A SAM-Residual U-Net Pipeline for Segmentation and Classification of Down Syndrome in Fetal Ultrasound Images

Down Syndrome is not a single disease but it is a group of diseases. It is a chromosomal defect which affects whole life of child/person. Therefore, it is crucial and very important task to detect the syndrome in early age of fetus so that parents can take decision as per Doctor’s(Gynecologist’s) advise. There are two techniques used for diagnosis of Down Syndrome:(a)invasive technique (b)non-invasive technique. Invasive technique includes various surgery or biopsy test means this screening is done through dissecting or puncturing skin of the body(fetus mother’s body). This can make complications during pregnancy like infection, bleeding and tissue damage. Non-invasive technique includes an screening without invading the body. It is safer and has low risk compared to invasive screening e.g- Ultrasound imaging, Magnetic Resonance Imaging, Electrocardiography and blood test(serum screening or cell-free DNA testing) to provide a scale for the fetus having Down Syndrome without direct touching fetal tissues. Here the proposed paper refers the non-invasive technique that is Ultrasound Imaging technique. In this paper there is an endeavor to segment the thickness of a fluid(NT-Nuchal Translucency) found behind the fetus neck and nasal bone availability using Segment Anything technique and classify whether an individuals with Down syndrome or not in ultrasound image of fetus are taken using Residual UNET. At last, we found overall accuracy 98%.