Carpal bone region segmentation method based on shape information

Abstract

A carpal bone region segmentation method based on shape information comprises the following steps: preprocessing X-ray images in sequence by adopting star median filtering and grayscale image binarization methods; determining the vertical coordinate position of the upper boundary of the carpal bone area by using a bidirectional iterative scanning algorithm according to the distribution characteristics of the carpal bone area; according to the wrist joint shape information, calculating gradient changes of edges on two sides of a wrist, and determining vertical coordinate positions of left and right candidate points of a lower boundary; integrating the upper boundary, the left candidate point, the right candidate point and the carpal bone area information, optimizing and correcting the coordinate position, and cutting the carpal bone area according to the obtained coordinates of the upper left corner and the lower right corner. According to the method, the carpal bone area in the wrist image is accurately segmented; when the upper boundary is determined, a bidirectional iterative scanning algorithm is adopted, so that the scanning rate is greatly improved; when the lower boundary isdetermined, the situation of the left boundary point and the right boundary point is considered at the same time, optimization selection is performed, and the influence of different palm joint positions on the positioning accuracy is weakened.

Description

technical field [0001] The invention relates to a carpal region segmentation method based on shape information. Background technique [0002] There are many methods for bone age assessment. The earliest bone age counting method in the early 20th century mainly observes the number of ossification centers at a certain age, mainly distributed in the carpal area. Later, when X-rays were discovered and put into the study of skeletal development, many experts successively proposed and revised more accurate skeletal maturation maps. Then there is the atlas method. During the evaluation, the X-ray film to be tested is compared with the atlas one by one, and the most similar one is taken as the bone age. Recently, with the wide application of artificial intelligence technology in the medical field and the major breakthrough of deep convolutional neural network in image processing technology, more and more scholars study bone age from the perspective of machine learning. [0003] At...

AI Technical Summary

Problems solved by technology

[0005]However, since the common X-ray images of human wrists do not have strict requirements on shooting positions, the wrist shapes of different people are different, and the palm joints are placed incorrectly. The image processing method brings many difficulties

For example, according to the characteristics of the shortest transverse width of the wrist joint to locate the caudal region of the wrist bone, a huge error will be introduced because some groups do not meet this characteristic

The wrist bone area predicted by the neural network depends largely on the training samples. For newly collected images with very different chromaticity and brightness, the prediction accuracy will be greatly reduced.

At the same time, due to the characteristics of automatic positioning, the size of the final carpal bone area is uncontrollable

Images