BACKGROUND:This study evaluated short-term knowledge change and learner perceptions following a virtual reality (VR)-based point-of-care ultrasound (POCUS) training program for undergraduate medical students. METHODS:A single-center observational pre-post educational evaluation was conducted within an undergraduate POCUS training course at the National University of Singapore. Students completed a self-directed online module followed by a VR simulation session covering cardiac, lung, and eFAST ultrasound. Knowledge was assessed using a 19-item multiple-choice questionnaire administered immediately before and after the VR session. Post-session survey responses assessed learner experience, usability, perceived learning benefit, and adverse effects. RESULTS:Of 290 students who completed the pre-training assessment and participated in the VR session, 275 matched pre-post datasets were available for analysis. Mean MCQ score increased from 0.689 out of 1 (SD 0.162) before training to 0.809 out of 1 (SD 0.153) after training, corresponding to a mean paired improvement of 0.120, or approximately 2.3 additional correct answers out of 19 items (95% CI 0.104 to 0.135; p < 0.001; Cohen's dz = 0.93). Survey responses demonstrated favorable learner perceptions. Participants reporting adverse effects rated the session as less engaging than those without adverse effects. CONCLUSIONS:VR-based POCUS simulation, delivered within a blended learning pathway, was associated with improved immediate MCQ-assessed knowledge and favorable learner perceptions. Further studies should assess effectiveness by practical skills assessment and long-term knowledge retention.
The rapid integration of Artificial Intelligence (AI) into Point-of-Care Ultrasound (POCUS) represents a transformative shift, offering the potential to democratize diagnostic expertise while simultaneously presenting significant risks regarding clinical validation, workforce preparedness, and health equity. Informed by a recent global survey indicating that while 81% of practitioners are optimistic about AI, major concerns remain regarding training and evidence gaps, the World Interactive Network Focused On Critical UltraSound (WINFOCUS) proposes a unified strategic framework. This manifesto outlines seven foundational pillars to guide the ethical and effective adoption of AI-augmented POCUS: (1) earning trust through rigorous, prospective evidence; (2) building an AI-literate workforce through evolved curricula; (3) championing global equity to prevent widening health disparities; (4) ensuring algorithmic transparency and accountability; (5) designing for seamless human-AI collaboration; (6) establishing a sustainable, privacy-centric data infrastructure; and (7) committing to continuous, patient-centered evaluation. We present this roadmap as a global call to action for clinicians, researchers, and industry partners to collectively shape a future where technology amplifies clinical wisdom and improves patient outcomes.
BACKGROUND:Ultrasound-guided (USG) peripheral intravenous catheter (PIVC) insertion improves cannulation success, particularly in patients with difficult intravenous access. Whether USG also improves post-insertion outcomes, including catheter failure, dwell time, and complications, remains uncertain. METHODS:We conducted a systematic review and meta-analysis of randomized controlled trials and comparative cohort studies comparing USG with landmark-guided PIVC insertion and reporting at least one post-insertion outcome. PubMed, Embase, Cochrane CENTRAL, and CINAHL were searched from January 2000 to March 2026. Risk of bias was assessed using RoB 2 and the Newcastle-Ottawa Scale. Random-effects meta-analysis used the DerSimonian-Laird estimator; REML with Hartung-Knapp-Sidik-Jonkman adjustment was performed as a post-hoc sensitivity analysis because few studies contributed to pooled outcomes. RESULTS:Fourteen studies (5 randomized trials, 9 cohort studies; 78,209 participants) were included. Catheter failure did not differ significantly between USG and landmark groups (k = 4; RR 1.23, 95% CI 0.99-1.51; p = 0.056), although the estimate was sensitive to model choice and exclusion of the largest study. Dwell time and extravasation were not pooled because of substantial heterogeneity (I² = 91.9% and 95.7%, respectively). Infiltration also showed no significant difference (k = 2; RR 0.68, 95% CI 0.12-3.83). Several studies reporting favorable USG outcomes were substantially confounded by systematic differences in catheter length or material between groups. CONCLUSIONS:Current evidence does not demonstrate a consistent post-insertion advantage of USG over landmark PIVC placement. Reported benefits in some studies may reflect catheter-specification differences rather than ultrasound guidance itself. Larger, well-controlled trials are needed.
