Automatic Fetal Measurements for Low-Cost Settings by Using Local Phase Bone Detection

Abstract

The estimation of gestational age is done mostly by measurement of foetal anatomical structures such as the head and femur. These measurement are also used in diagnosis and growth assessment. Manual measurements is operator dependent and hence subject to variability. The proposed method gives automatically fetal measurements which can be used to give gestational age and expected delivery date. The initial step is the detection of biometric elements in ultrasound fetal scans based on local image phase features, obtained using Log-Gabor filter banks, which are supposed to detect frequency nature of the soft tissue/bone interfaces. The second step is a regression method to relate the obtained features to the gestational age, predicting it. The lengths of 20 femurs are analyzed in our experiments. The resulting measurements are consistent with manual measurements obtained by two expert technicians. The method is fully automatic and requires little training.