Dataset Information

A mobile-optimized artificial intelligence system for gestational age and fetal malpresentation assessment.

ABSTRACT:

Background

Fetal ultrasound is an important component of antenatal care, but shortage of adequately trained healthcare workers has limited its adoption in low-to-middle-income countries. This study investigated the use of artificial intelligence for fetal ultrasound in under-resourced settings.

Methods

Blind sweep ultrasounds, consisting of six freehand ultrasound sweeps, were collected by sonographers in the USA and Zambia, and novice operators in Zambia. We developed artificial intelligence (AI) models that used blind sweeps to predict gestational age (GA) and fetal malpresentation. AI GA estimates and standard fetal biometry estimates were compared to a previously established ground truth, and evaluated for difference in absolute error. Fetal malpresentation (non-cephalic vs cephalic) was compared to sonographer assessment. On-device AI model run-times were benchmarked on Android mobile phones.

Results

Here we show that GA estimation accuracy of the AI model is non-inferior to standard fetal biometry estimates (error difference -1.4 ± 4.5 days, 95% CI -1.8, -0.9, n = 406). Non-inferiority is maintained when blind sweeps are acquired by novice operators performing only two of six sweep motion types. Fetal malpresentation AUC-ROC is 0.977 (95% CI, 0.949, 1.00, n = 613), sonographers and novices have similar AUC-ROC. Software run-times on mobile phones for both diagnostic models are less than 3 s after completion of a sweep.

Conclusions

The gestational age model is non-inferior to the clinical standard and the fetal malpresentation model has high AUC-ROCs across operators and devices. Our AI models are able to run on-device, without internet connectivity, and provide feedback scores to assist in upleveling the capabilities of lightly trained ultrasound operators in low resource settings.

SUBMITTER: Gomes RG

PROVIDER: S-EPMC9553916 | biostudies-literature | 2022

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

A mobile-optimized artificial intelligence system for gestational age and fetal malpresentation assessment.

Gomes Ryan G RG Vwalika Bellington B Lee Chace C Willis Angelica A Sieniek Marcin M Price Joan T JT Chen Christina C Kasaro Margaret P MP Taylor James A JA Stringer Elizabeth M EM McKinney Scott Mayer SM Sindano Ntazana N Dahl George E GE Goodnight William W Gilmer Justin J Chi Benjamin H BH Lau Charles C Spitz Terry T Saensuksopa T T Liu Kris K Tiyasirichokchai Tiya T Wong Jonny J Pilgrim Rory R Uddin Akib A Corrado Greg G Peng Lily L Chou Katherine K Tse Daniel D Stringer Jeffrey S A JSA Shetty Shravya S

Communications medicine 20221011

<h4>Background</h4>Fetal ultrasound is an important component of antenatal care, but shortage of adequately trained healthcare workers has limited its adoption in low-to-middle-income countries. This study investigated the use of artificial intelligence for fetal ultrasound in under-resourced settings.<h4>Methods</h4>Blind sweep ultrasounds, consisting of six freehand ultrasound sweeps, were collected by sonographers in the USA and Zambia, and novice operators in Zambia. We developed artificial ...[more]