BACKGROUND:The objective of this study was to evaluate whether deep learning radiomic (DLR) models utilizing B-mode ultrasound (BUS) and contrast-enhanced ultrasound (CEUS) could improve the preoperative prediction of lymph node metastasis (LNM) in patients with hilar cholangiocarcinoma (HCCA). METHODS:The study included 110 HCCA patients from two clinical centers, divided into primary and external validation cohorts. Pathological verification of lymph node status was performed, and the ResNet101 architecture was used to extract deep learning features (DLFs) from BUS and CEUS images. The Genetic Programming-based Symbolic Regression (GPSR) algorithm was applied to integrate radiomic features (RadFs) and DLFs, generating deep learning radiomic features (DLRFs). DLR models were subsequently constructed using the eXtreme Gradient Boosting (XGBoost) algorithm. RESULTS:Lymph node metastasis was identified in 48 out of 110 patients (43.64%). No significant differences in clinical characteristics were observed between LNM-positive and LNM-negative groups (P-values ranging from 0.14 to 0.98). A total of 837 RadFs and 4095 DLFs were initially extracted from each tumor region of interest (ROI). After feature selection, 10 RadFs (4 from BUS, 6 from CEUS) and 27 DLFs (5 from BUS, 22 from CEUS) were retained. Using the GPSR algorithm, 5 BUS-DLRFs, 10 CEUS-DLRFs, and 15 Combination-DLRFs were generated, leading to the development of three corresponding DLR models. In internal validation, the AUC values were 0.70 for the BUS-DLR model, 0.77 for the CEUS-DLR model, and 0.83 for the Combination-DLR model. In external validation, the AUC values were 0.66, 0.68, and 0.72, respectively. These results indicate that the integration of multiphasic CEUS and BUS data is essential for more comprehensively identifying LNM and achieving precise preoperative staging. CONCLUSIONS:The DLR models based on DLRFs demonstrated an enhanced ability to preoperatively predict LNM in HCCA patients, indicating that the integration of deep learning and RadFs from BUS and CEUS may offer improved predictive performance for LNM.
BACKGROUND:The role of venous Doppler ultrasound in assessing venous congestion in the setting of post-lung transplantation surgery is unknown. We performed a feasibility study on the use of venous Doppler ultrasound in the immediate post-operative period after lung transplantation in a retrospective, single-center observational evaluation. Patients had at least two venous congestion Doppler ultrasound exams involving interrogation of IVC diameter, hepatic vein, portal vein and renal vein Doppler assessments. Results: A total of 9 patients were included in our study with a median age of 60 years, 66.7% being male, and a median BMI of 20.9. All patients had VExUS grade ≥1 on their first venous Doppler exam. All nine patients had a decrease in portal vein pulsatility from initial to final exams. Correlation between portal venous pulsatility and net fluid balance was not statistically significant R= 0.33 (P= 0.39). Other parameters, including VExUS grade, IVC diameter, HV and RV Doppler changes from initial to final ultrasound exams showed variable changes in the post-operative period. CONCLUSION:Venous congestion Doppler ultrasound assessments appear feasible in immediate postoperative lung transplant patients. A multitude of factors beyond fluid status likely affect waveforms.
