Comparison of artificial intelligence versus real-time physician assessment of pulmonary edema with lung ultrasound

Artificial Intelligence in Emergency Medicine: A Primer for the Nonexpert

Artificial intelligence (AI) is increasingly being utilized to augment the practice of emergency medicine due to rapid technological advances and breakthroughs. AI applications have been used to enhance triage systems, predict disease-specific risk, estimate staffing needs, forecast patient decompensation, and interpret imaging findings in the emergency department setting. This article aims to help readers without formal training become informed end-users of AI in emergency medicine. The authors will briefly discuss the principles and key terminology of AI, the reasons for its rising popularity, its potential applications in the emergency department setting, and its limitations. Additionally, resources for further self-studying will also be provided.

The diagnostic performance of automatic B-lines detection for evaluating pulmonary edema in the emergency department among novice point-of-care ultrasound practitioners

PURPOSE

B-lines in lung ultrasound have been a critical clue for detecting pulmonary edema. However, distinguishing B-lines from other artifacts is a challenge, especially for novice point of care ultrasound (POCUS) practitioners. This study aimed to determine the efficacy of automatic detection of B-lines using artificial intelligence (Auto B-lines) for detecting pulmonary edema.

METHODS

A retrospective study was conducted on dyspnea patients treated at the emergency department between January 2023 and June 2024. Ultrasound documentation and electronic emergency department medical records were evaluated for sensitivity, specificity, positive likelihood ratio, and negative likelihood ratio of auto B-lines in detection of pulmonary edema.

RESULTS

Sixty-six patients with a final diagnosis of pulmonary edema were enrolled, with 54.68% having positive B-lines in lung ultrasound. Auto B-lines had 95.6% sensitivity (95% confidence interval [CI]: 0.92-0.98) and 77.2% specificity (95% CI: 0.74-0.80). Physicians demonstrated 82.7% sensitivity (95% CI: 0.79-0.97) and 63.09% sensitivity (95% CI: 0.58-0.69).

CONCLUSION

The auto B-lines were highly sensitive in diagnosing pulmonary edema in novice POCUS practitioners. The clinical integration of physicians and artificial intelligence enhances diagnostic capabilities.

[Cardiac resuscitation-associated lung edema (CRALE): evaluation of diagnostic and therapeutic approaches by an expert group of the German Cardiac Society]

Sudden cardiac arrest (CA) is one of the leading causes of death in Europe, with over 70,000 cases annually in Germany. This study aims to evaluate diagnostic and therapeutic approaches for pulmonary edema in the post-resuscitation phase among intensive care physicians in Germany. Methods: The Working Group on Cardiopulmonary Resuscitation (AG42) of the German Cardiac Society conducted a web-based survey among its members. The survey assessed diagnostic methods, therapeutic strategies, and risk factors related to pulmonary edema after resuscitation. Results: A total of 77 participants, with a mean age of 43.9 years (±9.6), took part in the survey. Among them, 54.5% had more than 10 years of clinical experience in acute and intensive care medicine. Most clinics have access to radiological and sonographic procedures as well as advanced hemodynamic monitoring. Diagnostic measures are predominantly performed immediately upon admission (49.4%) or within one hour (36.4%) and are typically monitored every eight hours (77.9%). The oxygenation index (paO2/FiO2) is used by 64.9% to assess the severity of pulmonary edema, followed by qualitative evaluation of chest X-rays (46.8%) and B-line scoring via lung ultrasound (33.8%). Therapeutic approaches focus on optimizing ventilation parameters, hemodynamic management, and the use of loop diuretics. To prevent pulmonary edema, participants favor a differentiated therapy involving volume and vasoactive agents, guided by invasive hemodynamic measurements. Conclusion: Pulmonary edema, alongside cardiac and cerebral dysfunctions, represents a significant challenge in managing post-resuscitation syndromes. The survey results reveal substantial variability in diagnostic and therapeutic approaches. Prospective studies are needed to better understand the complex pathological mechanisms and develop standardized protocols.

Automatic B-lines: a tool for minimizing time to diuretic administration in pulmonary edema patients in the emergency department of a developing country

BACKGROUND

Effective management of pulmonary edema in the emergency department (ED) is crucial given its significant global impact on health. This study aimed to investigate the hypothesis: "Does the utilization of Automatic B-lines via ultrasonography in patients with pulmonary edema facilitate faster diuretic administration in a developing country?"

METHODS

This retrospective observational study was conducted at a tertiary academic center in Thailand. Patients with pulmonary edema admitted to the ED between January 2023 and June 2024 were enrolled. Ultrasound documentation and electronic ED medical records were compared to assess the time of diuretic administration between patients who had lung ultrasounds utilizing automatic B-lines and those who had manual B-lines counted by physician eye inspection. Multivariate logistic regression was employed to examine the relationship between the use of automatic B-lines and early diuretic administration.

