Analysis and Identification of Spina Bifida Using Convolution Neural Network for Ultrasound Image of the Fetus

Abstract

Diagnosis of spina bifida may occur prenatally through ultrasound imaging or maternal serum screening, which can detect signs of the condition in utero. Postnatal diagnosis is confirmed through imaging studies such as ultrasound, MRI, or CT scans. Treatment and management of spina bifida depend on the severity of the condition and may include surgical repair of the spinal defect shortly after birth, management of associated complications such as hydrocephalus, Chiari malformation, and bladder and bowel dysfunction, as well as on going rehabilitation and support services to optimize function and quality of life. newlineA birth abnormality known as spina bifida occurs when the spinal cord malfunctions during pregnancy, potentially resulting in permanent paralysis. Early detection and treatment are key for successful management; ultrasound imaging is the preferred method for monitoring fetuses. Here, two deep-learning algorithms were used to segment and classify ultrasound images of the fetal spine into normal and pathological categories. Three-dimensional ultrasound exams were performed on 300 pregnant women at a hospital between November 2015 and November 2020. The pictures were cleared of noise using an adaptive bilateral filter (ABF). The dilated encoder-decoder (DED) method was used to segment images of fetal spina bifida, and a feature map-based differential convolutional network (FMDCN) was used to classify the images. The results revealed that the developed model effectively segmented and classified the spinal disorders in the Fetus by valuable detection tools newline