Background: Pneumonia (PNE) and cardiogenic pulmonary edema (CPE) are characterized by reduced air-spaces dimension and edema. Their distinction through gold standard computed tomography (CT) is challenging due to their pattern similarity. Lung Ultrasound (LUS) is a tool for monitoring the progression of lung pathologies. LUS is portable, real-time, and non-ionized; however, standard LUS (S-LUS) relies on the subjective visualization of imaging patterns, leading to poor reproducibility and lack of diagnostic specificity. To enhance LUS diagnostic utility, quantitative LUS (Q-LUS) was developed. Q-LUS quantifies imaging patterns and explores their correlation to different pathophysiological conditions. In literature, vertical artifacts (VA) quantification proved capable of differentiating PNE and CPE, however, this approach was never compared with gold standard. Methods: We statistically investigate and compare the clinical significance of CT, S-LUS, and Q-LUS, in differentiating PNE and CPE. From a cohort of 55 patients, CT, S-LUS, and Q-LUS data are acquired. CT and S-LUS data of each patient are evaluated to assign a semi-quantitative CT-score and S-LUS-score. Q-LUS radiofrequency data are acquired in multifrequency with convex (2, 3, and 4 MHz) and linear (3, 4, 5, and 6 MHz) probes. VA are manually segmented, quantified into three spectral quantities, and statistically analyzed to extract 15 features for each patient. The diagnostic significance of the scores is tested through Generalized Estimating Equation models. Results & Conclusions: Results show areas under the curve of 74%, 72%, 65%, and 53% for Q-LUS linear, Q-LUS convex, CT-score, and S-LUS-score, respectively, highlighting Q-LUS as the most significant tool.
BACKGROUND:Male infertility is a significant health issue in East Asia, affecting over 12 million men, primarily due to non-obstructive and obstructive azoospermia. Although scrotal color Doppler ultrasonography is widely used to assess male reproductive health, existing diagnostic standards derived from European populations remain unvalidated for Chinese men. This study aimed to establish population-specific reference values and evaluate the diagnostic utility of ultrasonographic parameters in differentiating azoospermia subtypes. METHODS:We enrolled 424 men aged 20-45 years, including 245 normozoospermic controls, 135 with non-obstructive azoospermia, and 44 with obstructive azoospermia from Guangdong Province, China. All participants underwent scrotal ultrasound following standardized protocols to measure testicular volume, epididymal dimensions, and hemodynamic parameters. Semen analysis and hormonal profiling were also performed. RESULTS:Results showed that testicular volumes in fertile men from Guangdong Province, China were significantly smaller than European reference values. Testicular volume was inversely correlated with follicle-stimulating hormone and luteinizing hormone levels and positively correlated with sperm count and concentration, but not with motility or morphology. Men with non-obstructive azoospermia had significantly smaller testicular volumes, while those with obstructive azoospermia showed significant dilation of the epididymal caput. CONCLUSIONS:This study provides the first scrotal ultrasound reference standards for men from Guangdong Province, China and suggests that a two-parameter approach based on testicular volume and epididymal caput diameter may aid in non-invasive differentiation of azoospermia subtypes, potentially reducing the need for invasive biopsies.
BACKGROUND:Optic nerve sheath diameter (ONSD) is a widely used non-invasive surrogate marker of intracranial pressure. Current recommendations state that measurements should be obtained with a neutral gaze. However, the effect of gaze deviation on ONSD has never been investigated. We aimed to evaluate the reliability of ONSD in case of gaze deviation and quantify the associated measurement error. A secondary aim was to assess whether bed inclination has a significant effect on ONSD values. METHODS:We conducted a prospective interventional study on 44 healthy volunteers during routine ocular ultrasound training. ONSD was measured bilaterally along the horizontal and vertical axes at 45° head-of-bed elevation in neutral and deviated gaze. After repositioning the bed to 0° and a 2-minute stabilization period, horizontal measurements were repeated. Measurements were performed according to current consensus recommendations and ALARA principles. Nonparametric analyses were used to analyse the results. RESULTS:Gaze deviation significantly reduced ONSD values across all axes and both eyes, with a consistent median difference of 0.8-1.0 mm (all p < 0.001) and large effect sizes. In contrast, no significant differences were observed between measurements obtained at 45° and 0° (p > 0.5). CONCLUSIONS:Gaze deviation induces a predictable "stretching" effect on the optic nerve sheath, resulting in clinically relevant ONSD reduction. If confirmed in neurocritical populations, this systematic difference may allow estimation of neutral-gaze ONSD in non compliant patients and could form the basis for a dynamic test that estimates optic nerve sheath compliance.