PMID: 36249461

Similar Datasets

Project description:BackgroundIn 2020, our center established a Tanner-Whitehouse 3 (TW3) artificial intelligence (AI) system using a convolutional neural network (CNN), which was built upon 9059 radiographs. However, the system, upon which our study is based, lacked a gold standard for comparison and had not undergone thorough evaluation in different working environments.MethodsTo further verify the applicability of the AI system in clinical bone age assessment (BAA) and to enhance the accuracy and homogeneity of BAA, a prospective multi-center validation was conducted. This study utilized 744 left-hand radiographs of patients, ranging from 1 to 20 years of age, with 378 boys and 366 girls. These radiographs were obtained from nine different children's hospitals between August and December 2020. The BAAs were performed using the TW3 AI system and were also reviewed by experienced reviewers. Bone age accuracy within 1 year, root mean square error (RMSE), and mean absolute error (MAE) were statistically calculated to evaluate the accuracy. Kappa test and Bland-Altman (B-A) plot were conducted to measure the diagnostic consistency.ResultsThe system exhibited a high level of performance, producing results that closely aligned with those of the reviewers. It achieved a RMSE of 0.52 years and an accuracy of 94.55% for the radius, ulna, and short bones series. When assessing the carpal series of bones, the system achieved a RMSE of 0.85 years and an accuracy of 80.38%. Overall, the system displayed satisfactory accuracy and RMSE, particularly in patients over 7 years old. The system excelled in evaluating the carpal bone age of patients aged 1-6. Both the Kappa test and B-A plot demonstrated substantial consistency between the system and the reviewers, although the model encountered challenges in consistently distinguishing specific bones, such as the capitate. Furthermore, the system's performance proved acceptable across different genders and age groups, as well as radiography instruments.ConclusionsIn this multi-center validation, the system showcased its potential to enhance the efficiency and consistency of healthy delivery, ultimately resulting in improved patient outcomes and reduced healthcare costs.

Project description:IntroductionEstimating a reliable gestational age (GA) is essential in providing appropriate care during pregnancy. With advancements in data science, there are several publications on the use of artificial intelligence (AI) models to estimate GA using ultrasound (US) images. The aim of this meta-analysis is to assess the accuracy of AI models in assessing GA against US as the gold standard.MethodsA literature search was performed in PubMed, CINAHL, Wiley Cochrane Library, Scopus, and Web of Science databases. Studies that reported use of AI models for GA estimation with US as the reference standard were included. Risk of bias assessment was performed using Quality Assessment for Diagnostic Accuracy Studies-2 (QUADAS-2) tool. Mean error in GA was estimated using STATA version-17 and subgroup analysis on trimester of GA assessment, AI models, study design, and external validation was performed.ResultsOut of the 1,039 studies screened, 17 were included in the review, and of these 10 studies were included in the meta-analysis. Five (29%) studies were from high-income countries (HICs), four (24%) from upper-middle-income countries (UMICs), one (6%) from low-and middle-income countries (LMIC), and the remaining seven studies (41%) used data across different income regions. The pooled mean error in GA estimation based on 2D images (n = 6) and blind sweep videos (n = 4) was 4.32 days (95% CI: 2.82, 5.83; l 2: 97.95%) and 2.55 days (95% CI: -0.13, 5.23; l 2: 100%), respectively. On subgroup analysis based on 2D images, the mean error in GA estimation in the first trimester was 7.00 days (95% CI: 6.08, 7.92), 2.35 days (95% CI: 1.03, 3.67) in the second, and 4.30 days (95% CI: 4.10, 4.50) in the third trimester. In studies using deep learning for 2D images, those employing CNN reported a mean error of 5.11 days (95% CI: 1.85, 8.37) in gestational age estimation, while one using DNN indicated a mean error of 5.39 days (95% CI: 5.10, 5.68). Most studies exhibited an unclear or low risk of bias in various domains, including patient selection, index test, reference standard, flow and timings and applicability domain.ConclusionPreliminary experience with AI models shows good accuracy in estimating GA. This holds tremendous potential for pregnancy dating, especially in resource-poor settings where trained interpreters may be limited.Systematic review registrationPROSPERO, identifier (CRD42022319966).

