Transfer Learning-Based B-Line Assessment of Lung Ultrasound for Acute Heart Failure

OBJECTIVE

B-lines assessed by lung ultrasound (LUS) outperform physical exam, chest radiograph, and biomarkers for the associated diagnosis of acute heart failure (AHF) in the emergent setting. The use of LUS is however limited to trained professionals and suffers from interpretation variability. The objective was to utilize transfer learning to create an AI-enabled software that can aid novice users to automate LUS B-line interpretation.

METHODS

Data from an observational AHF LUS study provided standardized cine clips for AI model development and evaluation. A total of 49,952 LUS frames from 30 patients were hand scored and trained on a convolutional neural network (CNN) to interpret B-lines at the frame level. A random independent evaluation set of 476 LUS clips from 60 unique patients assessed model performance. The AI models scored the clips on both a binary and ordinal 0-4 multiclass assessment.

RESULTS

A multiclassification AI algorithm had the best performance at the binary level when applied to the independent evaluation set, AUC of 0.967 (95% CI 0.965-0.970) for detecting pathologic conditions. When compared to expert blinded reviewer, the 0-4 multiclassification AI algorithm scale had a reported linear weighted kappa of 0.839 (95% CI 0.804-0.871).

CONCLUSIONS

The multiclassification AI algorithm is a robust and well performing model at both binary and ordinal multiclass B-line evaluation. This algorithm has the potential to be integrated into clinical workflows to assist users with quantitative and objective B-line assessment for evaluation of AHF.

Transfer Learning for Automated COVID-19 B-Line Classification in Lung Ultrasound

Lung ultrasound (LUS) as a diagnostic tool is gaining support for its role in the diagnosis and management of COVID-19 and a number of other lung pathologies. B-lines are a predominant feature in COVID-19, however LUS requires a skilled clinician to interpret findings. To facilitate the interpretation, our main objective was to develop automated methods to classify B-lines as pathologic vs. normal. We developed transfer learning models based on ResNet networks to classify B-lines as pathologic (at least 3 B-lines per lung field) vs. normal using COVID-19 LUS data. Assessment of B-line severity on a 0-4 multi-class scale was also explored. For binary B-line classification, at the frame-level, all ResNet models pretrained with ImageNet yielded higher performance than the baseline nonpretrained ResNet-18. Pretrained ResNet-18 has the best Equal Error Rate (EER) of 9.1% vs the baseline of 11.9%. At the clip-level, all pretrained network models resulted in better Cohen's kappa agreement (linear-weighted) and clip score accuracy, with the pretrained ResNet-18 having the best Cohen's kappa of 0.815 [95% CI: 0.804-0.826], and ResNet-101 the best clip scoring accuracy of 93.6%. Similar results were shown for multi-class scoring, where pretrained network models outperformed the baseline model. A class activation map is also presented to guide clinicians in interpreting LUS findings. Future work aims to further improve the multi-class assessment for severity of B-lines with a more diverse LUS dataset.

Tricuspid Regurgitant Jet Velocity Point-of-Care Ultrasound Curriculum Development and Validation

INTRODUCTION

The American College of Emergency Physicians (ACEP) recommends that Emergency Medicine physicians with advanced training can evaluate right ventricular (RV) pressures via point-of-care ultrasound (POCUS) by measuring a tricuspid regurgitant jet (TRJ). We were unable to find a published curriculum to deliver education for this at any skill level. Therefore, we developed, delivered, and evaluated a curriculum for the assessment of TRJ for novice physician sonographers.

METHODS

We designed an educational intervention for novice physician sonographers. The curriculum was created using a modified Delphi methodology. All novice sonographers participated in the educational intervention which consisted of a didactic lecture followed by hands-on-deliberate practice on healthy medical student volunteers with expert feedback in a simulated setting. Sonographer's knowledge was assessed at 3 time points: pre-intervention, immediately post-intervention, and 3 months post-intervention (retention assessment) by multiple choice exam.

RESULTS

Nine novice physician sonographers participated in the intervention. Mean exam performance increased from 55.6% [standard deviation (SD) 11.3%] on the pre-intervention exam to 94.4% (SD 7.3%) on the post-intervention exam and 92.9% (SD 12.5%) on the retention exam. The mean improvement between the pre- and post- exam was +38.9% (95% CI 31.8 - 46.0), and between the pre-exam and retention exam +37.1% (95% CI 22.3 - 52.0).