Background: Currently, the role of lung ultrasound (LUS) in the diagnosis and treatment of patients with septic shock is widely recognized. Various LUS protocols and scoring criteria have been proposed, yet a unified LUS protocol for assessing lung water in these patients remains lacking. Methods: Forty‑six septic shock patients underwent LUS with three scanning schemes (4‑region, 8‑region, BLUE) alongside pulse indicated continuous cardiac output (PiCCO) monitoring. Spearman correlation analysis was used to compare the correlation between the three ultrasound scoring schemes, PiCCO and other clinical laboratory indicators. At the same time, the value of three LUS protocols for assessing pulmonary water in septic shock patients was evaluated using receiver operating characteristic (ROC) analysis. Results: Spearman's correlation analysis showed a significant correlation between extravascular lung water index (EVLWI) and 4-region, 8-region, or the Bedside Lung Ultrasound Examination (BLUE) protocol. More importantly, the BLUE protocol showed a stronger correlation with EVLWI than the 4-region protocol and the 8-region protocol (r=0.634, 0.458, 0.546, p<0.001). The ROC curve analysis showed that ultrasound protocols could predict the early occurrence of pulmonary edema in septic shock patients (EVLWI>7 ml/kg), diagnosis of pulmonary edema (EVLWI>10 ml/kg), and evaluation of pulmonary fluid severity (EVLWI≥15 ml/kg). Conclusions: Our study found that septic shock patients were prone to pulmonary edema during fluid resuscitation, and ultrasound scoring could help us quantify pulmonary edema. Among the 4-region, 8-region and BLUE protocols, the BLUE protocol shows relatively better overall performance in predicting lung water elevation and evaluating edema severity.
Background: Bone fractures are common in acute care, and point-of-care ultrasound (POCUS) is an emerging diagnostic tool that can be complementary to or even in some cases an alternative to X-ray imaging. With the rise of artificial intelligence (AI) model incorporation in diagnostic interpretations, this systematic review and meta-analysis aimed to evaluate the diagnostic performance of deep learning models for detecting bone fractures on ultrasound images, using radiographic imaging with expert interpretation as the reference standard. Methods: Comprehensive literature searches were conducted from inception to 15 August 2025, in databases including MEDLINE (Ovid interface), Embase (Ovid interface), CINAHL Plus with Full Text (EBSCOhost interface), Web of Science, ACM Digital Library, Scopus, and Google Scholar. We included studies evaluating deep learning models applied to ultrasound images or sweeps for fracture classification. The included studies predominantly evaluated pediatric cohorts presenting to emergency departments with suspected upper extremity fractures. Sensitivity and specificity were pooled using a bivariate random-effects model on the logit scale, and summary receiver operating characteristic (sROC) curve was constructed. Risk of bias was assessed using the Quality Assessment of Diagnostic Accuracy Studies 2 (QUADAS-2) tool. Results: A total of 580 papers were identified in the preliminary literature search, and 333 studies were screened after duplicate removal. Screening 27 full-text articles resulted in eight studies that were included in the review. Across 21 reported model evaluations from five studies with extractable diagnostic data, pooled sensitivity was 0.77 (95% CI: 0.69-0.83) and pooled specificity was 0.90 (95% CI: 0.86-0.94), with an overall sROC AUC of 0.91. Patient-level evaluations yielded higher pooled sensitivity (0.89; 95% CI: 0.81-0.94) and specificity (0.94; 95% CI: 0.81-0.98) compared to video-level analyses. Conclusions: AI-assisted ultrasound demonstrates promising diagnostic performance for fracture detection with generally high specificity and variable sensitivity across architectures and anatomic targets. Future work should prioritize prospective, externally validated studies with standardized acquisition protocols and clinically meaningful reporting at the patient level.
