Identification of nasal bone for the early detection of down syndrome using Back Propagation Neural Network

Abstract

Down Syndrome is characterized by the absence of nasal bone during the late first trimester of pregnancy. Presently Downsyndrome is identified by visually examining the ultra sonogram of foetus of 11 to 13 weeks of gestation for the presence of nasal bone. Nasal bone is visually identified by differentiating the change in the contrast of nasal region of ultrasonogram. This method is prone to operator error as the nasal bone is a very small physical structure during the first trimester of pregnancy. Noise also introduce error during visual identification. This paper presents a new approach for the detection of nasal bone for ultrasonogram of foetus of 11 to 13 weeks of gestation. The proposed method is based on the extraction of image texture parameter of nasal bone region of ultra sonogram and their subsequent classification using Back Propagation Neural Network (BPNN). The features in the nasal region are extracted in the Spatial domain and Transform domain using Discrete Cosine Transform (DCT) and Daubechies D4 Wavelet transform. These features are extracted from images with nasal bone and images which don't have nasal bone. The extracted data is normalized and used to train Back Propagation Neural Network (BPNN). The trained BPNN is used to classify random ultrasonograms. The result shows that the proposed method can detect down syndrome with higher degree of accuracy. This method combined with the present detection methods can reduce operator error and overall enhance the down syndrome detection rate by analysing ultra sonogram.