Project description:Background: Manual bone age assessment (BAA) is associated with longer interpretation time and higher cost and variability, thus posing challenges in areas with restricted medical facilities, such as the high-altitude Tibetan Plateau. The application of artificial intelligence (AI) for automating BAA could facilitate resolving this issue. This study aimed to develop an AI-based BAA model for Han and Tibetan children. Methods: A model named "EVG-BANet" was trained using three datasets, including the Radiology Society of North America (RSNA) dataset (training set n = 12611, validation set n = 1425, and test set n = 200), the Radiological Hand Pose Estimation (RHPE) dataset (training set n = 5491, validation set n = 713, and test set n = 79), and a self-established local dataset [training set n = 825 and test set n = 351 (Han n = 216 and Tibetan n = 135)]. An open-access state-of-the-art model BoNet was used for comparison. The accuracy and generalizability of the two models were evaluated using the abovementioned three test sets and an external test set (n = 256, all were Tibetan). Mean absolute difference (MAD) and accuracy within 1 year were used as indicators. Bias was evaluated by comparing the MAD between the demographic groups. Results: EVG-BANet outperformed BoNet in the MAD on the RHPE test set (0.52 vs. 0.63 years, p < 0.001), the local test set (0.47 vs. 0.62 years, p < 0.001), and the external test set (0.53 vs. 0.66 years, p < 0.001) and exhibited a comparable MAD on the RSNA test set (0.34 vs. 0.35 years, p = 0.934). EVG-BANet achieved accuracy within 1 year of 97.7% on the local test set (BoNet 90%, p < 0.001) and 89.5% on the external test set (BoNet 85.5%, p = 0.066). EVG-BANet showed no bias in the local test set but exhibited a bias related to chronological age in the external test set. Conclusion: EVG-BANet can accurately predict the bone age (BA) for both Han children and Tibetan children living in the Tibetan Plateau with limited healthcare facilities.

Project description:ImportanceWith advancements in mobile technology and artificial intelligence (AI) methods, there has been a substantial surge in the availability of direct-to-consumer mobile applications (apps) claiming to aid in the assessment and management of diverse skin conditions. Despite widespread patient downloads, these apps exhibit limited evidence supporting their efficacy.ObjectiveTo identify and characterize current English-language AI dermatology mobile apps available for download, focusing on aspects such as purpose, supporting evidence, regulatory status, clinician input, data privacy measures, and use of image data.Evidence reviewIn this cross-sectional study, both Apple and Android mobile app stores were systematically searched for dermatology-related apps that use AI algorithms. Each app's purpose, target audience, evidence-based claims, algorithm details, data availability, clinician input during development, and data usage privacy policies were evaluated.FindingsA total of 909 apps were initially identified. Following the removal of 518 duplicates, 391 apps remained. Subsequent review excluded 350 apps due to nonmedical nature, non-English languages, absence of AI features, or unavailability, ultimately leaving 41 apps for detailed analysis. The findings revealed several concerning aspects of the current landscape of AI apps in dermatology. Notably, none of the apps were approved by the US Food and Drug Administration, and only 2 of the apps included disclaimers for the lack of regulatory approval. Overall, the study found that these apps lack supporting evidence, input from clinicians and/or dermatologists, and transparency in algorithm development, data usage, and user privacy.Conclusions and relevanceThis cross-sectional study determined that although AI dermatology mobile apps hold promise for improving access to care and patient outcomes, in their current state, they may pose harm due to potential risks, lack of consistent validation, and misleading user communication. Addressing challenges in efficacy, safety, and transparency through effective regulation, validation, and standardized evaluation criteria is essential to harness the benefits of these apps while minimizing risks.

Project description:ObjectiveTo compare fetal heart rate (FHR) patterns during the last hour of labor between small-for-gestational-age (SGA; birth weight less than the 10th percentile for gestational age) and appropriate-for-gestational-age (AGA; birth weight at the 10-90th percentile) neonates at 36 weeks of gestation or greater. We also compared the rate of cesarean delivery and composite neonatal morbidity among SGA and AGA newborns.MethodsThis is a secondary analysis of a randomized trial of intrapartum fetal electrocardiographic ST-segment analysis. We excluded women with chorioamnionitis, insufficient duration of FHR tracing in the hour before delivery, and anomalous newborns. Fetal heart rate patterns were categorized by computerized pattern recognition software (PeriCALM Patterns). Composite neonatal morbidity was defined as any of the following: intrapartum fetal death, Apgar score 3 or less at 5 minutes, cord artery pH 7.05 or less, base deficit 12 mmol/L or greater, neonatal seizure, intubation at delivery, neonatal encephalopathy, and neonatal death. Logistic regression was used to evaluate the association between FHR patterns and SGA adjusted for magnesium sulfate exposure and stage of labor.ResultsOf the 11,108 women randomized, 85% (n=9,402) met inclusion criteria, of whom 9% were SGA. In the last hour, the likelihood of accelerations was significantly lower among SGA than AGA neonates (72.4% vs 66.8%; P=.001). Variable decelerations lasting greater than 60 seconds, with depth greater than 60 beats per minute (bpm) or nadir less than 60 bpm, were significantly more common with SGA than AGA (all P<.001). The rate of late decelerations, prolonged decelerations, or bradycardia were similar between SGA and AGA (all P>.05). Cesarean delivery for fetal indications was significantly more common with SGA (7.0%) than AGA (4.0%; P<.001). The composite neonatal morbidity was 1.4% among SGA and 1.0% among AGA (odds ratio 1.40, 95% CI 0.74-2.64).ConclusionAlthough the FHR patterns in the last hour of labor differ among SGA and AGA neonates, as does the rate of cesarean delivery, the composite neonatal morbidity was similar.