CONCLUSION

Sonographer knowledge of TRJ assessment improved following a brief educational intervention as measured by exam performance. Given the expanding role of POCUS it is increasingly important to provide effective resources for teaching these skills. This work establishes the basis for further study and implementation of our TRJ curriculum.

Impact of point-of-care ultrasound on treatment time for ectopic pregnancy

BACKGROUND

Point-of-care ultrasound (POCUS) is useful in the evaluation of early pregnancy by confirming intrauterine pregnancy and recognizing hemorrhage from ectopic pregnancy. We sought to determine whether transabdominal POCUS by itself or in conjunction with consultative radiology ultrasound (RADUS), reduces Emergency Department (ED) treatment time for patients with ectopic pregnancy requiring operative care, when compared to RADUS alone. A secondary objective was to determine whether the incorporation of POCUS reduces time to operative care for patients with ruptured ectopic pregnancy specifically, when compared to RADUS alone.

METHODS

We performed a retrospective review of patients admitted for operative management of ectopic pregnancy. We excluded patients with known ectopic pregnancy and/or imaging prior to arriving to the treatment area, found not to have an ectopic pregnancy, or did not undergo operative care. Descriptive statistics, classical and nonparametric statistical analysis, and linear regression were performed.

RESULTS

There were 220 patients admitted with ectopic pregnancy, 111 met exclusion criteria, yielding 109 for analysis. Of 109, 36 received POCUS (23/36 also had RADUS), while 73 received RADUS only. Among the POCUS group 31/36 (86%) were classified as ruptured versus 47/73 (64%) in the RADUS group. The average ED treatment time in the POCUS group for all admitted ectopic pregnancies was 157.9 min (standard deviation [SD] 101.3) versus 206.3 min (SD 76.6) in the RADUS group (p = 0.0141). The median time to operating room (OR) for ruptured ectopic pregnancies was 203.0 min (interquartile range [IQR] 159.0) in the POCUS group versus 293.0 min (IQR 139.0) in the RADUS group (p = 0.0002). Regression analysis of the primary outcome was limited by multiple interactions and sample size. When controlling for race, positive shock index and ED visit time, POCUS was found to be associated with a significantly shorter time to OR among ruptured ectopic pregnancies compared to RADUS (p = 0.0052).

CONCLUSION

Compared to RADUS alone, incorporation of POCUS was associated with significantly faster ED treatment time for all ectopic pregnancies and significantly faster time to OR for ruptured ectopic pregnancies, even when combined with RADUS. When controlling for clinical differences, time to OR was still faster for patients who underwent POCUS. The integration of POCUS should be considered to expedite care for patients with ectopic pregnancy requiring operative care.

Novice Physician Ultrasound Evaluation of Pediatric Tricuspid Regurgitant Jet Velocity

Introduction

Pulmonary hypertension, associated with high mortality in pediatric patients, is traditionally screened for by trained professionals by measuring a tricuspid regurgitant jet velocity (TRJV). Our objective was to test the feasibility of novice physician sonographers (NPS) to perform echocardiograms of adequate quality to exclude pathology (defined as TRJV > 2.5 meters per second).

Methods

We conducted a cross-sectional study of NPS to assess TRJV by echocardiogram in an urban pediatric emergency department. NPS completed an educational course consisting of a didactic curriculum and hands-on workshop. NPS enrolled a convenience sample of patients aged 7–21 years. Our primary outcome was the proportion of echocardiograms with images of adequate quality to exclude pathology. Our secondary outcome was NPS performance on four image elements. We present descriptive statistics, binomial proportions, kappa coefficients, and logistic regression analysis.

Results

Eight NPS completed 80 echocardiograms. We found 82.5% (95% confidence interval [CI], 74.2–90.8) of echocardiograms had images of adequate quality to exclude pathology. Among image elements, NPS obtained a satisfactory, apical 4-chamber view in 85% (95% CI, 77.1–92.9); positioned the color box accurately 65% (95% CI, 54.5–75.5); optimized TRJV color signal 78.7% (95% CI, 69.8–87.7); and optimized continuous-wave Doppler in 55% (95% CI, 44.1–66.0) of echocardiograms.

