BACKGROUND:Deep learning (DL) shows promise for interpretation of lung ultrasound (LUS) images in humans, but its performance in animals remains underexplored. HYPOTHESIS/OBJECTIVES:Assess performance of a B-line detection algorithm (BLDA) trained on LUS images from humans when applied to images from dogs. ANIMALS:A total of 1,950 clips collected from 201 LUS examinations in 90 dogs across 4 studies. METHODS:Ultrasound clips were collated and labeled with A-line or B-line profiles by an expert reviewer. A DL model previously trained on LUS images from humans was applied to detect presence of B-lines at a framewise level. A clip classification algorithm was calibrated to maximize clip-level performance of the BLDA using a calibration data set. Performance of the DL model and BLDA was assessed on a held-out test set of LUS images. A heatmap-based explainability method was used to visualize regions most utilized by the model for predictions. RESULTS:When applied to the test set, the BLDA showed an overall accuracy of 87%, with sensitivity of 73% and specificity of 93%. The algorithm performed best when identifying images with strong (versus weak) B-line profiles. Assessing raw model predictions, zero-shot performance for identification of B-lines was excellent (area under the curve, 0.96). Heatmaps suggested that the DL model utilized image areas that were plausibly relevant for LUS interpretation. CONCLUSIONS AND CLINICAL IMPORTANCE:A LUS model trained on images from humans maintained strong performance when applied to dogs, supporting cross-species generalization to accelerate veterinary diagnostic innovations.

INTRODUCTION:Long-duration space missions demand reliable, portable, and autonomous medical diagnostic tools. Lung ultrasound (LUS) is ideal for space exploration due to its safety and versatility, but interpreting LUS images requires training. Artificial intelligence (AI) can assist with image interpretation by detecting lung sliding, an ultrasound sign produced by the movement of lung and chest wall during breathing. The presence of lung sliding helps to rule out pneumothorax, or a collapsed lung, which is a condition that may arise from barotrauma or rapid changes in pressurization. METHODS:LUS clips were acquired from two healthy volunteers during parabolic flight maneuvers simulating microgravity and lunar gravity, with +1-G clips used as controls. Clips were analyzed using an AI model to classify the presence of lung sliding and model performance and confidence were compared across gravity conditions. All clips were reviewed by an expert to establish whether lung sliding was present, which served as the reference standard for AI model evaluation. RESULTS:From 105 LUS clips, the model demonstrated an accuracy of 94%, with similar performance at +1 G (96%), lunar gravity (94%), and microgravity (92%). Prediction confidence varied by gravity, with median values of 93% at +1 G, 83% at lunar gravity, and 67% at microgravity. Confidence was significantly lower in microgravity compared with +1 G. DISCUSSION:These findings demonstrate that AI-assisted LUS can reliably detect lung sliding under reduced gravity, supporting its feasibility as an autonomous diagnostic support tool for spaceflight and highlighting the importance of further validation in microgravity environments. Côté M, Smith D, Orozco N, VanBerlo B, Huggard B, Arntfield R, Prager R. Artificial intelligence interpretation of point-of-care lung ultrasound in microgravity. Aerosp Med Hum Perform. 2026; 97(7):505-509.

Artificial intelligence (AI) is increasingly integrated into point-of-care ultrasound (POCUS) to enhance its utility in critical care settings. This manuscript explores the current state of AI applications in POCUS, focusing on key domains such as image acquisition, image interpretation, education, task automation, procedural guidance, program development, and quality assurance. AI-driven tools can potentially improve image quality, provide real-time feedback, and assist in the interpretation of ultrasound images, thereby democratizing the use of POCUS across varying levels of operator expertise. This narrative review highlights relevant studies demonstrating the clinical utility of AI in POCUS, discusses the challenges that remain, and provides insights into future developments. The goal is to equip intensivists with a comprehensive understanding of how AI can support POCUS practice today and what advancements are on the horizon.