RESULTS

The study included 134 patients with pulmonary edema. The time to diuretic administration was significantly shorter in the automatic B-lines group (median time [Q1-Q3], 55 min; range, 35-110 min) compared to the non-automatic B-lines group (median time, 100 min; range, 75-145 min). In the multivariable logistic regression analysis, early diuretic administration within 60 min of triage was significantly more likely in the automatic B-lines group (adjusted odds ratio, 1.45; 95% confidence interval, 1.10-2.45) than in the non-automatic B-lines group.

CONCLUSIONS

In a developing country, patients with pulmonary edema who had lung ultrasound evaluation with automated B lines experienced a fastest diuresis compared to those who utilized ultrasonography without automatic B lines. Implementing automatic B-lines as an early screening protocol could enhance clinical practice in the ED.

Diagnostic accuracy of artificial intelligence for identifying systolic and diastolic cardiac dysfunction in the emergency department

INTRODUCTION

Cardiac point-of-care ultrasound (POCUS) can evaluate for systolic and diastolic dysfunction to inform care in the Emergency Department (ED). However, accurate assessment can be limited by user experience. Artificial intelligence (AI) has been proposed as a model to increase the accuracy of cardiac POCUS. However, there is limited evidence of the accuracy of AI in the clinical environment. The objective of this study was to determine the diagnostic accuracy of AI for identifying systolic and diastolic dysfunction compared with expert reviewers.

METHODS

This was a prospective, observational study of adult ED patients aged ≥45 years with risk factors for systolic and diastolic dysfunction. Ultrasound fellowship-trained physicians used an ultrasound machine with existing AI software and obtained parasternal long axis, parasternal short axis, and apical 4-chamber views of the heart. Systolic dysfunction was defined as ejection fraction (EF) < 50 % in at least two views using visual assessment or E-point septal separation >10 mm. Diastolic dysfunction was defined as an E:A < 0.8, or ≥ 2 of the following: septal e' < 7 cm/s or lateral e' < 10 cm/s, E:e' > 14, or left atrial volume > 34 mL/m2. AI was subsequently used to measure EF, E, A, septal e', and lateral e' velocities. The gold standard was systolic or diastolic dysfunction as assessed by two independent physicians with discordance resolved via consensus. We performed descriptive statistics (mean ± standard deviation) and calculated the sensitivity, specificity, positive likelihood ratio (LR+), and negative likelihood ratio (LR-) of the AI in determining systolic and diastolic dysfunction with 95 % confidence interval (CI). Subgroup analyses were performed by body mass index (BMI).

RESULTS

We enrolled 220 patients, with 11 being excluded due to inadequate images, resulting in 209 patients being included in the study. Mean age was 60 ± 9 years, 51.7 % were women, and the mean BMI was 31 ± 8.1 mg/kg2. For assessing systolic dysfunction, AI was 85.7 % (95 %CI 57.2 % to 98.2 %) sensitive and 94.8 % (95 %CI 90.6 % to 97.5 %) specific with a LR+ of 16.4 (95 %CI 8.6 to 31.1) and LR- of 0.15 (95 % CI 0.04 to 0.54). For assessing diastolic dysfunction, AI was 91.9 % (95 %CI 85.6 % to 96.0 %) sensitive and 94.2 % (95 %CI 87.0 % to 98.1 %) specific with a LR+ of 15.8 (95 %CI 6.7 to 37.1) and a LR- of 0.09 (0.05 to 0.16). When analyzed by BMI, results were similar except for lower sensitivity in the BMI ≥ 30 vs BMI < 30 (100 % vs 80 %).

CONCLUSION

When compared with expert assessment, AI had high sensitivity and specificity for diagnosing both systolic and diastolic dysfunction.

Artificial intelligence application for identifying toxic plant species: A case of poisoning with Datura stramonium

INTRODUCTION

The management of plant poisonings in the emergency department (ED) presents various challenges. Foremost among these is the identification of the specific botanical species responsible for the toxic effect. In cases of plant poisoning, it is crucial to accurately identify the plant in order to promptly evaluate if it has cardiotoxic, neurotoxic, hepatotoxic, or anticholinergic properties. Furthermore, it is typically not possible to determine the identity of these plants through blood tests conducted in the ED.

CASE REPORT

An otherwise healthy 23-year-old male patient presented to the ED with symptoms of restlessness, altered mental state, and hallucinations that occurred 2 h after consuming herbal tea. On physical examination, he was tachypneic, tachycardic, and disoriented. The pupils were bilaterally mydriatic. The patient's symptoms were consistent with both sympathomimetic and anticholinergic (antimuscarinic) toxidromes. We were unable to promptly reach a botanist to identify the plant to which the patient had been exposed. Therefore, we employed Google Gemini, an artificial intelligence software, to ascertain the plant's identity. Google Gemini identified the plant we photographed as Datura stramonium, commonly known as jimson weed, which is known to cause anticholinergic toxicity. The botanist we contacted later confirmed that the plant was D. stramonium. The patient's symptoms were alleviated with the use of intravenous diazepam and physostigmine.

