Wrist reference bone segmentation method

Abstract

The invention discloses a wrist reference segmentation method, which comprises the following steps of 1, extracting and preprocessing a reference bone region of interest image; 2, segmenting the reference bone region of interest image; and 3, post-processing the reference bone region of interest. According to the scoring method, the bone age is calculated by evaluating the mature indications of the reference bone combination, and the bone age calculated through the method is small in deviation. Due to the influence of the factors, such as the irregular shapes of the reference bones, the low contrast ratio of the reference bones and the soft tissues, the adjacent positions of the reference bones, etc., difficulty is increased for the experts to accurately evaluate the maturation indicationsof the reference bones. The reference bone segmentation accuracy is guaranteed, and meanwhile it is guaranteed that the reference bone segmentation method has higher universality. The method comprises the following steps of firstly, extracting a reference bone ROI image by adopting a Faster R-CNN; then, inputting the preprocessed reference bone ROI image into the ARU-Net to be segmented, adding the residual error mapping and an attention mechanism to the ARU-Net on the basis of traditional U-Net to improve the accuracy of reference bone segmentation and the utilization rate of features; and finally, eliminating the phenomena of edge sawteeth, holes and the like in a prediction result by adopting a flooding filling algorithm and the corrosion and expansion operation.

Description

technical field [0001] The present invention relates to a wrist reference bone segmentation method. Background technique [0002] The scoring method calculates skeletal age by assessing the maturation indications of the reference bone assembly, and the skeletal age calculated by this method has a small deviation. Due to the irregular shape of the reference bone, the low contrast between the reference bone and soft tissue, and the adjacent position of the reference bone, it is more difficult for experts to accurately evaluate the maturity indicators of the reference bone. How to accurately segment the reference bone from the wrist is a challenge. [0003] Most of the current common reference bone segmentation methods are based on threshold, region and edge. When the interval between the reference bones is large, these algorithms can accurately segment the reference bones, but when the interval between the reference bones is small or even fused, the extraction accuracy of th...

AI Technical Summary

Problems solved by technology

When the interval between the reference bones is large, these algorithms can accurately segment the reference bone, but when the interval between the reference bones is small or even fused, the extraction accuracy of these algorithms will drop sharply

Images