Ultrasonic Marker Pattern Recognition and Measurement Using Artificial Neural Network

Ultrasound screening is performed during early pregnancy for assessment of fetal well being and prenatal diagnosis of fetal chromosomal anomalies including measurement of nuchal translucency (NT) thickness. The drawback of current NT measurement technique is restricted with inter and intro-observer variability and inconsistency of results. Hence, we present an automated detection and measurement method for NT in this study. Artificial neural network was trained to locate the region of interest (ROI) that contains NT. The accuracy of the trained network was achieved at least 93.33 percent which promise an efficient method to recognize NT automatically. Border of NT layer was detected through automatic computerized algorithm to find the optimum thickness of the windowed region. Local measurements of intensity, edge strength and continuity were extracted and became the weighted terms for thickness calculation. Finding showed that this method is able to provide consistent and more objective results. Key-Words: nuchal translucency, ultrasound, fetal, pattern recognition, artificial neural network