BACKGROUND:Different pathogens cause pneumonia with overlapping symptoms, labs, and imaging, complicating early diagnosis. The modified lung ultrasound score (MLUS) shows value in lung disease evaluation, but its ability to differentiate pathogens is unclear. This study assessed MLUS for early identification of mycoplasma, viral, and bacterial pneumonia in children. METHODS:We enrolled 186 children with suspected pneumonia (Jan-Dec 2023). Clinical data, labs, lung ultrasound findings (A-lines, B-lines, solid lesions, pleural effusion), and pathogens were recorded. MLUS was assigned. Based on pathogens, cases were grouped into mycoplasma (n=74), viral (n=63), and bacterial (n=49) pneumonia for comparison. RESULTS:1. The median age of the mycoplasma pneumonia group was 6 (4.58,8) years, greater than that of the bacterial pneumonia group (3 [1.1,4.83] years) and the viral pneumonia group (3.41 [1.16,4.75] years) (P＜ 0.05). The mycoplasma pneumonia group had more febrile symptoms, extensive solid lung lesions, and fewer asthmatic symptoms than the other two pathogen groups (P < 0.05). 2. The median MLUS for mycoplasma pneumonia group was 15 (10,22), significantly higher than the median score of 9 (5,16) in the bacterial pneumonia group and 8 (5,15) in the viral pneumonia group (P < 0.05). 3. Coughing up sputum and small lung solid changes were significantly more common in the mycoplasma pneumonia group than in the bacterial pneumonia group (P < 0.05). CONCLUSIONS:A high modified lung ultrasound score, extensive solid lung lesions, older age, fever, and fewer shortness-of-breath symptoms strongly suggest mycoplasma pneumoniae infection. These findings provide critical evidence for early, targeted clinical management.
Background and objective: Traditional point-of-care ultrasound (PoCUS) teaching that relies on hands-on practice faces challenges of limited scale and infection-control constraints during pandemics. We designed a low-contact curriculum integrating near-peer teaching (NPT) with faculty remote supervision (RS) and evaluated its effectiveness versus on-site faculty-led teaching. Methods: In this randomized controlled pilot trial, 69 senior medical students were assigned to NPT+RS (n=34) or faculty-led control group (n=35). Both groups received identical didactic and hands-on training, with the NPT+RS group incorporating multi-camara telemedicine supervision by the faculty to support NPT. The primary outcome was the Entrustable Professional Activity (EPA)-based Objective Structured Clinical Examination (OSCE) score with four major domains measured 1 month after training. The secondary outcomes included the technology acceptance model (TAM) survey and the OSCE feedback. Results: Total OSCE scores (74.9 vs 76.8; p=0.614) and the Indication-Acquisition-Interpretation-Medical decision domain scores did not differ between two groups. For single items, only a lower interpretation performance at the posterolateral alveolar/pleural syndrome in NPT+RS group (4.6 vs 6.5; p=0.013) was found. The NPT+RS group reported higher perceived usefulness (efficiency, usefulness, performance, productivity; all p<0.05), with no between-group difference in perceived ease-of-use. The OSCE was more often perceived by the NPT+RS group as revealing weaknesses (p=0.002) and meriting routine post-training implementation (p=0.034). Conclusions: A low-contact PoCUS curriculum integrating NPT with RS achieved overall EPA-based OSCE performance comparable to traditional faculty-led instruction while enhancing perceived usefulness and assessment acceptance. Implementation of this model allows for greater training capacity and decreased contact. The capacity of NPT for image-interpretation skills warrants investigation.