Conclusion

NPS obtained images of adequate quality to exclude pathology in a majority of studies; however, optimized acquisition of specific image elements varied. This work establishes the basis for future study of NPS assessment of TRJV pathology when elevated pulmonary pressures are of clinical concern.

Point-of-care Lung Ultrasound Is More Sensitive than Chest Radiograph for Evaluation of COVID-19

INTRODUCTION

Current recommendations for diagnostic imaging for moderately to severely ill patients with suspected coronavirus disease 2019 (COVID-19) include chest radiograph (CXR). Our primary objective was to determine whether lung ultrasound (LUS) B-lines, when excluding patients with alternative etiologies for B-lines, are more sensitive for the associated diagnosis of COVID-19 than CXR.

METHODS

This was a retrospective cohort study of all patients who presented to a single, academic emergency department in the United States between March 20 and April 6, 2020, and received LUS, CXR, and viral testing for COVID-19 as part of their diagnostic evaluation. The primary objective was to estimate the test characteristics of both LUS B-lines and CXR for the associated diagnosis of COVID-19. Our secondary objective was to evaluate the proportion of patients with COVID-19 that have secondary LUS findings of pleural abnormalities and subpleural consolidations.

RESULTS

We identified 43 patients who underwent both LUS and CXR and were tested for COVID-19. Of these, 27/43 (63%) tested positive. LUS was more sensitive (88.9%, 95% confidence interval (CI), 71.1-97.0) for the associated diagnosis of COVID-19 than CXR (51.9%, 95% CI, 34.0-69.3; p = 0.013). LUS and CXR specificity were 56.3% (95% CI, 33.2-76.9) and 75.0% (95% CI, 50.0-90.3), respectively (p = 0.453). Secondary LUS findings of patients with COVID-19 demonstrated 21/27 (77.8%) had pleural abnormalities and 10/27 (37%) had subpleural consolidations.

CONCLUSION

Among patients who underwent LUS and CXR, LUS was found to have a higher sensitivity than CXR for the evaluation of COVID-19. This data could have important implications as an aid in the diagnostic evaluation of COVID-19, particularly where viral testing is not available or restricted. If generalizable, future directions would include defining how to incorporate LUS into clinical management and its role in screening lower-risk populations.

False Elevation of Volume Determined by Bladder Scanner Secondary to Bowel Obstruction

Bladder scanners allow for quick determination of bladder volumes (BV) with minimal training. BV measured by a machine is generally accurate; however, circumstances exist in which falsely elevated BVs are reported. This case details a patient with a significant small bowel obstruction (SBO) due to superior mesenteric artery syndrome causing a falsely elevated BV. We believe this is the first case report of a SBO causing an elevated BV by bladder scanner. Emergency physicians should be aware of the pitfalls of using bladder scanners, and use their point-of-care ultrasound skills when possible to expand their differential.

Ultrasound vs. Computed Tomography for Severity of Hydronephrosis and Its Importance in Renal Colic

Introduction

Supporting an “ultrasound-first” approach to evaluating renal colic in the emergency department (ED) remains important for improving patient care and decreasing healthcare costs. Our primary objective was to compare emergency physician (EP) ultrasound to computed tomography (CT) detection of hydronephrosis severity in patients with suspected renal colic. We calculated test characteristics of hydronephrosis on EP-performed ultrasound for detecting ureteral stones or ureteral stone size >5mm. We then analyzed the association of hydronephrosis on EP-performed ultrasound, stone size >5mm, and proximal stone location with 30-day events.

Methods

This was a prospective observational study of ED patients with suspected renal colic undergoing CT. Subjects had an EP-performed ultrasound evaluating for the severity of hydronephrosis. A chart review and follow-up phone call was performed.

Results

We enrolled 302 subjects who had an EP-performed ultrasound. CT and EP ultrasound results were comparable in detecting severity of hydronephrosis (x²=51.7, p<0.001). Hydronephrosis on EP-performed ultrasound was predictive of a ureteral stone on CT (PPV 88%; LR+ 2.91), but lack of hydronephrosis did not rule it out (NPV 65%). Lack of hydronephrosis on EP-performed ultrasound makes larger stone size >5mm less likely (NPV 89%; LR− 0.39). Larger stone size > 5mm was associated with 30-day events (OR 2.30, p=0.03).

