Study On The Registration Of Thoracic Aorta CTA And X-ray Fluoroscopic Images

Thoracic endovascular aortic repair(TEVAR)has become a reliable method for the treatment of thoracic aortic diseases by implant the stent in a minimally invasive method.However,during surgery,doctors could not obtain clear aortic shape and location from intraoperative X-ray images,but preoperative CTA images can well reproduce the location of aortic lesions.Therefore,by registration the preoperative CTA images and intraoperative X-ray images,a clear preoperative vascular structure is overlaid into the intraoperative X-ray images,which can provide advanced image navigation for stent implantation and improve the safety and success rate of the surgery.Based on this background,this paper proposes a preoperative CTA images and intraoperative X-ray image registration algorithm based on Digitally Reconstructed Radiograph(DRR)library.Specific work content and innovations include:(1)The registration first needs to generate a two-dimensional DRR image by projecting the 3D CTA images,and then register the X-ray image and the DRR image in the same dimension.Therefore,the digitally reconstructed radiograph operation of 3D CTA images and the optimized search of similarity measure of X-ray image and DRR image are two key steps in the registration.(2)For the time-consuming projections of 3D volume data into 2D images,a preoperative CTA images and intraoperative X-ray fluoroscopic image registration algorithm based on the superimposed DRR library is proposed.First,high-density sampling is performed on the projection parameters that cause shape changes of the intraoperative X-ray fluoroscopic image.Using the series of projection deformations obtained from the sampling,the digitally reconstructed radiograph operation of the CT full-image and the segmented CT bone structure images is performed,and the projection results are superimposed to generate a superimposed DRR library.Then,the DRR image with the most similarity measure to the X-ray image is searched in the superposition DRR library,and the final spatial transformation parameters are calculated to complete the registration of the CTA images and the X-ray image.The method based on the superimposed DRR library avoids the operation of projecting the 3D CTA images,which takes an average of 3.58 seconds and improves the registration efficiency.(3)For the optimization search of similarity measure,an improved hill-climbing strategy is proposed.Combining the actual clinical operation rules of the C-arm machine during surgery,the projection angles that cause the most obvious change in the shape of the X-ray image are sampled at equal intervals.The interval distance is the initial optimization step of the hill-climbing algorithm,which avoids the mis-registration cause by the similarity measure falling into the local extremum.The experimental results on real patient data demonstrate that our method achieves higher efficiency compared to traditional algorithms and achieves favorable registration of preoperative CTA images and intraoperative X-ray images.