CONCLUSION

We propose that the utilization of artificial intelligence applications with visual recognition capabilities could be beneficial for physicians, patients, and foragers of edible wild plants to accurately identify plants and distinguish toxic species.

Thoracic Ultrasound in Others Scenarios: An Expanding Tool

Modern management of thoracic disease is dominated by ultrasound assessment with strong evidence supporting its use in many clinical settings, providing both diagnostic and procedural. Thoracic ultrasound is a pivotal step in the management of chronic lung disease and pulmonary vascular disease, in early assessment as in therapeutic monitoring. Development and validation of novel ultrasound biomarkers of activity and prognostic, especially those linked to advanced ultrasound techniques, are expected in the coming years. Assessing and treating respiratory muscle dysfunction is crucial for patients with both acute and chronic respiratory failure. To explore novel techniques, including imaging with ultrasound is important. Artificial intelligence (AI) excels at automatically recognizing complex patterns and providing quantitative assessment for imaging data, showing high potential to assist physicians in acquiring more accurate and reproducible results. Finally, a training system with structured proficiency and competency standards, about the use of TU is necessary. We offer our perspective on the challenges and opportunities for the clinical practice in other scenarios.

Interpretation of Artificial Intelligence Models in Healthcare: A Pictorial Guide for Clinicians

Artificial intelligence (AI) models can play a more effective role in managing patients with the explosion of digital health records available in the healthcare industry. Machine-learning (ML) and deep-learning (DL) techniques are two methods used to develop predictive models that serve to improve the clinical processes in the healthcare industry. These models are also implemented in medical imaging machines to empower them with an intelligent decision system to aid physicians in their decisions and increase the efficiency of their routine clinical practices. The physicians who are going to work with these machines need to have an insight into what happens in the background of the implemented models and how they work. More importantly, they need to be able to interpret their predictions, assess their performance, and compare them to find the one with the best performance and fewer errors. This review aims to provide an accessible overview of key evaluation metrics for physicians without AI expertise. In this review, we developed four real-world diagnostic AI models (two ML and two DL models) for breast cancer diagnosis using ultrasound images. Then, 23 of the most commonly used evaluation metrics were reviewed uncomplicatedly for physicians. Finally, all metrics were calculated and used practically to interpret and evaluate the outputs of the models. Accessible explanations and practical applications empower physicians to effectively interpret, evaluate, and optimize AI models to ensure safety and efficacy when integrated into clinical practice.

Artificial intelligence and point-of-care ultrasound: Benefits, limitations, and implications for the future

The utilization of artificial intelligence (AI) in medical imaging has become a rapidly growing field as a means to address contemporary demands and challenges of healthcare. Among the emerging applications of AI is point-of-care ultrasound (POCUS), in which the combination of these two technologies has garnered recent attention in research and clinical settings. In this Controversies paper, we will discuss the benefits, limitations, and future considerations of AI in POCUS for patients, clinicians, and healthcare systems.

Results of the implementation of a double-check protocol with point-of-care ultrasound for acute heart failure in the emergency department

OBJECTIVE

To determine the effectiveness of a double-check protocol using Point-of-Care Ultrasound in the management of patients diagnosed with Acute Heart Failure in an Emergency Department.

METHOD

Prospective analytical cross-sectional observational study with patients diagnosed with Acute Heart Failure by the outgoing medical team, who undergo multi-organ ultrasound evaluation including cardiac, pulmonary, and inferior vena cava ultrasound.

RESULTS

96 patients were included. An alternative diagnosis was found in 33% of them. Among the 77% where AHF diagnosis was confirmed, 73.4% had an underlying cause or condition not previously known (Left Ventricular Ejection Fraction less than 40% or moderate-severe valvulopathy). The introduction of the protocol had a clinically relevant impact on 47% of all included patients.

CONCLUSIONS

The implementation of a double-check protocol using POCUS, including cardiac, pulmonary, and inferior vena cava assessment in patients diagnosed with Acute Heart Failure, demonstrates a high utility in ensuring accurate diagnosis and proper classification of these patients.

A Novel Automatic Algorithm to Support Lung Ultrasound Non-Expert Physicians in Interstitial Pneumonia Evaluation: A Single-Center Study

INTRODUCTION

Lung ultrasound (LUS) is widely used in clinical practice for identifying interstitial lung diseases (ILDs) and assessing their progression. Although high-resolution computed tomography (HRCT) remains the gold standard for evaluating the severity of ILDs, LUS can be performed as a screening method or as a follow-up tool post-HRCT. Minimum training is needed to better identify typical lesions, and the integration of innovative artificial intelligence (AI) automatic algorithms may enhance diagnostic efficiency.

