Deep Learning Based Medical Image Segmentation in Prenatal Prediction of Neurodevelopmental Disorders

Authors:
Publisher website
Google Scholar
Abstract
The onset of autism related neurodevelopmental disorders happening in the embryonic stage itself. The effectiveness of treatment is greatly dependent on the early detection. Deep learning has made significant strides in the field of artificial intelligence. To perform semantic segmentation on the ultrasound image of the developing embryo, this research attempts to design a framework based on 3D Convolutional Neural Network, a deep learning model. The segments of the fetal brain can be more effectively separated by combining many convolutional neural networks with different configurations. Transfer learning can be used to lessen the paucity of data by using pretrained models. Two CNNs with different configurations can process a single image, and fusion is used to integrate the two feature maps. The normalization of the image takes place as a pre-processing step. The performance is measured with the evaluation parameters loss, training accuracy and testing accuracy.

This publication has 11 references indexed in Scilit: