Predicting Fetal Weight and Pregnancy Advisory Using Deep Learning Algorithm

For the benefit of the unborn kid and the mother, controlling pregnancy and making correct fetal weight predictions are essential. To address this, this study, proposes a deep learning model that effectively combines multimodal data, such as text-based medical records and ultrasound images to predict fetal weight and provide personalized pregnancy advisory services. The model is trained on a sizable dataset that includes information from the mother's clinical evaluation and images from fetal ultrasounds that were gathered from Nigst Elleni Hospital in Hossana and the Ethiopian Demographic and Health Survey 2016, respectively. We ran tests on a real-world dataset of expectant mothers and compared several models to assess the effectiveness of our suggested framework. In this study, we used multiclass target classes. The results revealed that combining CNN, LSTM, MLP, Autoencoder, or DBN models significantly improved accuracy in classifying and predicting fetal weight. With scores of 92.3%, 93.4%, 92.2%, and 92.8%, respectively, the CNN and LSTM fusion model coupled yielded the best results in terms of accuracy, precision, recall, and F1 score. Additionally, this model demonstrated the lowest MAE, MSE, RMSE, and the highest R2, measuring 0.45, 0.34, 0.58, and 0.5, respectively.