A method and system for 2D and 3D medical image registration based on deep learning

Abstract

The invention discloses a two-dimensional and three-dimensional medical image registration method and system based on deep learning. First, the target lesion in the three-dimensional medical image is segmented by deep learning technology, and then the segmented image is reconstructed. According to the clinical prior information, the edge information of the lesion is obtained from the reconstructed 3D image, and several binary images are obtained. Obtain the part that the doctor needs to observe through the camera, use deep learning technology to calculate the edge of the target lesion, and obtain another binary image. Using image registration technology to register the binary images of these two sources, the transformation matrix between the two types of images is obtained, and this transformation matrix is ​​applied to the original image, which completes the image registration problem . This method enables doctors to have a pair of real-time "see-through eyes" in the process of diagnosis and treatment, better complete the medical image registration task, make full use of the patient's imaging examination information, and assist clinical treatment.

Description

technical field [0001] The invention belongs to the field of medical image registration, mainly applies deep learning technology, and integrates clinical prior knowledge to perform multi-modal and multi-dimensional medical image registration. Background technique [0002] In the medical field, medical imaging examinations can provide imaging information of internal lesions in the body, such as the shape, size, capsule, and blood supply of lesions. By reading the films, doctors integrate the information provided by imaging studies with their own knowledge, so as to diagnose and treat patients (such as surgery). In the process of surgery or treatment, doctors cannot see deep information through superficial tissues, and can only use a layer-by-layer approach to approach the lesion. This process requires doctors to have solid theoretical literacy and surgical experience. If the imaging pictures containing the three-dimensional information of the patient's lesion can be processe...

AI Technical Summary

Problems solved by technology

However, the existing work has the following deficiencies: 1. The clinical prior information is not fully utilized; 2. The clinical work is not fully integrated.

However, in the above-mentioned patents, the registration accuracy is not high, and the image containing the three-dimensional information of the lesion cannot be projected onto the two-dimensional image or the projection effect is not good. Unlike the above-mentioned invention or existing similar inventions, the present invention uses deep learning technology Extract image features, use clinical prior knowledge to reduce the amount of calculation, and disclose a method for registration of medical 3D images and 2D images

Images