Deep Learning-Based Hybrid Model for Childbirth Mode Prediction Using Maternal Health Data

The current study proposes the use of a hybrid deep learning model to predict delivery modalities (normal, caesarean, emergency, and aided delivery) from maternal health data. The model employs Convolutional Neural Networks (CNNs) for feature extraction and Long Short-Term Memory (LSTM) networks for time series analyses to learn sequential and feature-based patterns in pregnancy-related data that yield reliable predictions. The LSTM module may capture any sequential relationships in the event of pregnancy, while the CNN architecture is responsible for most feature extraction from maternal health data, contributing towards predicting delivery modality with accuracy. The model may take further advantage of Bayesian hyperparameter tuning to enhance model resilience and generalizability. XGBoost and Random Forest have been chosen as the benchmark models to evaluate them. Moreover, from the experimental data, the newly proposed hybrid LSTM-CNN model seems to outperform conventional machine-learning techniques with a much more accurate prediction capability. The conclusion regarding the improved performance of the model is corroborated through metrics such as recall, accuracy, precision, and F1 measures. Hence, through providing more accurate and reliable childbirth predictions, this research posits hybrid deep learning models to be capable of predicting maternal health outcomes and endorses their feasibility in a clinical setting.