Project description:BackgroundThorough dietary assessment is essential to obtain accurate food and nutrient intake data yet challenging because of the limitations of current methods. Image-based methods may decrease energy underreporting and increase the validity of self-reported dietary intake. Keenoa is an image-assisted food diary that integrates artificial intelligence food recognition. We hypothesized that Keenoa is as valid for dietary assessment as the automated self-administered 24-hour recall (ASA24)-Canada and better appreciated by users.ObjectiveWe aimed to evaluate the relative validity of Keenoa against a 24-hour validated web-based food recall platform (ASA24) in both healthy individuals and those living with diabetes. Secondary objectives were to compare the proportion of under- and overreporters between tools and to assess the user's appreciation of the tools.MethodsWe used a randomized crossover design, and participants completed 4 days of Keenoa food tracking and 4 days of ASA24 food recalls. The System Usability Scale was used to assess perceived ease of use. Differences in reported intakes were analyzed using 2-tailed paired t tests or Wilcoxon signed-rank test and deattenuated correlations by Spearman coefficient. Agreement and bias were determined using the Bland-Altman test. Weighted Cohen κ was used for cross-classification analysis. Energy underreporting was defined as a ratio of reported energy intake to estimated resting energy expenditure <0.9.ResultsA total of 136 participants were included (mean 46.1, SD 14.6 years; 49/136, 36% men; 31/136, 22.8% with diabetes). The average reported energy intakes (kcal/d) were 2171 (SD 553) in men with Keenoa and 2118 (SD 566) in men with ASA24 (P=.38) and, in women, 1804 (SD 404) with Keenoa and 1784 (SD 389) with ASA24 (P=.61). The overall mean difference (kcal/d) was -32 (95% CI -97 to 33), with limits of agreement of -789 to 725, indicating acceptable agreement between tools without bias. Mean reported macronutrient, calcium, potassium, and folate intakes did not significantly differ between tools. Reported fiber and iron intakes were higher, and sodium intake lower, with Keenoa than ASA24. Intakes in all macronutrients (r=0.48-0.73) and micronutrients analyzed (r=0.40-0.74) were correlated (all P<.05) between tools. Weighted Cohen κ scores ranged from 0.30 to 0.52 (all P<.001). The underreporting rate was 8.8% (12/136) with both tools. Mean System Usability Scale scores were higher for Keenoa than ASA24 (77/100, 77% vs 53/100, 53%; P<.001); 74.8% (101/135) of participants preferred Keenoa.ConclusionsThe Keenoa app showed moderate to strong relative validity against ASA24 for energy, macronutrient, and most micronutrient intakes analyzed in healthy adults and those with diabetes. Keenoa is a new, alternative tool that may facilitate the work of dietitians and nutrition researchers. The perceived ease of use may improve food-tracking adherence over longer periods.

Dataset Information

A mobile-optimized artificial intelligence system for gestational age and fetal malpresentation assessment.

Background

Methods

Results

Conclusions

Publications

A mobile-optimized artificial intelligence system for gestational age and fetal malpresentation assessment.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets