,
Dhanalakshmi Srinivasan University , Tiruchirappalli , India
Dhanalakshmi Srinivasan University , Tiruchirappalli , India
Down Syndrome (DS) is a genetic disorder due to a partial copy of the 21 st chromosome in the fetus, usually detected by invasive techniques like Amniocentesis and Chorionic Villus Sampling (CVS) which have chances of miscarriage. This paper suggests a completely automated and computer-aided method of non-invasive detection of DS by ultrasound fetus (USF) images, as a way of overcoming the constraints of the conventional diagnostic methods. The proposed system applies Non-Sub Sampled Contourlet Transform (NSCT) in transforming the image, and extracting features and classifying them using a custom Convolutional Neural Network (CNET). The system was trained and tested with normal and abnormal USF images with considerable results. On the Mendeley data, the system had Fetal NT Sensitivity (FNSE) of 99.24, Fetal NT Specificity (FNSP) of 99.19, Fetal NT Accuracy (FNA) of 99.23, Fetal Positive predictive rate (FPPR) of 99.28, and Fetal Negative predictive rate (FNPR) of 99.24. In the case of Kaggle Fetus (KF) dataset, the system achieved FNSE of 99.24, FNSP of 99.33, FNA of 99.1, FPPR of 99.29, and FNPR of 99.26. Mendeley and KF datasets had the average Fetus Detection Time (FDT) of 0.51 ms and 0.46 ms, respectively. The results of this study prove that the proposed method performs better in terms of detection accuracy, sensitivity, and speed and is a promising tool to conduct early non-invasive screening of the fetus and minimize the use of invasive diagnostic methods.
This is an open access article distributed under the Creative Commons Attribution Non-Commercial License (CC BY-NC) License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
The statements, opinions and data contained in the journal are solely those of the individual authors and contributors and not of the publisher and the editor(s). We stay neutral with regard to jurisdictional claims in published maps and institutional affiliations.
