Labor Induction Failure Prediction Based on B-Mode Ultrasound Image Processing Using Multiscale Local Binary Patterns

Labor induction is defined as the artificial onset of labor for the purpose of vaginal birth. Cesarean section is one of the potential risks of labor induction as it occurs in about 20% of the inductions. A ripe cervix (soft and distensible) is needed for a successful labor. Changes occurring during the ripening process, will affect the interaction between cervical tissues and sound waves during ultrasound transvaginal scanning and will be perceived as gray level intensity variations in the echography image. Thus, a non-invasive method using image processing of ultrasound images may help in predicting the outcome of labor induction. In this paper a set of echography images from labor induction patients is analyzed using a multiscale methodology based on Local Binary Patterns and circular Gabor filters. Results show that it is feasible to predict the outcome of a labor induction procedure using this method with a good accuracy.