Background:Pneumothorax (PTX) is a frequent complication after chest tube removal, and timely detection is important to inform monitoring and potential intervention. Chest radiograph (CXR) remains the standard modality after chest-tube-removal PTX detection, despite its limited sensitivity and frequent delays in acquisition. Lung ultrasound (LUS) has superior accuracy and portability but is highly operator dependent, limiting its usability. The objective of this study is to evaluate whether artificial intelligence-assisted LUS (AI-LUS) enables novice users to accurately detect PTX after chest tube removal, compared with expert interpretation and CXR. Research Question:Does AI-LUS improve the ability of novice users to detect findings associated with PTX after chest tube removal compared with expert interpretation and CXR? Study Design and Methods:We conducted a prospective diagnostic accuracy study in adult patients undergoing chest tube removal at a tertiary academic hospital. LUS clips were acquired by novice operators using a handheld ultrasound device. A previously trained artificial intelligence model was then calibrated and used to detect the absence or presence of lung sliding. The reference standard was expert consensus LUS interpretation, with CXR serving as a secondary reference standard. Sensitivity and specificity were calculated at 2 time points: immediately after removal and after routine CXR. Results:A total of 76 patients were enrolled, yielding 848 LUS clips across 2 time points. Data from the first 12 patients were used to calibrate the model, with the remaining 64 forming the validation cohort. Compared with expert LUS interpretation, AI-LUS demonstrated a sensitivity of 0.775, a specificity of 0.831, and a negative predictive value of 0.96 for identifying absent lung sliding. When compared with CXR, AI-LUS achieved a sensitivity of 1.0 immediately after removal and 0.923 after CXR for PTX detection. Interpretation:Our results show that novice-performed AI-LUS demonstrated moderate diagnostic accuracy for detecting absent lung sliding after chest tube removal. Its very high sensitivity and excellent negative predictive value for identifying cases with absent lung sliding associated with PTX relative to CXR highlights a potential role for AI-LUS as a rapid triage tool that may reduce reliance on routine CXR, while acknowledging that PTX inference requires clinical correlation and additional ultrasound findings.

BACKGROUND:Artificial intelligence (AI) has the potential to address training limitations and inter-operator variability that constrain the use of lung ultrasound (LUS) in austere and prehospital settings. This pilot study evaluated whether AI-based decision support could improve the diagnostic accuracy and confidence of United States Marine Corps Corpsmen in identifying absent lung sliding, a key indicator of pneumothorax, during LUS interpretation. METHODS:This pilot-prospective multi-reader, multi-case study involved five military medics, all novices in point-of-care ultrasound, each interpreting 50 de-identified LUS video clips twice, once without AI assistance (control) and once with AI assistance (ATLAS, Deep Breathe Inc., London, Canada), in randomized order with at least a 2-hour washout between sessions. Expert consensus served as a reference standard. Diagnostic performance was assessed using area under the receiver operating characteristic curve (AUROC), sensitivity, specificity, and accuracy. Differences were analyzed using the Random-Reader Random-Case method. Per-clip reader confidence ratings were compared using the Stuart-Maxwell test. RESULTS:AI assistance significantly improved diagnostic performance across all measured outcomes. The mean AUROC increased from 0.72 (SD 0.16) without AI to 0.93 (SD 0.04) with AI (P=.03). Sensitivity rose from 0.63 (SD 0.14) to 0.90 (SD 0.09), specificity from 0.70 (SD 0.15) to 0.86 (SD 0.10), and overall accuracy from 0.67 (SD 0.10) to 0.88 (0.06) (McNemar's test, P<.001). Reader confidence also improved, with high-confidence ratings nearly doubling from 20% to 37%, and low-confidence ratings decreasing from 38% to 33%. These distributional changes were statistically significant (Stuart-Maxwell χ², P<.001). CONCLUSION:AI support markedly improved the diagnostic accuracy and confidence of novice LUS interpretation for detecting absent lung sliding. These findings suggest that real-time AI-based decision support may help improve access to high-quality LUS in military and other resource-limited care settings.