AIM

This study aims to assess the effectiveness of a novel AI algorithm in automatic ILD recognition and scoring in comparison to an expert LUS sonographer. The "SensUS Lung" device, equipped with an automatic algorithm, was employed for the automatic recognition of the typical ILD patterns and to calculate an index grading of the interstitial involvement.

METHODS

We selected 33 Caucasian patients in follow-up for ILDs exhibiting typical HRCT patterns (honeycombing, ground glass, fibrosis). An expert physician evaluated all patients with LUS on twelve segments (six per side). Next, blinded to the previous evaluation, an untrained operator, a non-expert in LUS, performed the exam with the SensUS device equipped with the automatic algorithm ("SensUS Lung") using the same protocol. Pulmonary functional tests (PFT) and DLCO were conducted for all patients, categorizing them as having reduced or preserved DLCO. The SensUS device indicated different grades of interstitial involvement named Lung Staging that were scored from 0 (absent) to 4 (peak), which was compared to the Lung Ultrasound Score (LUS score) by dividing it by the number of segments evaluated. Statistical analyses were done with Wilcoxon tests for paired values or Mann-Whitney for unpaired samples, and correlations were performed using Spearman analysis; p < 0.05 was considered significant.

RESULTS

Lung Staging was non-inferior to LUS score in identifying the risk of ILDs (median SensUS 1 [0-2] vs. LUS 0.67 [0.25-1.54]; p = 0.84). Furthermore, the grade of interstitial pulmonary involvement detected with the SensUS device is directly related to the LUS score (r = 0.607, p = 0.002). Lung Staging values were inversely correlated with forced expiratory volume at first second (FEV1%, r = -0.40, p = 0.027), forced vital capacity (FVC%, r = -0.39, p = 0.03) and forced expiratory flow (FEF) at 25th percentile (FEF25%, r = -0.39, p = 0.02) while results directly correlated with FEF25-75% (r = 0.45, p = 0.04) and FEF75% (r = 0.43, p = 0.01). Finally, in patients with reduced DLCO, the Lung Staging was significantly higher, overlapping the LUS (reduced median 1 [1-2] vs. preserved 0 [0-1], p = 0.001), and overlapping the LUS (reduced median 18 [4-20] vs. preserved 5.5 [2-9], p = 0.035).

CONCLUSIONS

Our data suggest that the considered AI automatic algorithm may assist non-expert physicians in LUS, resulting in non-inferior-to-expert LUS despite a tendency to overestimate ILD lesions. Therefore, the AI algorithm has the potential to support physicians, particularly non-expert LUS sonographers, in daily clinical practice to monitor patients with ILDs. The adopted device is user-friendly, offering a fully automatic real-time analysis. However, it needs proper training in basic skills.

Building and validating an artificial intelligence model to identify tracheobronchopathia osteochondroplastica by using bronchoscopic images

BACKGROUND

Given the rarity of tracheobronchopathia osteochondroplastica (TO), many young doctors in primary hospitals are unable to identify TO based on bronchoscopy findings.

OBJECTIVES

To build an artificial intelligence (AI) model for differentiating TO from other multinodular airway diseases by using bronchoscopic images.

DESIGN

We designed the study by comparing the imaging data of patients undergoing bronchoscopy from January 2010 to October 2022 by using EfficientNet. Bronchoscopic images of 21 patients with TO at Anhui Chest Hospital from October 2019 to October 2022 were collected for external validation.

METHODS

Bronchoscopic images of patients with multinodular airway lesions (including TO, amyloidosis, tumors, and inflammation) and without airway lesions in the First Affiliated Hospital of Guangzhou Medical University were collected. The images were randomized (4:1) into training and validation groups based on different diseases and utilized for deep learning by convolutional neural networks (CNNs).

RESULTS

We enrolled 201 patients with multinodular airway disease (38, 15, 75, and 73 patients with TO, amyloidosis, tumors, and inflammation, respectively) and 213 without any airway lesions. To find multinodular lesion images for deep learning, we utilized 2183 bronchoscopic images of multinodular lesions (including TO, amyloidosis, tumor, and inflammation) and compared them with images without any airway lesions (1733). The accuracy of multinodular lesion identification was 98.9%. Further, the accuracy of TO detection based on the bronchoscopic images of multinodular lesions was 89.2%. Regarding external validation (using images from 21 patients with TO), all patients could be diagnosed with TO; the accuracy was 89.8%.

CONCLUSION

We built an AI model that could differentiate TO from other multinodular airway diseases (mainly amyloidosis, tumors, and inflammation) by using bronchoscopic images. The model could help young physicians identify this rare airway disease.