Neurosono-net: deep learning for the detection of fetal brain abnormalities using ultrasonography

Abstract

Diagnosis of fetal brain abnormalities through the use of ultrasound is important because congenital brain malformations are one of the most common anomalies. The prevalence of these abnormalities is unknown, as some cases do not manifest until later in life and most occur in people who are not in risk groups. Many of these anomalies have no available treatment and can be severe and permanent, so it is important to make an accurate diagnosis and prognosis in order to provide multidisciplinary advice and management. The objective of this work is to implement a deep learning architecture that uses ultrasound images of the fetal brain for the automatic detection of abnormalities in standard planes. For this we will use the Faster R-CNN architecture, which could be very useful for remote places where there is no professional specialist in ultrasound; in turn, it can also serve as support for the ultrasound training of professionals who perform this procedure.