Conclusion

Using an ultrasound-first approach to detect hydronephrosis may help physicians identify patients with renal colic. The lack of hydronephrosis on ultrasound makes the presence of a larger ureteral stone less likely. Stone size >5mm may be a useful predictor of 30-day events.

Prediction of In-hospital Mortality in Emergency Department Patients With Sepsis: A Local Big Data-Driven, Machine Learning Approach

OBJECTIVES

Predictive analytics in emergency care has mostly been limited to the use of clinical decision rules (CDRs) in the form of simple heuristics and scoring systems. In the development of CDRs, limitations in analytic methods and concerns with usability have generally constrained models to a preselected small set of variables judged to be clinically relevant and to rules that are easily calculated. Furthermore, CDRs frequently suffer from questions of generalizability, take years to develop, and lack the ability to be updated as new information becomes available. Newer analytic and machine learning techniques capable of harnessing the large number of variables that are already available through electronic health records (EHRs) may better predict patient outcomes and facilitate automation and deployment within clinical decision support systems. In this proof-of-concept study, a local, big data-driven, machine learning approach is compared to existing CDRs and traditional analytic methods using the prediction of sepsis in-hospital mortality as the use case.

METHODS

This was a retrospective study of adult ED visits admitted to the hospital meeting criteria for sepsis from October 2013 to October 2014. Sepsis was defined as meeting criteria for systemic inflammatory response syndrome with an infectious admitting diagnosis in the ED. ED visits were randomly partitioned into an 80%/20% split for training and validation. A random forest model (machine learning approach) was constructed using over 500 clinical variables from data available within the EHRs of four hospitals to predict in-hospital mortality. The machine learning prediction model was then compared to a classification and regression tree (CART) model, logistic regression model, and previously developed prediction tools on the validation data set using area under the receiver operating characteristic curve (AUC) and chi-square statistics.

RESULTS

There were 5,278 visits among 4,676 unique patients who met criteria for sepsis. Of the 4,222 patients in the training group, 210 (5.0%) died during hospitalization, and of the 1,056 patients in the validation group, 50 (4.7%) died during hospitalization. The AUCs with 95% confidence intervals (CIs) for the different models were as follows: random forest model, 0.86 (95% CI = 0.82 to 0.90); CART model, 0.69 (95% CI = 0.62 to 0.77); logistic regression model, 0.76 (95% CI = 0.69 to 0.82); CURB-65, 0.73 (95% CI = 0.67 to 0.80); MEDS, 0.71 (95% CI = 0.63 to 0.77); and mREMS, 0.72 (95% CI = 0.65 to 0.79). The random forest model AUC was statistically different from all other models (p ≤ 0.003 for all comparisons).

CONCLUSIONS

In this proof-of-concept study, a local big data-driven, machine learning approach outperformed existing CDRs as well as traditional analytic techniques for predicting in-hospital mortality of ED patients with sepsis. Future research should prospectively evaluate the effectiveness of this approach and whether it translates into improved clinical outcomes for high-risk sepsis patients. The methods developed serve as an example of a new model for predictive analytics in emergency care that can be automated, applied to other clinical outcomes of interest, and deployed in EHRs to enable locally relevant clinical predictions.

Revival of the use of ultrasound in screening for appendicitis in young adult men

PURPOSE

Our primary aim was to evaluate the use of ultrasound (US) as an initial screening test for diagnosing appendicitis in young adult men. Secondary exploratory analyses included the effects of using US for initial screening in these patients, compared with the use of CT, on radiation exposure, length of stay (LOS), and cost of imaging.

METHODS

We retrospectively gathered data from the records of male patients 18-39 years old who had been admitted with appendicitis between June 2006 and September 2011. We investigated the diagnostic tests performed, the patients' characteristics, and the pathologic testing findings and compared the results obtained on US with those obtained on CT.

