Imaging Screening Research Articles

Workload and clinical impact of MRI-based extension of reperfusion therapy time window in acute ischaemic stroke-a prospective single-centre study.

Current European Stroke Organisation (ESO) guidelines recommend extended time window reperfusion therapies (4.5-9h for thrombolysis, 6-24h for thrombectomy) based on advanced imaging. However, the workload and clinical benefit of this strategy on a population basis are not known. To determine the caseload, treatment rates, and outcomes in the extended as compared to the standard time windows. All consecutive ischaemic stroke patients within 24h of last known well between 1st March 2021 and 28th February 2022 were included in a prospective single-centre study. Treatment eligibility in the extended time windows or wake-up strokes recognized within 4h was based on current ESO guideline criteria using MRI DWI-PWI or DWI-FLAIR mismatch. MRI was only available during working hours (8-20h); otherwise, CT/CTA was used. Clinical outcome in treated patients was assessed at three months. Among the 777 admitted patients, 252 (32.4%) had MRI. The thrombolysis rate was 119/304 (39.1%) in standard and 14/231 (6.1%) in the extended time window. The thrombectomy rate was 34/386 (8.8%) in standard and 15/391 (3.8%) in the extended time window. Independent clinical outcomes (mRS ≤ 2) were not statistically different in early and late-treated patients both for thrombolysis (48% vs. 28.6%, p = 0.25) and thrombectomy (38.4% vs. 33.3%, p = 0.99). Even with a limited availability of advanced imaging extending therapeutic time windows resulted in an 11.7% increase in thrombolysis and a 44% increase in thrombectomy with comparable clinical outcomes in early and late-treated patients at the price of a twofold burden in clinical and advanced imaging screening.

ResGloTBNet: an interpretable deep residual network with global long-range dependency for tuberculosis screening of sputum smear microscopy images

ResGloTBNet: an interpretable deep residual network with global long-range dependency for tuberculosis screening of sputum smear microscopy images

Cell membrane biomimetic magnetic fluorescent bifunctional nanoplatform for drug lead discovery.

Biomimetic nanoplatforms based on membrane coating strategies have received increasing attention in the field of medical research. However, it cannot perform biomedical imaging screening, which is essential for real-time identification. As a rich source of new drug discovery, traditional Chinese medicine (TCM) has made important contributions to the treatment of many diseases. Therefore, a magnetic fluorescent bifunctional nanomaterial was developed for screening the active ingredients in Fuzi. The magnetic fluorescence nanoparticles Fe3O4@GO@RhB synthesized by one-step synthesis showed low toxicity, excellent fluorescence properties and fast magnetic separation. After coating the magnetic nanoparticle with cell membrane, the screening procedure was optimized with positive drug verapamil hydrochloride. Under the optimal conditions, the active ingredients in aconite were screened, and the three active ingredients were identified as benzoyl-mesaconine, benzoyl-aconine and benzoyl-hypacoitine by UPLC-MS/MS. In addition, in situ imaging technique was used to further verify the proliferative activity of the screened active ingredients. The magnetic fluorescence bifunctional nanoplatform for cell membrane bionic screening combined with in situ imaging technology established in this paper not only can rapidly and effectively target screening and separation of active ingredients in TCM, but also provides a new platform for bioimaging screening.

Scalable synthesis of Sb-doped Rb3InCl6 nanocrystals for high-resolution X-ray imaging screen

Scalable synthesis of Sb-doped Rb3InCl6 nanocrystals for high-resolution X-ray imaging screen

Eye Diseases Classification Using Transfer Learning Technique of Deep Neural Network

The human eye is a complex organ susceptible to various diseases and abnormalities that can impair vision. Identifying retinal diseases early and precisely categorizing them is crucial to ensure effective treatment is administered promptly. In this proposed study, we implemented Convolutional Neural Networks (CNNs) along with pretrained neural network architectures. We use existing methods as the foundation and develop models tailored for detecting eye diseases, such as Diabetic Retinopathy (DR), Cataracts, and Glaucoma, specifically among diabetic patients. These models are applied for screening of retinal fundus images for accurate diagnosis. Prompt identification and suitable treatment can aid in preventing the commencement and advancement of diabetic retinopathy, cataracts, and glaucoma among patients. Accurate identification of these conditions is paramount in the healthcare sector. Our research aims to categorize eye diseases through the utilization of retinal fundus images for multiclass classification. The multiclass classification outcomes, implemented on the Kaggle platform, indicated that CNN, MobileNet, DenseNet121, ResNet and EfficientNetB3 yielded training accuracies of 98.50%, 99.94%, 99.37%, 99.90% and 99.97 as well as the validation accuracies of 88.22%, 95.67%, 93.99%, 93.30% and 95.97% respectively.

