Research On Multiple Phase Magnetic Resonance Image Registration And Detection Of Hepatocellular Carcinomas

The resolution of images produced by Magnetic Resonance Imaging are rather high.The intensity of images are changing during different phases after using Dynamic Contrast Enhanced Magnetic Resonance Imaging,which can help the doctors observe the lesion areas and make a diagnosis,so MRI are often used to diagnose hepatocellular carcinomas.But deformations may be occur in MRI because of patient's breathing,exercise and so on.That will make against subsequent processing of MRI and interference diagnosis of doctors,so there is a very necessary to register the MRI from different phases.In addition,there are a lot of noise in MRI of the liver and the contrast of some tumors relative to liver parenchyma are low.So it can help doctors to diagnose hepatocellular carcinomas with the detection of these weak tumors.This paper studies the registration between MRI and detection of weak hepatocellular carcinomas on the basis of the former.The main research content of this paper are as follows(1)Aiming at the problem that registration of MRI from different phases can't be achieved precisely using intensity alone,this paper realizes registration using intensity combined with point features.Firstly,the corresponding points in MRI from different phases are extracted.Then point sets are plugged into the distance metric of point sets.Finally distance metric acts as intensity metric' penalty during the registration,which can suppress the irregular deformation caused by registration using intensity alone because of constraint of the point features to their neighborhood.(2)Aiming at the problem that the extracted feature points in liver MRI are not enough,this paper supplements improved HARRIS.The number of feature points extracted is more,because improved HARRIS can detect more feature points in images with rich texture information.And more feature points are distributed in the liver tissues and the marginal areas of the liver,which will improve performance registration based on point features.(3)This paper proposes a method of changing weight of distance metric of point sets adaptively during registration aiming at multi-resolution registration model.It reduces weight of distance metric of point sets with the increase of the image resolution in the process of registration gradually.This strategy can reduce impact of each other between intensity metric and distance metric of point sets because it pays attention to correspondence between point sets when the resolution of images are low and it pays more attention to correspondence of intensity when the resolution of images are high,which can improve the accuracy of registration.(4)This paper improves method of grey level scattergrams to detect weak tumors in MRI of the liver.The detection results obtained by the method of grey level scattergrams have a lot of noise because of the characteristics of MRI of the liver.This paper calculates grey level scattergrams using the method of Parzen Window.The improved method improves the smoothness of the grey level scattergrams,and it not only increases the contrast of the tumors area relative to the liver parenchyma greatly but also reduces the noise in the detection results.