Spine segmentation method based on 3D full convolution neural network

Abstract

The invention discloses a spine segmentation method based on a 3D full convolution neural network. The method comprises the following steps: 1, a 3D full convolution neural network for spine segmentation is built; 2, data sets are prepared, wherein the data sets comprise a training set plus a test set; 3, the training set is used to train the neural network to obtain a network model; and 4, the obtained network model is used to segment spine CT data in the test set to obtain a segmentation result. With the help of the 3D full convolution neural network, the spine is segmented, full automationis achieved, and the segmentation precision is high.

Description

technical field [0001] The invention belongs to the field of medical image processing, and specifically relates to a spine segmentation method based on a 3D full convolutional neural network. Background technique [0002] The spine forms an important support structure in the human body, consisting of 7 cervical, 12 thoracic and 5 lumbar vertebrae. Due to reduced physical activity and modern sedentary office work, spinal problems are becoming more and more serious problems in modern society. Common such as idiopathic scoliosis, severe scoliosis will affect the health of the human body, compress the heart and lung function, and often require surgery correction. [0003] Spine segmentation plays an important role in spinal surgery. Before spinal surgery, the doctor needs to make a treatment plan based on the patient's image data, such as determining the implantation angle and depth of the pedicle screw, which can be determined by measuring the aggregate information of the spi...

AI Technical Summary

Problems solved by technology

Traditional segmentation methods have problems such as heavy workload, time-consuming, and a lot of preprocessing.

Images