Automatic detection and measurement of fetal biometrics to determine the Gestational Age

An Ultrasonography (USG) feature extraction and classification system has been developed for detecting and analyzing organ structure in fetal body. This study aims to detect and measure features of fetus from scanned image of fetal head (biparietal diameter) and femur length. We propose Haar-like feature is used to extract feature from cropping image object, meanwhile AdaBoost classifier is used for object detection and Randomized Hough Transform is applied for biometry measurement. In this research, we used 300 biparietal head image data and 200 image data of femur. After the data processing stage, we obtained the detection and measurement of biparietal as many as 44 images with an average error of 0.039 and Correlation Coefficient result of 0.984, while the results for the detection and measurement of fetal femur error as many as 18 with an average of 0.101 and Correlation Coefficient result of 0.763. Predicting Gestational Age(GA) using our method are match with manually predicted GA, average error of predicting GA with BPD is 0.005, and with femur length is 0.033. The result of this research can be optimized further to realize a fully integrated system that can detect and measure fetal organ with usable user interaction and affordable price.