PURPOSE OF REVIEW:Cardiopulmonary monitoring is fundamental to critical care, yet traditional approaches rely on simplified thresholds that capture only a fraction of the rich information contained within waveforms, imaging, and continuous physiological data. This review examines emerging applications of artificial intelligence (AI) and machine learning (ML) that enhance waveform interpretation, automate point-of-care ultrasound (POCUS), enable predictive monitoring, and extend advanced assessment capabilities into low-resource settings. RECENT FINDINGS:AI models now identify deterioration earlier than conventional tools, derive complex hemodynamic variables from noninvasive signals, and predict events such as hypotension, cardiac arrest, and sepsis hours in advance. In POCUS, AI enables real-time acquisition guidance and automated cardiac and pulmonary interpretation, allowing novice users to obtain expert-quality studies. Cloud and edge-based architectures further support AI-driven monitoring in austere environments. Despite these advances, most AI systems remain in early development; fewer than 2% have undergone clinical integration, and challenges persist related to generalizability, bias, heterogeneous data quality, and limited prospective evaluation. SUMMARY:AI-assisted cardiopulmonary monitoring has the potential to transition critical care from reactive assessment to dynamic, anticipatory management. Realizing this promise will require rigorous validation, workflow integration, and evidence demonstrating true clinical benefit.

In intensive care, hemodynamic and respiratory failure are the two main types of organ failure and are frequently associated. In cases of shock and/or acute dyspnea, thoracic ultrasound involves the simultaneous use of lung ultrasound and transthoracic echocardiography. While rigorous clinical assessment remains fundamental, the current literature shows that adding thoracic ultrasound to conventional examination improves diagnostic accuracy and enables rapid, adequate therapeutic decisions during shock and/or acute respiratory failure in a non-invasive way. However, its impact on survival is still being debated. This narrative review summarizes the criteria that help identify the five main hemodynamic profiles during shock, as well as the endpoints that allow discrimination between cardiogenic pulmonary edema, acute respiratory distress syndrome, pneumonia, pleural effusion and pneumothorax. It also proposes pragmatic algorithms for managing shock and/or respiratory failure at the bedside. Current data suggest that artificial intelligence and automated measurements will greatly improve the diagnostic performance, reproducibility, ease of use, and teachability of thoracic ultrasound in the near future.

Rationale Lung ultrasound, the most precise diagnostic tool for pleural effusions, is underutilized due to healthcare providers' limited proficiency. To address this, deep learning models can be trained to recognize pleural effusions. However, current models lack the ability to diagnose effusions in diverse clinical contexts, which presents significant challenges. Objective To develop and validate a deep learning model for detecting pleural effusions in lung ultrasound images, with adaptable performance characteristics tailored to specific clinical scenarios. Methods A retrospective study was conducted at two Canadian tertiary hospitals to evaluate the detection of pleural effusions of varying sizes and complexities using lung ultrasound. A deep learning model incorporating a frame-level convolutional neural network and a clip-level prediction algorithm was developed and validated against expert annotations. Results The model was evaluated using a holdout dataset of 103 lung ultrasound clips from 46 patients with pleural effusion and 136 clips from 83 patients without effusion. The general model achieved a sensitivity of 0.90 for small-to-large effusions, with a specificity of 0.89. The large effusion model demonstrated a sensitivity of 0.97 for large effusions while maintaining a specificity of 0.90. The trauma model showed high sensitivity to all effusions, including trace (0.91) and small (0.97) effusions. Conclusion Our research highlights the development of a deep learning model that effectively detects pleural effusions of varying sizes and complexities on lung ultrasound in different clinical settings. This tool has the potential to enhance emergency physicians' ability to quickly and accurately diagnose effusions, particularly in time-sensitive situations.

