Ultrasonic imaging systems are in an important position in medical imaging due to their advantages of noninvasive, non-radiation to humans, low cost and real-time, and are increasingly used in medical diagnosis, especially in obstetric examinations. It is widely used.However, the physical characteristics of the ultrasonic imaging hardware cause the signal-to-noise ratio of the ultrasonic image to be low, which has an adverse impact on the analysis of the image.At present, the feature extraction and analysis methods for ultrasonic images are still lagging behind the development of ultrasonic imaging technology itself. Therefore, it is of great practical significance to introduce automated intelligent ultrasonic image analysis schemes in ultrasonic image analysis in order to achieve automatic identification and measurement of targets.
Obstetric ultrasound image analysis is an important means to evaluate the intrauterine growth and development of the fetus. Measuring the head circumference of the fetus through ultrasound images and further estimating the gestational age and fetal weight of the fetus is one of the important indicators to measure the intrauterine growth and development of the fetus.At present, manual measurement is mainly used in ultrasonic two-dimensional image measurement. The tester manually delineates the contour of the fetal head circumference by moving the trackball. This method is repetitive and boring, time-consuming and laborious, and is affected by random errors and the tester's own visual errors. The accuracy of the measurement is unsatisfactory.The ultrasonic image with a low signal-to-noise ratio as shown in the figure below is a huge test for manual and accurate measurement.
First of all, in the early stages of the problem, we visited many clinicians and learned about their common operating steps when manually measuring the fetal head circumference, that is, first locate the upper end of the fetal head circumference, then locate the lower end of the fetal head circumference, and finally roll the trackball to determine the ratio of the long and short axis of the ellipse to complete the tracing of the outline of the fetal head.Inspired by this, this scheme subdivides the analysis steps into three steps: locating the upper end of the fetal head circumference, locating the lower end of the fetal head circumference, and determining the ratio of the long and short axis.
First, the two-dimensional characteristic information of the endpoints on the fetal head circumference is generated, and the maximum frequency response of this characteristic information is searched in the ultrasound image and the coordinates are determined.
Secondly, the coordinates of the lower end point are determined according to the method of step 1. On this basis, we can determine the rotation angle of the fetal head.
Again, the angle of the head is corrected and the ultrasound image is corrected.
Finally, on the more positive image, the fetal head circumference is automatically measured by combining the horizontal and vertical axes.
The effect of this intelligent recognition scheme is as shown in the figure below