Background: Preoperative identification of Luminal B breast cancer remains a clinical challenge. This study aimed to develop an ultrasound radiomics framework integrating tumoral and peritumoral information for preoperative identification of Luminal B subtype and prediction of Ki-67 status. Methods: We retrospectively analyzed 1,944 patients from three centers. The development cohort from Centers One and Two was divided by stratified sampling into a training set (n = 1,434) and an internal test set (n = 253), and an independent cohort from Center Three (n = 257) was used for external validation. Lesion-containing ROIs were processed using deep learning-assisted segmentation and standardized for downstream analysis. Radiomic features were extracted, and a genetic algorithm (GA) was coupled with a random forest (RF) classifier to construct two models: one for Luminal B classification and another for predicting Ki-67 expression. Results: The combined tumor-peritumoral model achieved the highest performance, with the Luminal B classifier showing AUCs of 0.876 (training), 0.693 (test), and 0.786 (external validation). The Ki-67 prediction model yielded AUCs of 0.890 (training) and 0.858 (test), though external validation (AUC=0.661) was limited by dataset distribution. The Delong test confirmed that combined ROIs significantly outperformed tumor-only models, with NRI and IDI tests further validating the added value of peritumoral features. Conclusions: Ultrasound radiomics integrating tumoral and peritumoral regions can support the preoperative identification of Luminal B breast cancer, and peritumoral region analysis significantly enhances predictive performance. The framework also shows potential for predicting Ki-67 status within this subtype.
Background: Thoracic ultrasound (TUS) is integrated into clinical practice across various medical specialties to aid with diagnosis and procedural safety. Significant variability in training approaches and standards exists between nations and individual centres. Aim and objectives: To review how differences in published TUS training methods impact upon learning. Methods: A literature search following PRISMA guidelines was conducted in Medline, Embase, Cochrane Library and Scopus, from Jan 2017 until May 2025. Studies involving TUS education using pre- and post-assessments of learning were included. Results: 12,460 studies were screened, 235 full texts assessed, and 68 were included in the review. Studies were mainly observational cohorts targeting different healthcare professionals. Ten randomized controlled trials, four non-randomized comparison and one switching replications evaluating different interventions were identified. Physicians were the primary audience in 65% while 29% focused on other healthcare professionals. Class-based teaching was the most common educational tool (66%) sometimes combined with web-based (16%). 84% of studies involved practical training, with training on humans alone (in 63% of cases). Most educational tools led to significant improvements in test scores, although assessment validity was rarely addressed. Conclusion: Our review confirms various TUS educational methods are used in heterogenous participants. While most interventions demonstrated positive learning effects, few studies addressed how this translates to safe clinical use.
BACKGROUND:Post-extubation stridor (PES) is a common complication in children in the pediatric intensive care unit (PICU). Despite the evaluation of various predictive tests, no single method has consistently demonstrated strong predictive value for PES. We evaluated the diagnostic utility of subglottic diameter ultrasonography for predicting PES within 48-h post-extubation in children. MATERIAL AND METHODS:The prospective observational study in university-affiliated PICU was done in children aged 1 month to 15 years who successfully passed a spontaneous breathing trial (SBT) and were scheduled for extubation were enrolled between January 2024 and January 2025. During SBT and cuff deflation, subglottic ultrasonography was performed at a peak inspiratory pressure (PIP) of 15 cmH2O, and the subglottic ratio (SR) and intracricoid peritubular free space (IPFS) were calculated. Leak percentage was calculated as the ratio of the difference between inspiratory and expiratory volumes over inspiratory volume at each PIP level. RESULTS:PES occurred in 33 of 176 (18.75%) patients. In children aged ≥1 year, SR was significantly associated with PES (adjusted OR, 3.78; 95% CI: 1.28-12.21). An SR <1.30 showed 70% sensitivity, 65% specificity, and 66% accuracy. IPFS <2.08 mm demonstrated 80% sensitivity, 65% specificity, and 68% accuracy. A leak percentage <36.77% at 25 cmH2O yielded 80% sensitivity, 47% specificity, and 53% accuracy. AUC values for SR, IPFS, and leak percentage were 0.664, 0.691, and 0.648, respectively. CONCLUSION:Among the evaluated parameters, IPFS showed the highest diagnostic performance for predicting PES, slightly outperforming the subglottic ratio in patients aged ≥1 year.