Background: Chest point of care ultrasound (POCUS) is a first-line diagnostic test to identify lung sliding, an important artifact to diagnose or rule out pneumothorax. Despite enthusiastic adoption of this modality, the interrater reliability for physicians to identify lung sliding is unknown. Additionally, the relative diagnostic performance of physicians interpreting B-mode and M-mode ultrasound is unclear. We sought to determine the interrater reliability of physicians to detect lung sliding on B-mode and M-mode POCUS. Methods: We performed a cross-sectional interrater agreement study surveying acute care physicians on their interpretation of 20 B-mode and M-mode POCUS clips. Two experienced clinicians determined the reference standard diagnosis. Respondents reported their interpretation of each POCUS B-mode clip or M-mode image. The primary outcome was the interrater agreement, determined by an intra-class correlation coefficient (ICC). Results: From September to November 2023, there were 20 survey respondents. Fourteen (70%) respondents were resident physicians. Respondents were confident or very confident in their skill performing chest POCUS in 14 (70%) cases, with 19 (90%) performing chest POCUS every week or more frequently. The ICC on B-mode was 0.44 and for M-mode was 0.43, indicating moderate agreement. There were no significant differences in interrater reliability between subgroups of confidence or experience. Conclusion: There is only moderate interrater reliability between clinicians to diagnose lung sliding. Clinicians have superior accuracy on B-mode compared to M-mode clips.

OBJECTIVE:To determine the impact of using dynamic measures of fluid responsiveness in guiding the resuscitation of adult patients with sepsis and septic shock. DATA SOURCE:We searched MEDLINE, Embase, and unpublished sources from inception to February 3, 2025. STUDY SELECTION:We included randomized controlled trials (RCTs) that evaluated the use of dynamic measures of fluid responsiveness to guide resuscitation compared with any other method in patients with sepsis and septic shock. DATA EXTRACTION:We collected data regarding study and patient characteristics, definitions of fluid responsiveness, modality for assessing fluid responsiveness, and outcome data. We performed a random-effects meta-analysis and rated the certainty of the evidence using the Grading of Recommendations Assessment, Development, and Evaluation framework. DATA SYNTHESIS:We included nine eligible RCTs (n = 698 patients). The use of dynamic measures of fluid responsiveness to guide IV fluid (IVF) administration of patients with septic shock probably reduces 28-day mortality (relative risk] 0.61; 95% CI, 0.42-0.90, moderate certainty), may reduce the risk of acute kidney injury (AKI) (RR 0.66; 95% CI, 0.44-0.98, low certainty), and cumulative fluid balance on day 3 (mean difference -1.57L; 95% CI, -2.44 L to -0.69 L, low certainty). The use of dynamic measures of fluid responsiveness has an uncertain effect on ICU mortality, ICU and hospital length of stay, need for and duration of mechanical ventilation, need for renal replacement therapy, vasoactive medication administration, duration of vasopressor use, and IVF administration on day 1. CONCLUSIONS:In adult patients with sepsis and septic shock, using dynamic measures of fluid responsiveness may improve survival and reduce the risk of AKI. Future studies should evaluate the impact of this intervention on other important clinical outcomes and determine the comparative efficacy of specific modalities for assessing fluid responsiveness.

Pneumothorax (PTX) is an acute respiratory condition in which air accumulates between the lung and chest wall. PTX can become life threatening if not identified and treated; therefore, early detection and prompt intervention are critical for patients at risk. Lung ultrasound (LUS) is a real-time, portable imaging modality suitable for bedside use. In contrast to chest X-ray for PTX detection, LUS does not use ionizing radiation and has demonstrated superior diagnostic performance. However, the operator dependence of handheld ultrasound probes limits the feasibility of continuous PTX monitoring which would enable early detection of PTX and be particularly beneficial for patients at ongoing risk, such as those under mechanical ventilation or aeromedical transport. We have developed a flexible, lightweight, and thin wearable ultrasonic sensor (WUS) for motion mode (M-mode) ultrasound image acquisitions. The WUS consists of a single-element ultrasonic transducer made of a polyvinylidene difluoride piezoelectric polymer film. Its simple fabrication and component materials make it low-cost and suitable for disposable use. The WUS is well-suited for long-term, continuous monitoring due to its hands-free operation. In this study, we demonstrate the feasibility of the WUS for PTX detection by using M-mode image features with lung tissue-mimicking phantoms and in-vivo human subjects. [Work supported by the Natural Sciences and Engineering Research Council of Canada.]

