CT image skeleton segmentation method and device based on convolutional neural network

Abstract

The invention discloses a CT image skeleton segmentation method and device based on a convolutional neural network, and relates to the technical field of computer vision and deep learning. The method comprises the following steps: acquiring CT image data and inputting the CT image data into a skeleton recognition model; performing convolution operation processing on the CT image data to obtain a feature output graph C1; performing multiple composite operation processing on the feature output graph C1 through a bottleneck module to obtain a feature output graph CN; performing convolution operation processing on the feature output graph CN to obtain a feature output graph CN + 1; inputting the feature output graph CN + 1 into a plurality of pooling cores for average pooling operation to obtain a feature output graph PL; performing stacking operation on the plurality of feature output graphs PL to obtain a fusion feature PF1; performing stacking operation on the fusion feature PF1 and a feature output graph CN + 1 to obtain a fusion feature PF2; and carrying out convolution operation on the fusion feature PF2 to obtain a segmentation image S. According to the method, skeleton segmentation can be accurately and quickly performed on the CT image, the time cost is saved, and the requirement on the GPU is not high.

Description

technical field [0001] The invention relates to the technical fields of computer vision and deep learning, in particular to a method and device for bone segmentation of CT images based on a convolutional neural network. Background technique [0002] Automatic bone removal on CT images is an important step for doctors to evaluate patients' CT images. The bones in CT images will affect the doctor's judgment on vascular diseases. At present, the bone segmentation method is realized by the method in traditional digital image processing, that is, the threshold value is determined first, and then the binary image is segmented. This method is ideal for segmentation results only in special cases, and the segmentation results are not accurate when the blood vessels are densely distributed. [0003] In the prior art, two CT scans are generally used for CT image bone removal, one is a normal CT scan, and the second is a contrast CT scan, and the bone is marked with a contrast agent to...

AI Technical Summary

Problems solved by technology

However, when a doctor consults a patient's CT image, because blood and bone are relatively similar in CT images, the effect of using traditional digital image processing methods to separate bone and blood is not ideal.

However, the existing deep learning method has a slow calculation speed and cannot achieve fast segmentation in a short time.

Considering that large-scale networks also have certain requirements for GPU, it will lead to problems such as high cost

Images