Patients supported with veno-arterial extracorporeal membrane oxygenation (VA-ECMO) may require left ventricular (LV) unloading to prevent pulmonary congestion and adverse hemodynamic interactions. However, defining the need and timing of unloading remains challenging, as current approaches rely predominantly on cardiac and pulmonary parameters, often neglecting systemic and organ-level congestion. We report the case of a 53-year-old woman admitted with cardiogenic shock following pericardial drainage and mediastinal mass biopsy. Due to rapid hemodynamic deterioration, VA-ECMO was initiated, resulting in stabilization. Early echocardiographic assessment showed severe biventricular dysfunction but evidence of partial aortic valve opening without LV distension. To further characterize the hemodynamic profile, splanchnic Doppler ultrasound was performed, demonstrating preserved renal arterial flow (resistive index <0.72), continuous intrarenal venous flow, and low portal vein pulsatility (<30%), consistent with a non-congestive phenotype. Based on this integrated assessment, LV unloading was deferred. Subsequent pulmonary artery catheterization confirmed low filling pressures despite reduced cardiac output. A diagnosis of stress-induced cardiomyopathy was suspected, and levosimendan was administered, leading to rapid improvement in cardiac function and successful ECMO weaning. Final pathology revealed a thymic neuroendocrine carcinoma. This case highlights the limitations of relying solely on cardiac indices to guide LV unloading decisions during VA-ECMO. Splanchnic Doppler provided a non-invasive, real-time evaluation of the perfusion-congestion balance at the organ level, complementing echocardiographic and invasive hemodynamic data. The integration of multimodal monitoring may allow a more comprehensive understanding of patient-device interaction and support individualized management strategies. Further research is needed to validate the role of splanchnic ultrasound in guiding unloading decisions and optimizing outcomes in patients with cardiogenic shock supported by VA-ECMO.
Background: The rapid proliferation of artificial intelligence (AI) in medical education has outpaced the development of quality assurance methods for AI-generated content. This study provides the first systematic evaluation of AI-generated instructional materials for lung ultrasound (LUS) training. Methods: The ATLAS study employed a cross-sectional, multi-rater evaluation design comparing six instruction sources (five AI systems and human-generated content) across ten LUS content sessions. Expert evaluators (n=39) assessed materials using five standardized domains: Medical Accuracy, Evidence Completeness, Clarity, Practical Utility, and Pedagogical Quality. Statistical analysis included Kruskal-Wallis tests and pairwise comparisons with Bonferroni correction. Results: Significant differences existed between instruction sources (H = 92.582, p < 0.001). Manus AI achieved the highest overall rating (4.55±0.83) and significantly outperformed human instructions in Medical Accuracy (p = 0.0002) and Evidence Completeness (p < 0.001). Gemini AI (3.94±0.97) performed statistically equivalent to human instructions (4.23±1.00). ChatGPT (2.62±1.35) and Meta (1.53±1.02) performed significantly worse than human instructions (p < 0.001). Clarity emerged as the most discriminating criterion with the widest performance range (1.53-4.54). Conclusions: Certain AI systems can generate high-quality LUS instructional materials that match or exceed human-generated content. However, significant quality variations across AI systems emphasize the critical importance of systematic evaluation before implementation. These findings support cautious but optimistic integration of high-performing AI systems into medical education with appropriate quality assurance measures.