Fish Evo‐Devo: Moving Toward Species‐Specific and Knowledge‐Based Interactome

ABSTRACTA knowledge‐based interactome maps interactions among proteins and molecules within a cell using experimental data, computational predictions, and literature mining. These interactomes are vital for understanding cellular functions, pathways, and the evolutionary conservation of protein interactions. They reveal how interactions regulate growth, differentiation, and development. Transitioning to functionally validated interactomes is crucial in evolutionary developmental biology (Evo‐Devo), especially for non‐model species, to uncover unique regulatory networks, evolutionary novelties, and reliable gene interaction models. This enhances our understanding of complex trait evolution across species. The European Evo‐Devo 2024 conference in Helsinki hosted the first fish‐specific Evo‐Devo symposium, highlighting the growing interest in fish models. Advances in genome annotation, genome editing, imaging, and molecular screening are expanding fish Evo‐Devo research. High‐throughput molecular data have enabled the deduction of gene regulatory networks. The next steps involve creating species‐specific interactomes, validating them functionally, and integrating additional molecular data to deepen the understanding of complex regulatory interactions in fish Evo‐Devo. This short review aims to address the logical steps for this transition, as well as the necessities and limitations of this journey.

'Such a lot of bother': Qualitative results of a home trial of a wearable electronic vision enhancement system for people with age-related macular degeneration.

Wearable electronic low vision enhancement systems (wEVES) improve visual function but are not widely adopted by people with vision impairment. Here, qualitative research methods were used to investigate the usefulness of wEVES for people with age-related macular degeneration (AMD) after an extended home trial. Following a 12-week non-masked randomised crossover trial, semi-structured interviews were completed with 34 participants with AMD, 64.7% female, mean age 80.2 (±6.0) years, mean distance visual acuity 0.81logMAR (±0.32). Reflexive thematic analysis was used to analyse the data. Four themes were developed: (i) early positivity and potential; (ii) you're not good enough: performance barriers of the device; (iii) you're annoying: practicality barriers of the device and (iv) we can fix this together. First, participants expressed joy in an aesthetically appealing device perceived as potentially enabling, different and complementary to their current solutions. Imagined usefulness included not only reading, shopping and television but also resuming abandoned hobbies. The second theme captured performance barriers that restricted numerous activities but were most acutely noted with manipulation tasks. Barriers included image quality, screen size and short-lived adverse effects. The third theme conveyed the multiple practical challenges that caused annoyance, preventing imagined usage even when performance appeared superior to other solutions. Slow start-up times and the inability to use wEVES dynamically prevented integration within users' lifestyles. The final theme reflected that wEVES remained a desirable concept, but future iterations require inclusive design methodology to ensure development is directed by consumers' needs. Performance and practicality barriers limit the usefulness of a device initially seen as desirable. Current devices do not align with users' requirements for flexible use, even when performance is good. Improvements in technology may solve performance barriers, but these changes must be inclusively designed and evaluated to ensure the device integrates more successfully into the lives of users with AMD.

The impact of brain MRI screening on stage IV NSCLC patients: A real world look at guidelines based care.

Brain metastases frequently occur in patients with non-small cell lung cancer (NSCLC) and are associated with poor prognosis and shortened overall survival, despite the advances in both imaging and therapeutic fields. Data are scarce regarding the utility of brain magnetic resonance imaging (MRI) screening in NSCLC patients. We aimed to characterize the impact of brain MRI screening on stage IV NSCLC patients and their survival. A retrospective analysis was performed in 609 newly-diagnosed patients with stage IV NSCLC treated at our center during 2019-2020. Patients with neurological symptoms at presentation were excluded (n=230). The remaining 379 patients comprised the study group and were divided into two groups - those who underwent brain MRI screening within 12weeks of diagnosis (n=170), and those who did not (n=209). The clinical data retrieved from patients' medical files included demographics, performance status (PS), brain metastases during follow-up, tumor molecular profiling, and oncology treatment. Median survival among patients who underwent brain MRI screening was 24months from diagnosis, versus 18months for those who did not (p=0.003). Among patients with good PS (ECOG 0-2), median OS was longer among patients who underwent screening MRI (25months), versus those who did not (21months) (p=0.025). Among patients with low PS (ECOG 3-4), the between-group difference for OS did not reach statistical significance. Our study supports the use of brain MRI screening among good PS patients diagnosed with stage IV NSCLC lung cancer. Brain MRI screening correlated with better survival among patients with good functional status in this patient population.