RESULTS

Of 451 included patients, 86 had undergone US initially. Its sensitivity was only 57% (95% confidence interval, 46-67.6), but its positive predictive value was 98% (95% confidence interval, 93.8-100). The mean LOS was significantly shorter for patients who had undergone US only (214 minutes) than it was for those who had undergone CT only (276 minutes; p < 0.001). We estimated a 57% reduction in CT use and radiation exposure if US were to be performed initially; this would lead to a 45% decrease in imaging costs at our institution.

CONCLUSIONS

Screening US should be considered first for diagnosing appendicitis because of its high positive predictive value, but even if US results are negative for appendicitis, one should not exclude the possible existence of pathology because US has poor sensitivity in this situation. We speculate that the use of screening US can decrease radiation exposure, imaging costs, and LOS.

Emergency physician focused cardiac ultrasound improves diagnosis of ascending aortic dissection

STUDY OBJECTIVE

Ascending aortic dissection (AAD) is an uncommon, time-sensitive, and deadly diagnosis with a nonspecific presentation. Ascending aortic dissection is associated with aortic dilation, which can be determined by emergency physician focused cardiac ultrasound (EP FOCUS). We seek to determine if patients who receive EP FOCUS have reduced time to diagnosis for AAD.

METHODS

We performed a retrospective review of patients treated at 1 of 3 affiliated emergency departments, March 1, 2013, to May 1, 2015, diagnosed as having AAD. All autopsies were reviewed for missed cases. Primary outcome measure was time to diagnosis. Secondary outcomes were time to disposition, misdiagnosis rate, and mortality.

RESULTS

Of 386547 ED visits, targeted review of 123 medical records and 194 autopsy reports identified 32 patients for inclusion. Sixteen patients received EP FOCUS and 16 did not. Median time to diagnosis in the EP FOCUS group was 80 (interquartile range [IQR], 46-157) minutes vs 226 (IQR, 109-1449) minutes in the non-EP FOCUS group (P = .023). Misdiagnosis was 0% (0/16) in the EP FOCUS group vs 43.8% (7/16) in the non-EP FOCUS group (P = .028). Mortality, adjusted for do-not-resuscitate status, for EP FOCUS vs non-EP FOCUS was 15.4% vs 37.5% (P = .24). Median rooming time to disposition was 134 (IQR, 101-195) minutes for EP FOCUS vs 205 (IQR, 114-342) minutes for non-EP FOCUS (P = .27).

CONCLUSIONS

Patients who receive EP FOCUS are diagnosed faster and misdiagnosed less compared with patients who do not receive EP FOCUS. We recommend assessment of the thoracic aorta be performed routinely during cardiac ultrasound in the emergency department.

The "5Es" of emergency physician-performed focused cardiac ultrasound: a protocol for rapid identification of effusion, ejection, equality, exit, and entrance

Emergency physician (EP)-performed focused cardiac ultrasound (EP FOCUS) has been increasingly recognized as a crucial tool to help clinicians diagnose and treat potentially life-threatening conditions. The existing literature demonstrates a variety of EP FOCUS applications and protocols; however, EP FOCUS is not taught, practiced, or interpreted consistently between institutions. Drawing on over 12 years of experience in a large-volume, high-acuity academic emergency department, we have developed a protocol for teaching and performing EP FOCUS known as "The 5Es," where each E represents a specific assessment for immediately relevant clinical information. These include pericardial effusion, qualitative left ventricular ejection, ventricular equality, exit (aortic root diameter), and entrance (inferior vena cava diameter and respirophasic variation). Each of these assessments has been well described in the emergency medicine literature and is within the scope of EP-performed echocardiography. This approach provides a reliable and easily recalled framework for assessing, teaching, and communicating EP FOCUS findings that are essential in caring for the patient in the emergency setting.

Clinical pathologic conference: A 65-year-old male with left-sided chest pain. A case of an unexpected occupational hazard

The authors present a case of a 65-year-old male who presented four times to the emergency department (ED) with left-sided chest pain. On the first three visits, the patient was admitted with a different diagnosis related to his chest pain. On the final visit, an abnormality on an imaging study performed in the ED led to the ultimate diagnostic test revealing the cause of the patient's symptoms. The patient's clinical presentation and ultimate clinical course are summarized, and a discussion of the differential diagnoses of his condition is presented.