Objectives: To explore a fusion model designed for the quality evaluation of ultrasound images utilized in fetal crown-rump length (CRL) measurement, and to use SHapley Additive exPlanations (SHAP) method to elucidate the model's decision-making processes. Methods: We retrospectively collected 1149 images of midsagittal planes of the entire fetus during early pregnancy from two hospitals. Two senior radiologists categorized the images into standard and non-standard planes. Seven image segmentation models were trained to select the best model for automatically segmenting the region of interest. The radiomics features and deep transfer learning (DTL) features were extracted and selected to establish radiomics models and DTL models. We also constructed fusion models to enhance the classification performance and the optimal one underwent comparison with radiologists. The SHAP method was employed to interpret and visualize the model. Results: The DeepLabV3 ResNet101 segmentation model demonstrated the highest performance (DSC: 97.15%). The early fusion model exhibited superior classification performance in validation set (AUC: 0.947, 95% CI: 0.924-0.970, accuracy: 88.4%, sensitivity: 83.0%, specificity: 92.7%, PPV: 90.1%, NPV:87.3%, precision: 90.1%). The model demonstrated performance commensurate with that of senior radiologists while surpassing junior radiologists. Notably, when leveraging the model's support, there was a substantial improvement in their overall performance. Conclusions: The early fusion model demonstrated satisfactory performance in the intelligent quality evaluation of ultrasound images for CRL measurement. It has the potential to enhance the professional skills of junior radiologists.
Background: Quality assurance (QA) review is essential to any point-of-care ultrasound (POCUS) program that credentials emergency clinicians or hosts an Emergency Ultrasound Fellowship Accreditation Council (EUFAC) accredited fellowship. Commercially designed QA solutions exist but may not be accessible to all programs due to budgetary or institution-specific administration constraints. Objectives: The authors aimed to develop a robust, standalone, HIPAA-compliant POCUS QA database which satisfies both credentialing requirements and EUFAC review goals. Methods: An indexed, searchable electronic QA database was developed using Google workspace. The database was inspected reviewer disagreement trends that led to feedback at the individual and departmental level and was used to satisfy QA review requirements for ultrasound fellows. Results: Since deployment of the database in July 2021, 2742 studies have undergone QA review out of approximately 15,000 studies performed (18.3%). Reviewers disagreed with some portion of the interpretation in 848 (24.8%) exams, most commonly abdominal aortic aneurysm evaluation (n = 45 exams, 31.1%) and basic echocardiography (n = 1203 exams, 27.9%). Disagreement was least common for male genitourinary exams (n = 19 exams, 0%) and advanced echocardiograms (n=57, 15.8%), likely reflecting operator experience. For 209 (7.6%)studies, feedback to the clinicians was warranted and delivered via direct communication. This process has resulted in department-level education in consistently deficient areas. Conclusion: We present our lightweight, easily reproducible alternative to commercial POCUS QA solutions that has led to targeted feedback and department-level education while also satisfying accreditation requirements.
BACKGROUND:Ultrasound Localization Microscopy (ULM) is a milestone in the medical vascular imaging context, enabling the precise characterization of micro-vascular structures using ultrasound imaging. By accurately localizing contrast microbubbles (MBs) flowing in the circulatory system, ULM generates micro-resolved vascular images, overcoming the ultrasonic diffraction limit. However, as ULM relies on precise localization and tracking of individual MBs, high MB concentrations yield to increased localization errors and, ultimately, ULM failure. This constraint limits ULM to low MB concentrations, resulting in long acquisition times that pose challenges in clinical settings. Methods: Here, we show the feasibility of uncoupling a bi-disperse MB population, composed of two monodisperse MB populations. The uncoupling is performed through a signal processing pipeline that exploits the strong nonlinear response of MBs having resonance frequency tuned with the transmission frequency. After uncoupling, ULM density and velocity flow maps are generated. Results: Density and velocity maps are generated after uncoupling, when injecting the bi-disperse population individually and simultaneously in a vascular 3D-printed phantom. Furthermore, density maps generated after uncoupling are compared with the one obtained using standard ULM. Results demonstrate the capability of the proposed uncoupling pipeline to separate the bi-disperse population. Conclusion: This work presents a signal processing pipeline to uncouple a bi-disperse MB population, formed by two monodisperse MB populations. Results are validated in a 3D-printed phantom and demonstrate the feasibility of the uncoupling which, in turn, would enable higher concentrations and reduce acquisition times for micro-vascular imaging.
