A medical image registration method based on super pixel guide

Abstract

The invention discloses a medical image registration method based on a superpixel guide, comprising the following steps: 1) rigidly pre-registering a reference image and a floating image by adopting an FFD registration algorithm based on a B-spline to obtain a pre-registered reference image and a floating image; 2) downsampling the pre-registered reference image and the floating image to form a multi-resolution pyramid with a series of l; 3) the image registration of the multi-resolution scheme starts from the coarsest resolution, and gradually registers from coarse to fine. Under the currentresolution l, after iteration, the obtained spatial transform ul is up-sampled to obtain ul-1, and then use the space transformation ul Pairs of Pyramids 1-1 level image is transformed, and the obtained image is iterated again until the end of the lowest layer of the pyramid image; 4) the final obtained spatial transformation u is applied to the floating image after rigid registration to obtain the final registration result image. The invention obtains the registration accuracy and robustness superior to the current mainstream DIR registration models such as the FFD model and the DD model based on the B-spline.

Description

technical field [0001] The invention relates to the technical field of medical image processing and application, in particular to a medical image registration method based on a superpixel guide. Background technique [0002] Lung cancer is one of the malignant tumors with the fastest increasing morbidity and mortality in the world and the greatest threat to human health and life. Registration between chest follow-up CT (Computed Tomography, CT) images after lung cancer resection is to predict postoperative lung function, monitor postoperative tumor recurrence and metastasis, evaluate postoperative residual lung compensation effect, judge prognosis and guide treatment. other key indicators. However, due to the inherent deformability of the lung, coupled with the characteristics of organ peristalsis, respiratory motion, and patient movement, there are complex and large deformations between chest follow-up images, and their precise registration remains a difficult task. In re...

AI Technical Summary

Problems solved by technology

However, different doctors may produce different delineation results, and even the same doctor may delineate different contours at different time points, so the accuracy and consistency of the registration results are greatly affected by the doctor's subjective experience, and the manual delineation process takes a long time , analyzing large data sets can be very laborious

Images