Pulmonary nodule detection and segmentation method in virtual medical treatment based on Mask-RCNN deep learning

Abstract

The invention relates to a pulmonary nodule detection and segmentation method in virtual medical treatment based on Mask-RCNN deep learning and belongs to the field of image processing. The method specifically comprises the following steps: S1, establishing a training sample: firstly, preprocessing a three-dimensional lung CT image sample, then synthesizing a cross section, a sagittal plane and acoronal plane of a pulmonary nodule into a three-channel picture to obtain a training sample set, and finally, expanding the sample set by adopting a data enhancement method; S2, establishing a pulmonary nodule segmentation network; the method comprises the steps of establishing a backbone network, a feature pyramid network, a region generation network, an ROI generation and alignment network andthree function branches; S3, training a pulmonary nodule segmentation network: training the pulmonary nodule segmentation network by using the training sample to obtain a pulmonary nodule segmentationdevice; and S4, reconstructing a pulmonary nodule and a pulmonary three-dimensional diagram in the virtual medical environment to realize detection of the pulmonary nodule. According to the method, the accuracy of the model can be improved under the condition of not consuming more resources, and real-time interaction is realized.

Description

technical field [0001] The invention belongs to the field of image processing, relates to the fields of medical image analysis and computer vision, in particular to a method for detecting and segmenting pulmonary nodules in virtual medical treatment based on Mask-RCNN deep learning. Background technique [0002] Lung cancer is the cancer with the highest mortality rate among various cancers, among which the mortality rate of lung cancer is 13% for men and 19.5% for women. About 70% of patients are diagnosed at the advanced stage of lung cancer, and the 5-year survival rate in this case is only about 16%. However, if lung cancer is diagnosed early, the five-year survival rate can reach 70%. Studies have shown that 75% of lung cancers have been shown in early images, and lung nodules are the early forms of lung cancer, which also makes the detection of lung nodules very important. At the same time, CT images have the characteristics of high resolution and high contrast of an...

AI Technical Summary

Problems solved by technology

[0006](2) Deep learning needs the support of a large number of samples, but there is a lack of high-quality pulmonary nodule image annotation data

However, they can only obtain two-dimensional tomographic images, which only display pathological information at a single level. Medical personnel can only estimate the size and shape of the region of interest based on past experience, and it is difficult to have an intuitive understanding. Three-dimensional visualization technology The use of the 2D sequence generated by CT, MRI and other equipment can be drawn into a 3D medical model through a series of reconstruction operations and expressed in the window, so that the complex spatial characteristics and complex spatial characteristics of human organs or tissues can be understood more clearly and intuitively. Mutual positioning relationship

Images