Research On Medical Image Registration By The ICP Algorithm

With the development of the medical imaging technology, various medical image modescan be obtained. It is often necessary to combine the medical images which come from differentmodes, different times or different imaging conditions. Fusing two or multiple different medicalimages to one image can combine the advantages of different images so that the image containsthe richer information and the higher diagnostic value. However, the spatial locations of theimages which come from the different modes and conditions are not often consistent. Andtherefore images can not be fused directly. The image registration should be conducted in theearly stage. Therefore, the image registration is the base of image fusion. Image registrationtechniques have important application in the medical image processing.Iterative Closest Point（ICP）algorithm is based on the feature points. When the distance ofthe points is the nearest by continuously iterating, the spatial position of the set of points reachesconsistent. In this case, spatial transformation parameters are the image transformationparameters. The ICP algorithm enables the distances of the image locations to become closequickly. If the nearest distance reaches a certain extent value, the value will no longer beimproved. So the immune algorithm is used to improve the accuracy of the registration. Theimmune algorithm has very important applications in the objective function optimization. Thealgorithm is able to achieve a desired accuracy by setting the parameters. However, theconvergence rate is very slow for optimizing the distance function because of having noexperience values. So there is no practical significance. ICP algorithm is selected to conductearly steps of the registration at first. Then immune algorithm is used to optimize the distancefunction on the base of the ICP algorithm. Satisfactory results are achieved at last.The experimental results show that ICP algorithm has the fastest convergence rate. It adjuststo the registration of both MR-CT images and T1-T2images. It is very stable for the medical images which have a little difference in the spatial location. When the accuracy reaches a certainvalue, it can not be improved. It is the only drawback of the ICP algorithm. The disadvantage ofthe immune algorithm is that the convergence rate is too slow. But it can achieve the higheraccuracy. The ICP algorithm combines the two algorithms. The calculation speed of the immuneICP algorithm is faster than the calculation speed of the immune algorithm. The calculationaccuracy of the immune ICP algorithm is higher than the calculation accuracy of the ICPalgorithm.