Occult Abdominal Trauma Screening in the Evaluation of Suspected Child Physical Abuse

BACKGROUND AND OBJECTIVES Occult abdominal trauma (OAT) screening with transaminases, followed by abdominal computed tomography (CT) for transaminase values >80 IU/L, has been recommended in cases of suspected physical abuse. This study aimed to evaluate case characteristics associated with OAT evaluation and determine OAT prevalence in these children. METHODS Injured children <60 months old undergoing Child Abuse Pediatrics (CAP)consultation for suspected physical abuse from February 2021 to May 2023 were identified in CAPNET, a multicenter research network. Children with symptoms or signs of intra-abdominal injury were excluded. We identified case characteristics associated with transaminase screening and abdominal CT imaging using logistic regression and determined OAT prevalence. RESULTS 3982/6161 (64.6%) eligible children underwent transaminase screening; 687/3982 (17.3%) had transaminases >80 IU/L with 298/687 (43.4%) undergoing abdominal CT imaging. Variability in screening and imaging practices was identified between CAPNET sites. In a fully adjusted model, transaminase screening was associated with age <6 months, greater clinical severity, and site. CT imaging was associated with site, inpatient status, and higher transaminase range. We identified 16 OAT cases in children with transaminases >80 IU/L, representing 2.3% of CAP-evaluated children with positive transaminase screening and 0.3% of all eligible children. CONCLUSIONS Providers often perform transaminase screening but not abdominal CT imaging despite transaminases >80 IU/L. The low prevalence of OAT suggests that routine transaminase screening in suspected child physical abuse evaluations may not be necessary when all signs and symptoms of abdominal injury are absent.

Predicting cell morphological responses to perturbations using generative modeling

Advancements in high-throughput screenings enable the exploration of rich phenotypic readouts through high-content microscopy, expediting the development of phenotype-based drug discovery. However, analyzing large and complex high-content imaging screenings remains challenging due to incomplete sampling of perturbations and the presence of technical variations between experiments. To tackle these shortcomings, we present IMage Perturbation Autoencoder (IMPA), a generative style-transfer model predicting morphological changes of perturbations across genetic and chemical interventions. We show that IMPA accurately captures morphological and population-level changes of both seen and unseen perturbations on breast cancer and osteosarcoma cells. Additionally, IMPA accounts for batch effects and can model perturbations across various sources of technical variation, further enhancing its robustness in diverse experimental conditions. With the increasing availability of large-scale high-content imaging screens generated by academic and industrial consortia, we envision that IMPA will facilitate the analysis of microscopy data and enable efficient experimental design via in-silico perturbation prediction.

A thermoplastic chip for 2D and 3D correlative assays combining screening and high-resolution imaging of immune cell responses.

We present an easy-to-use, disposable, thermoplastic microwell chip designed to support screening and high-resolution imaging of single-cell behavior in two- and three-dimensional (2D and 3D) cell cultures. We show that the chip has excellent optical properties and provide simple protocols for efficient long-term cell culture of suspension and adherent cells, the latter grown either as monolayers or as hundreds of single, uniformly sized spheroids. We then demonstrate the applicability of the system for single-cell analysis by correlating the dynamic cytotoxic response of single immune cells grown under different metabolic conditions to their intracellular cytolytic load at the end of the assay. Additionally, we illustrate highly multiplex cytotoxicity screening of tumor spheroids in the chip, comparing the effect of environment cues characteristic of the tumor microenvironment on natural killer (NK)-cell-induced killing. Following the functional screening, we perform high-resolution 3D immunofluorescent imaging of infiltrating NK cells within the spheroid volumes.

One Scan, Multiple Insights: A Review of AI-Driven Biomarker Imaging and Composite Measure Detection in Lung Cancer Screening

One Scan, Multiple Insights: A Review of AI-Driven Biomarker Imaging and Composite Measure Detection in Lung Cancer Screening

Automated Detection of Normal Pressure Hydrocephalus Using CT Imaging for Calculating the Ventricle-to-Subarachnoid Volume Ratio.

