Deep Learning with Modified Loss function to Predict Gestational Age of the Fetal Brain

TL;DR: This study uses deep learning to classify and predict fetal gestational age through MRI images, and changes the cross-entropy loss function of the network model to Fnlloss and Flloss to have a better improvement in the accuracy of classification.

Abstract: Fetal magnetic resonance imaging (MRI) sequences are often used for fetal development judgment and disease diagnosis. In clinical practice, doctors usually rely on the naked eye to identify and make corresponding judgments based on the shape, depth and appearance time of the sulci and gyri between different gestational ages. This study uses deep learning to classify and predict fetal gestational age. Because ultrasound detection is accurate in the first trimester, and after this time, it is impossible to accurately judge and predict the corresponding gestational age, so this is a problem that we need to solve. We use MRI images to classify and predict the gestational age of the next three months. We select 13 gestational ages of the coronal sequence of 22 to 34 as the classification prediction data. For adequate training, we changed the cross-entropy loss function of the network model to Fnlloss and Flloss we proposed. The model after modifying the loss function has a better improvement in the accuracy of classification compared to the original cross-entropy loss used by the network model