Method and system for automatically detecting ultrasound image of fetal head circumference

Abstract

The invention provides a method and a system for automatically detecting an ultrasound image of a fetal head circumference. The method comprises the following steps: acquiring the ultrasound image of the fetal head circumference, and acquiring a corresponding binary image; acquiring all curve arc sections of the binary image, and forming a curve arc section set by the acquired curve arc sections; using each curve arc section in the curve arc section set as a base arc section in sequence; classifying other curve arc sections, which are different curve arc sections on the same ellipse with respect to the current base arc section, into one group, to form curve arc section groups; acquiring the curve arc section group with a maximum ellipse matching rate; judging whether the maximum ellipse matching rate is greater than a preset ellipse matching rate threshold of the system or not, if so, determining an ellipse matched with the current curve arc section group as a fetal head circumference ellipse, and acquiring a perimeter of the fetal head circumference according to the fetal head circumference ellipse; otherwise, judging that a matching result is wrong, and acquiring the ultrasound image of the fetal head circumference again. The method and the system for automatically detecting the ultrasound image of the fetal head circumference are simple in calculation, accurate in detection and high in robustness.

Description

technical field [0001] The invention belongs to the technical field of medical ultrasonic diagnosis, and mainly relates to an automatic detection method and a detection system for fetal head circumference ultrasonic images. Background technique [0002] Ultrasound imaging has become one of the most widely used clinical diagnostic tools because of its non-invasive, real-time, convenient operation, and low price. Ultrasound color flow imaging is an important imaging mode in ultrasound imaging, and it is widely used in clinical practice. [0003] Ultrasound examination has no radiation damage and is often used for prenatal diagnosis of fetuses. For the measurement of fetal signs, doctors mainly perform manual measurements on ultrasound images. This process is cumbersome and time-consuming, and is related to the experience of imaging doctors. Therefore, the automatic measurement of the fetal head circumference has attracted extensive attention. Correspondingly, various implemen...

AI Technical Summary

Problems solved by technology

[0004] One embodiment proposes to detect the head circumference ellipse through random Hough transform. Although the random Hough transform has certain advantages in calculation and memory requirements compared with the traditional Hough transform ellipse detection, it has no target for images with complex backgrounds. The random sampling will introduce a large number of invalid samples, the detection results will be highly random, and the performance of the algorithm will be reduced.

[0005] One embodiment proposes ellipse detection based on arc segmentation. This method is prone to errors in combination of arcs; especially in the case of complex background and missing boundary parts such as ultrasound fetal head circumference, automatic measurement cannot be accurately performed.

[0006] One embodiment proposes to use non-iterative ellipse fitting for ellipse determination. This method first constructs a classifier based on random forest algorithm or Adaboost to detect the ROI of the region of interest around the head circumference, and then detects the edge of the detected ROI through local phase symmetry. , and finally the ellipse fits the head circumference ellipse. During the implementation of this method, a large amount of data is required to construct a classifier, and the effect of the classifier will directly affect the measurement results. This method increases the average speed of automatic measurement to about 1 second / frame, although it is relatively slow The performance of the first two implementations is improved but the real-time performance is still insufficient

Images