Normal pressure hydrocephalus (NPH) is a diagnostic challenge because its clinical symptoms and imaging appearance resemble normal aging and other forms of dementia. Identifying NPH is essential so that patients can receive timely treatment to improve gait distortion and quality of life. An automated marker of NPH was developed and evaluated on clinical CT images, and its utility was assessed in a large patient cohort. A retrospective review was conducted of CT images from 306 tap test-responsive patients with NPH between January 2015 and January 2022. Control CT images were obtained from patients in the emergency department who were evaluated for headache and had unremarkable CT findings between June 2021 and August 2022. The ventricle-to-subarachnoid volume ratio (VSR) was automatically calculated by the imaging software and used as a predictor of NPH in linear regression modeling with adjustment for age and sex. The correlations of VSR with age, sex, and the receiver operating characteristic were computed. VSR was significantly greater in patients with NPH than controls (P < .001). Importantly, VSR was not significantly correlated with age (P = .56, R2 = 0.001). VSR identifies NPH with a sensitivity and specificity of 94.1% and 92.5%, respectively, with an area under the receiver operating characteristic curve of 0.99 (95% CI 0.975-0.995). Automated assessment of the VSR on head CT images identified probable NPH with 93% accuracy. The assessment of a large cohort of patients with NPH supports the generalizability of clinical screening of CT images. Moreover, the results support the utility of ventricle-to-sulcal concordance often used by radiologists but not currently a part of the accepted guidelines for imaging markers of NPH.

Imaging Findings and Management Strategies for Liver Masses in Children with Predisposition Disorders: A Review by the Pediatric LI-RADS Group.

Liver masses in children with underlying systemic disease or a predisposing syndrome can be benign or malignant, ranging from focal fat to hepatocellular carcinoma (HCC). Knowledge of the underlying condition, the pathophysiologic effect on the liver, and the development of liver disease and specific liver lesions allows radiologists to guide imaging with regard to modality and frequency and give recommendations for biopsy when appropriate. In some predisposition disorders, such as Beckwith Wiedemann spectrum, familial adenomatous polyposis syndrome, and tuberous sclerosis complex, established guidelines for imaging screening exist. In many of the syndromes discussed, masses may occur outside of the liver and the liver may not be the primary focus of screening. For other entities, no consensus recommendations exist. Screening recommendations may be based on the risk of development of chronic liver disease. Once cirrhosis occurs, the risk of developing HCC is elevated. The authors summarize the spectrum of liver lesions that may be encountered in children with predisposing syndromes and systemic diseases, the imaging appearance of the lesions with various modalities, and screening guidelines where published. ©RSNA, 2024 See the invited commentary by Rutten and Chavan in this issue.

High-Dose Cyproterone Acetate and Intracranial Meningioma: Impact of the Risk Minimisation Measures Implemented in France in 2018-2019.

To measure the impact of national regulatory actions implemented in France in August 2018 and June 2019 to reduce the risk of meningioma associated with the use of cyproterone acetate (CPA). Using the French National Healthcare database, we calculated the monthly number of CPA users among cisgender women, men and transgender women in 2010-2021, the monthly proportion of users with cerebral imaging screening, and the annual rate of meningioma surgery associated with CPA use. CPA discontinuations and switches were analysed. Between 2018 and 2021, the number of individuals exposed to CPA fell by 85% (55 000 in August 2018 versus 7900 users of high-dose CPA in December 2021), corresponding to two waves of decrease in both use and initiation. This drop was greater among cisgender women (88%) than men (69%) or transgender women (50%). Cerebral imaging screening increased from 11% in June 2018 to 70% in June 2021 for ciswomen (13%-51% for men, 9%-60% for transwomen). After CPA discontinuation, no massive shift to a single product was observed, but, instead, dispersion towards other hormonal therapies. The overall annual rate of meningioma surgery associated with CPA exposure spectacularly decreased between 2017 and 2021 (-93% for ciswomen and -86% for men). In France, high-dose CPA use sharply decreased after the implementation of national regulatory measures without a massive switch to other hormonal therapies. The increase in cerebral imaging screening did not result in an increase in meningioma surgery associated with CPA, but rather a massive drop of over 90%.

