were acquired in high resolution, anonymized, and included in a pictorial database in random sequences. Seventeen clinicians with different LU experience were asked to blindly assess the pictorial database. Interrater agreement and interpretation reliability were analyzed. Subanalyses according to expertise and probe, and multivariate linear regression (including an "expertise× probe" interaction factor), were also performed. The agreement tends to be lower and more heterogeneous for residents (intraclass correlation coefficient [ICC], 0.82 [95%CI, 0.74-0.9], P< .001; I2, 67%, P= .04) and for fellows (ICC, 0.93 [95%CI, 0.9-0.97], P< .001; I2, 69%, P= .04), especially when using nonlinear probes, compared with senior physicians (ICC, 0.95 [95%CI, 0.93-0.96], P< .001; I2, 0%, P= .433). Area under the curve (AUC) values were high for all probes (L15, 0.96 [95%CI, 0.93-0.99]; C8, 0.91 [95%CI, 0.85-0.98]; S7, 0.86 [95%CI, 0.82-0.91]) and physicians (senior physicians, 0.95 [95%CI, 0.83-0.99]; fellows, 0.95 [95%CI, 0.75-0.99]; residents, 0.86 [95%CI, 0.5-0.99]). Worse reliability and higher heterogeneity were found when the evaluation was performed by residents (AUC, 0.9 [95%CI, 0.85-0.94], P< .01; I2, 93.6%, P< .001) than by fellows (AUC, 0.99 [95%CI, 0.9-0.999], P< .001; I2, 34.3%, P= .09) and/or by senior physicians (AUC, 0.99 [95%CI, 0.9-0.999], P< .001; I2, 18%, P= .236). The "expertise× probe" interaction factor was associated with lower ICC (standardized regression coefficient β, -0.69; P< .0001; adjusted R2, 0.99) and AUC (standardized regression coefficient β, -0.76; P< .0001; adjusted R2, 0.98). LU interpretation in neonates shows good interrater agreement and reliability, irrespective of the probe and rater expertise. The use of nonlinear probes by novice operators is associated with the lowest agreement and reliability.

The Utility of Bedside Lung Ultrasound Findings in Bronchiolitis

Recent literature suggests that bedside lung ultrasound may have a role in the evaluation of infants with bronchiolitis. B lines, which are multiple and diffuse vertical artifacts spreading from the lung pleural interface to the edge of the ultrasound screen, have been associated with thickened interlobular septa, extravascular lung water, and diffuse parenchymal disease. The aims of this study were (1) to describe the prevalence of B lines in children younger than 24 months presenting to the emergency department with wheezing, (2) to determine the interrater reliability of lung ultrasound findings in this setting, and (3) to determine the association of B lines with atopy and other clinical findings. This was a pilot, prospective, observational study of a convenience sample of patients younger than 2 years presenting with wheezing to a large academic pediatric hospital emergency department. Investigators performed lung ultrasound examinations, and a second provider reviewed the ultrasound examinations to determine interrater reliability. We performed univariate analyses to test for associations between ultrasound findings and atopy, acute illness severity, age, and treatment response. Studies were obtained on 29 patients (mean [SD] age, 291 [187] days; 62% male). Twenty-one patients (72%) had compact B lines. B lines were significantly associated with older age and an absence of atopic features. There was poor correlation of lung ultrasound examination interpretation among enrolling providers. In this small sample of patients with bronchiolitis, B lines were associated with older age and an absence of atopic features. Lung ultrasound interpretation had poor interrater reliability.