OBJECTIVES:To determine the impact of short-acting beta-blocker therapy on outcomes in adult patients with septic shock. DATA SOURCES:We searched MEDLINE, Embase, and unpublished sources from inception to April 19, 2024. STUDY SELECTION:We included randomized controlled trials (RCTs) that evaluated short-acting beta-blockers compared with usual care in patients with septic shock. DATA EXTRACTION:We collected data regarding study and patient characteristics, beta-blocker administration, and clinical, hemodynamic, and biomarker outcomes. DATA SYNTHESIS:Twelve RCTs proved eligible ( n = 1170 patients). Short-acting beta-blockers may reduce 28-day mortality (relative risk [RR], 0.76; 95% CI, 0.62-0.93; low certainty) and probably reduce new-onset tachyarrhythmias (RR, 0.37; 95% CI, 0.18-0.78; moderate certainty) but may increase the duration of vasopressors (mean difference [MD], 1.04 d; 95% CI, 0.37-1.72; low certainty). Furthermore, there is an uncertain effect as to whether short-acting beta blockers impact 90-day mortality (RR, 0.98; 95% CI, 0.73-1.31), ICU length of stay (MD, -0.75 d; 95% CI, -3.43 to 1.93 d), hospital length of stay (MD, 1.03 d; 95% CI, -1.92 to 3.98 d), duration of mechanical ventilation (MD, -0.10 d; 95% CI, -1.25 to 1.05 d) (all very low certainty), bradycardia episodes (RR, 3.14; 95% CI, 0.91-14.01), and hypotension episodes (RR, 4.74; 95% CI, 1.62-14.01) (all very low certainty). CONCLUSIONS:In patients with septic shock, short-acting beta-blockers may improve survival and reduce new-onset tachyarrhythmias. However, these findings were based on low certainty evidence and given ongoing concerns regarding adverse effects and the increase duration of vasopressor use, we need larger and more rigorous RCTs to evaluate this intervention.

We sought to conduct a systematic review to determine the diagnostic test accuracy of point-of-care ultrasound (POCUS) for the specific etiologies and subtypes of shock. We searched MEDLINE, Embase, and the grey literature for prospective studies in adult populations with shock. We collected data on study design, patient characteristics, operator characteristics, POCUS protocol, and true and false positives and negatives, and assessed the risk of bias. We found 18 eligible studies with a total of N = 2,088 patients. The pooled sensitivity and specificity of POCUS for determining shock subtype were 90 CRD42020160001 ); first submitted 1 December 2019.

BACKGROUND:Persistently increasing healthcare spending, paired with growing healthcare demand, highlights the need to identify mechanisms for cost savings. Chest radiography (CXR) is commonly performed following intrathoracic procedures to rule out pneumothorax (PTX) even if the clinical pretest probability is low. However, lung ultrasound (LUS) is known to have superior sensitivity, possibly representing a promising cost-saving tool. In response, we conducted an economic analysis comparing LUS and CXR to exclude PTX after invasive intrathoracic procedures. METHODS:A retrospective review of the radiology case-costing center was performed at an academic cardiothoracic surgical institution to identify the activity and cost of CXRs performed to rule out PTX following intrathoracic procedures. This cost was then compared to the theoretical cost of LUS. RESULTS:CXRs performed to rule out iatrogenic PTX were common with 22,274 radiographs completed and were economically burdensome, with an associated cost of $1.4 million. Portable CXR cost $75.46 per test, while CXR posteroanterior/lateral costs $41.64. Comparatively, LUS cost $38.38. Implementation would lead to cost savings of $559,537.10 or $41.58, on average, per patient. CONCLUSION:Given the superiority of LUS in terms of sensitivity and accuracy for PTX diagnosis, these findings underscore the compelling rationale for its broader integration into clinical practice.

