RESUMEN
BACKGROUND: Birthweight is an important indicator during the fetal development process to protect the maternal and infant safety. However, birthweight is difficult to be directly measured, and is usually roughly estimated by the empirical formulas according to the experience of the doctors in clinical practice. METHODS: This study attempts to combine multiple electronic medical records with the B-ultrasonic examination of pregnant women to construct a hybrid birth weight predicting classifier based on long short-term memory (LSTM) networks. The clinical data were collected from 5,759 Chinese pregnant women who have given birth, with more than 57,000 obstetric electronic medical records. We evaluated the prediction by the mean relative error (MRE) and the accuracy rate of different machine learning classifiers at different predicting periods for first delivery and multiple deliveries. Additionally, we evaluated the classification accuracies of different classifiers respectively for the Small-for-Gestational-age (SGA), Large-for-Gestational-Age (LGA) and Appropriate-for-Gestational-Age (AGA) groups. RESULTS: The results show that the accuracy rate of the prediction model using Convolutional Neuron Networks (CNN), Random Forest (RF), Linear-Regression, Support Vector Regression (SVR), Back Propagation Neural Network(BPNN), and the proposed hybrid-LSTM at the 40th pregnancy week for first delivery were 0.498, 0.662, 0.670, 0.680, 0.705 and 0.793, respectively. Among the groups of less than 39th pregnancy week, the 39th pregnancy week and more than 40th week, the hybrid-LSTM model obtained the best accuracy and almost the least MRE compared with those of machine learning models. Not surprisingly, all the machine learning models performed better than the empirical formula. In the SGA, LGA and AGA group experiments, the average accuracy by the empirical formula, logistic regression (LR), BPNN, CNN, RF and Hybrid-LSTM were 0.780, 0.855, 0.890, 0.906, 0.916 and 0.933, respectively. CONCLUSIONS: The results of this study are helpful for the birthweight prediction and development of guidelines for clinical delivery treatments. It is also useful for the implementation of a decision support system using the temporal machine learning prediction model, as it can assist the clinicians to make correct decisions during the obstetric examinations and remind pregnant women to manage their weight.
Asunto(s)
Recién Nacido Pequeño para la Edad Gestacional , Aprendizaje Automático , Peso al Nacer , Femenino , Humanos , Recién Nacido , Modelos Logísticos , Redes Neurales de la Computación , EmbarazoRESUMEN
Objective: The aim of the study was to investigate whether the accuracy of ultrasound estimates of fetal weight (EFW) was dependent on maternal obesity.Study design: A prospective cross-sectional study of 1064 singleton pregnant women classified according to body mass index (BMI) into two categories: normal (BMI < 25 kg/m2, n = 863) and obese (BMI ≥ 35 kg/m2, n = 201) was conducted. EFW were calculated using Hadlock's formula, and the difference between EFW and the actual birthweight (absolute percent error) was analyzed in both groups. Spearman's correlation was used to assess the relationship between ultrasound performance (absolute error), maternal BMI, and actual birth weight.Results: Median absolute error of sonographic EFW was 5.90 and 6.47% for the normal and obese groups, respectively (p .38). A correlation between EFW and birth weight (BW) was found in both groups, r = 0.755 (p < .001) and r = 0.753 (p < .001), respectively. The correlation between absolute error, maternal BMI, and fetal birth weight was poor.Conclusions: Maternal obesity is unrelated to the accuracy of sonographic EFW, and regardless of maternal or fetal size, ultrasound is currently an accurate method of prediction for both obese and normal weight pregnant women.
Asunto(s)
Obesidad Materna , Peso al Nacer , Estudios Transversales , Femenino , Peso Fetal , Edad Gestacional , Humanos , Embarazo , Estudios Prospectivos , Ultrasonografía , Ultrasonografía PrenatalRESUMEN
OBJECTIVE: To develop and validate birthweight prediction models using fetal fractional thigh volume (TVol) in an Indian population, comparing them with existing prediction models developed for other ethnicities. METHODS: A prospective observational study was conducted among 131 pregnant women (>36 weeks) attending a tertiary hospital in New Delhi, India, for prenatal care between December 1, 2014, and November 1, 2016. Participants were randomly divided into formulating (n=100) and validation (n=31) groups. Multiple regression analysis was performed to generate four models to predict birthweight using various combinations of two-dimensional (2D) ultrasonographic parameters and a three-dimensional (3D) ultrasonographic parameter (TVol). The best fit model was compared with previously published 2D and 3D models. RESULTS: The best fit model comprised biparietal diameter, head circumference, abdominal circumference, and TVol. This model had the lowest mean percentage error (0.624 ± 8.075) and the highest coefficient of determination (R2 =0.660). It correctly predicted 70.2% and 91.6% of birthweights within 5% and 10% of actual weight, respectively. Compared with previous models, attributability for the 2D and 3D models was 0.65 and 0.55, respectively. Accuracy was -0.05 ± 1.007 and -2.54 ± 1.11, respectively. CONCLUSION: Models that included TVol provided good prediction of birthweight in the target population.