Improved Consistency of Lung Nodule Categorization in CT Scans with Heterogeneous Slice Thickness by Deep Learning-Based 3D Super-Resolution

Background/Objectives: Accurate volumetric assessment of lung nodules is an essential element of low-dose lung cancer screening programs. Current guidance recommends applying specific thresholds to measured nodule volume to make the following clinical decisions. In reality, however, CT scans often have heterogeneous slice thickness which is known to adversely impact the accuracy of nodule volume assessment. Methods: In this study, a deep learning (DL)-based 3D super-resolution method is proposed for generating thin-slice CT images from heterogeneous thick-slice CT images in lung cancer screening. We evaluated the performance in a qualitative way by radiologist’s perceptual assessment as well as in a quantitative way by accuracy of nodule volume measurements and agreement of volume-based Lung-RADS nodule category. Results: A 5-point Likert scale tabulated by two radiologists showed that the quality of DL-generated thin-slice images from thick-slice CT images were on a par with the image quality of ground truth thin-slice CT images. Furthermore, thick- and thin-slice CT images had a nodule volume difference of 52.2 percent on average which was reduced to a 15.7 percent difference with DL-generated thin-slice CT. In addition, the proposed method increased the agreement of lung nodule categorization using Lung-RADS by 74 percent. Conclusions: The proposed DL approach for slice thickness normalization has a potential for improving the accuracy of lung nodule volumetry and facilitating more reliable early lung nodule detection.

Follow-up Adherence After Teleretinal Screening for Diabetic Retinopathy.

Determine the follow-up rate for patients referred following an abnormal teleretinal imaging (TRI) screening and examine patient characteristics predictive of follow-up nonadherence. A cross-sectional study of patients screened between August 2014 and July 2016 in the Harris Health System (HHS) in Houston, TX. All diabetic patients referred for in-person examination, who had data for all study variables, and who did not have established ophthalmic care in the HHS within the previous 2 years were included. Data collected included TRI findings, clinic location, age, race, gender, hemoglobin A1C (HbA1c), and insulin use. The primary outcome was whether or not a patient attended an in-person follow-up visit. Attendance was the dependent variable in a multivariable logistic regression with the aforementioned exposures used as independent variables. There were 1695 patients included in the study. The follow-up rate (ie, the percentage of patients who went for an in-person examination with an ophthalmologist following a positive screening test and referral) was 54.9%. TRI findings of proliferative diabetic retinopathy (PDR) + referable diabetic macular edema (DME) predicted poorer compliance with follow-up recommendations compared to patients with PDR alone (OR 0.64, 95% CI 0.42-0.99; P=0.046). Age, race, gender, glycemic control, and insulin use did not predict compliance. Age, race, gender, glycemic control, and insulin use were not found to be predictive of follow-up compliance rates; the most severe disease (PDR + DME) predicted noncompliance. Diabetic retinopathy screening programs may consider more aggressive measures in emphasizing compliance in patients with the worst disease as these patients are more likely to miss follow-up appointments. More studies are needed to characterize risk factors for noncompliance.

A preliminary test of ChatGPT’s ESG literacy

PurposeMotivated by recent research on the pivotal role of artificial intelligence (AI) in the interplay of financial activities and ESG issues, we examine ChatGPT’s ability to assist the public, investors, corporate boards, shareholders and companies with ESG-related issues.Design/methodology/approachWe conduct a preliminary test of ChatGPT’s knowledge of ESG by feeding ChatGPT with questions from three sources: Bloomberg, Corporate Finance Institute and Alison.com. We compare ChatGPT-4o mini to random guessing, Google’s Gemini and ChatGPT-4o. We test ChatGPT-4o mini’s familiarity with Bloomberg terminal functions related to ESG.FindingsWe find ChatGPT’s accuracy is 100% on Bloomberg questions. We also find that ChatGPT-4o mini’s performance is slightly better than that of Gemini and ChatGPT-4o. However, we document that ChatGPT-4o mini sometimes fabricates non-existing functions, which is evidence of hallucinations, a built-in flaw of ChatGPT. Lastly, we show that ChatGPT-4o mini can analyze Bloomberg terminal screen images, assess companies’ actual ESG reports and draft CEO compensation contracts with integrated ESG metrics.Originality/valueOur study is the first attempt to test ChatGPT’s knowledge in the domain of ESG. Our findings suggest the potential use of ChatGPT by the public to educate themselves on ESG issues, by investors to integrate ESG in portfolio construction, by corporate boards to incorporate ESG metrics in CEO compensation contracts, by companies to file ESG reports to regulators, by ESG-conscious shareholders to engage the management, etc.

Parkinson's disease subtypes and their association with probable rapid eye movement sleep behavior disorder severity: a brainstem tractography and machine learning investigation.

Rapid Eye Movement (REM) sleep behavior disorder (RBD) affects nearly half of Parkinson's disease (PD) patients. However, the structural heterogeneity within the brainstem, which regulates REM sleep, remains largely unexplored in PD. Our objective was to identify distinct PD subtypes based on microstructural characteristics in the brainstem and examine their associations with the severity of RBD. Data, including diffusion tensor imaging and REM sleep behavior disorder screening questionnaire (RBDSQ) responses, were obtained from 124 PD patients and 61 healthy controls through the Parkinson's Progression Marker Initiative database. Mean Quantitative Anisotropy (QA) values, representing axonal density, were extracted from 14 brainstem tracts and input into the semi-supervised machine learning algorithm, Heterogeneity through Discriminative Analysis (HYDRA), to cluster subtypes. Applying HYDRA, we identified two distinct PD subtypes (Subtype 1: n = 66, Subtype 2: n = 58). Subtype 2 exhibited reduced QA across assessed brainstem tracts and significantly higher RBDSQ scores than Subtype 1 and healthy controls (p < 0.001). Conversely, Subtype 1, characterized by lower RBDSQ scores, exhibited increased QA, notably in the right medial longitudinal fasciculus, when compared to Subtype 2 and controls (p < 0.001). These findings suggest that heterogeneous axonal damage in brainstem circuits correlates with variations in RBD severity, providing insights into the neurobiological underpinnings of early PD.

Cost-effectiveness analysis of AI-based image quality control for perinatal ultrasound screening

PurposeThis study aimed to compare the cost-effectiveness of AI-based approaches with manual approaches in ultrasound image quality control (QC).MethodsEligible ultrasonographers and pregnant volunteers were prospectively recruited from the Hunan Maternal and Child Health Hospital in May 2023. The ultrasonographers were randomly and evenly assigned to either the AI or Manual QC groups with baseline scores determined in June-July. From August to October, these groups received real-time AI or post-scan manual QC with post-interventional scores recorded monthly. We applied the repeated measures analysis of variance to analyze the between-subject and within-subject effectiveness and time trends in effectiveness (QC score improvement) assessment. An extra 50 pregnant volunteers underwent real-time manual QC, with their screening images utilized for post-scan AI and manual QC. The time cost of real-time AI QC was zero since it only required trainees’ involvement. We used Friedman’s M and Quade tests to compare multiple independent medians in cost assessment.ResultsThis study recruited 14 ultrasonographers, equally divided into the AI and Manual QC groups. No significant difference existed between the groups concerning age, service year in perinatal diagnosis, male proportion, and QC frequency. The simple effect of the group revealed that the AI QC method outperformed the Manual QC method at least once (F = 13.113, P = 0.004, η2 = 0.522). The simple effect of the month in the AI QC groups indicated an improvement in the mean QC scores (F = 9.827, P = 0.003, η2 = 0.747) while that of manual QC groups suggested no improvement (F = 0.144, P = 0.931, η2 = 0.041). Baseline scores were equal in June-July (F = 0.031, P = 0.864, η2 = 0.003). However, the AI QC group surpassed the Manual QC group in August (F = 14.579, P = 0.002, η2 = 0.549), September (F = 28.590, P < 0.001, η2 = 0.704), and October (F = 35.411, P < 0.001, η2 = 0.747). Within the Manual QC group, no significant differences were found in scores between June-July and August, September, or October (all P values of 1.000, nominal significance level of 0.0083). In contrast, the AI QC group showed significantly higher scores in August, September, and October compared to June-July (all P values of 0.001, nominal significance level of 0.0083). The time costs of real-time AI QC, post-scan AI QC, post-scan manual QC, and real-time manual QC were 0 s, 13.76 s (interquartile range, IQR: 4.79–46.79 s), 1239.50 s (IQR: 1141.00–1311.25 s), and 1541.00 s (IQR: 1453.50–1635.25 s), with significant differences in both overall and multiple comparisons.ConclusionsThe AI QC method, more cost-effective than the manual method, shows great potential for application in image QC scenarios. The AI QC can enhance operators’ skills in perinatal ultrasound screening, while the manual method can only maintain the existing level.