Objective:This is a pilot study to determine the feasibility of a multicentre stepped wedge cluster randomized trial of implementing the 2013 World Society of the Intraabdominal Compartment Syndrome (WSACS) guidelines as an intervention to treat intraabdominal hypertension (IAH) and abdominal compartment syndrome (ACS) in critically ill patients. Design:Single-centre before-and-after trial, with an observation / baseline period of 3 months followed by a 9-month intervention period. Setting:A 35 bed medical-surgical-trauma intensive care unit in a tertiary level, Canadian hospital. Patients:Recruitment from consecutively admitted adult intensive care unit patients. Intervention:In the intervention period, treatment teams were prompted to implement WSACS interventions in all patients diagnosed with IAH. Measurements and Main Results:129 patients were recruited, 59 during the observation period and 70 during the intervention period. Only 17.0% and 12.9%, respectively, met diagnostic criteria for IAH. Many recruited patients did not have intraabdominal pressures measured regularly per study protocol. There was no difference in ICU mortality for patients in either cohort or between those with and without IAH. Conclusions:The incidence of IAH in our patient population has decreased significantly since 2015. This is likely due to a significant change in routine care of critically ill patients, especially with respect to judicious goal-directed fluid resuscitation. Patient recruitment and protocol adherence in this study were low, exacerbated by other staffing and logistical pressures during the study period. We conclude that a larger multicentre trial is unlikely to yield evidence of a detectable treatment effect.

Data augmentation is a central component of joint embedding self-supervised learning (SSL). Approaches that work for natural images may not always be effective in medical imaging tasks. This study systematically investigated the impact of data augmentation and preprocessing strategies in SSL for lung ultrasound. Three data augmentation pipelines were assessed: (1) a baseline pipeline commonly used across imaging domains, (2) a novel semantic-preserving pipeline designed for ultrasound, and (3) a distilled set of the most effective transformations from both pipelines. Pretrained models were evaluated on multiple classification tasks: B-line detection, pleural effusion detection, and COVID-19 classification. Experiments revealed that semantics-preserving data augmentation resulted in the greatest performance for COVID-19 classification—a diagnostic task requiring global image context. Cropping-based methods yielded the greatest performance on the B-line and pleural effusion object classification tasks, which require strong local pattern recognition. Lastly, semantics-preserving ultrasound image preprocessing resulted in increased downstream performance for multiple tasks. Guidance regarding data augmentation and preprocessing strategies was synthesized for developers working with SSL in ultrasound.

ABSTRACT:To realize the transformative potential of artificial intelligence (AI) in health care, physicians must learn how to use AI-based tools effectively, safely, and equitably. Continuing professional development (CPD) activities are one way to learn how to do this. The purpose of this article is to describe a theory-based approach for assessing health professionals' motivation to participate in CPD on AI-based tools. An online survey, based on an AI competency framework developed from existing literature and expert consultations, was administered to practicing physicians in Ontario, Canada. Across eight subcompetencies for using AI-based tools (eg, appraise AI-based tools for their regulatory and legal status), the survey measured physicians' perception they could successfully enact the competency, the importance of the competency in meeting their practice needs, and the desirability of participating in CPD activities on the competency. Motivation scores were calculated by multiplying the three scores together. Ninety-five physicians completed the survey. The highest motivation scores were for the subcompetency of identifying AI-based tools based on clinical needs, while the lowest motivation scores were for appraising tools' regulatory and legal status. All AI subcompetencies were generally rated as important, and CPD activities were generally perceived as desirable. This survey demonstrates the utility of a theory-based approach for assessing physicians' motivation to learn. Although the survey results are context specific, the approach may be useful for other CPD providers to support decision making about future AI-related CPD activities.

This narrative review explores the integration of artificial intelligence (AI) within critical care settings in KSA in alignment with the goals outlined in Saudi Vision 2030. As part of the Health Sector Transformation Program, the incorporation of AI technologies aims to enhance patient outcomes, optimize workflows, and improve the operational efficiency of intensive care units (ICUs). Key applications include automated clinical documentation, predictive analytics for early detection of clinical deterioration, and AI-assisted imaging techniques, such as chest X-ray and ultrasound interpretation. These innovations can support clinicians by reducing their administrative burden as well as enabling timely interventions, particularly in resource-constrained environments. In addition, AI-powered tele-ICU command centers are explored to extend critical care expertise to underserved regions and enhance equitable access to specialized care. This review was conducted using a structured narrative approach by synthesizing peer-reviewed literature, national policy documents, and expert perspectives from ICU physicians in KSA, Canada, and other international settings.