Research On The Segmentation Of MR Medical Images Based On FCM-DS Theory

Medical imaging technology, especially magnetic resonance imaging(MRI), as a medical examination mean, has a widely application in clinicalapplication. And the medical image analysis is able to provide certainreference to help the doctor’s diagnosis on lesions, so image segmentation isone of the indispensably important link. MRI images include many kinds ofmode such as T1, T2, DWI, etc. The doctors give diagnosis and treatmentplan through a synthetic judgment analysis on variety of modal images. Thispaper is a research on segmentation on MR images based on FCM-DSalgorithm.Fuzzy c-means clustering algorithm (FCM) is one of the clusteringalgorithms which is based on the objective function, and it has beensuccessfully used in image analysis as unsupervised clustering algorithm. Indealing with medical MRI images, peripheral gray information has positiveeffects on lesions segmentation, so this paper gives an appropriateimprovement on the FCM algorithm, combined with spatial gray levelinformation, so that the clustering algorithm can be more quickly andeffectively to do image analysis, achieving good segmentation results on thelesions. Dempster-Shafer (DS) theory can do the regional division ofdifferent mode under the same image of the organization structure accordingto the uncertainty of information, and reduce the scope of the uncertain parts.This paper combines FCM algorithm and the DS theory on applicationresearch of the breast, brain MRI lesion segmentation. In the algorithmvalidation of synthesis images, the fusion of two images with12.24%and 65.28%error rate will eventually reduce to8.24%by using FCM-DSalgorithm.In the application of breast MR image lesions segmentation, choose T1,DCE and DWI sequence according to the characteristics of breast MRI, andmake fusion clustering of different modals: respectively on T1and DCE,DWI and DCE images based on FCM-DS algorithm to finish lesionssegmentation. The experimental results show that in breast imagesegmentation, image can be divided into four types (C=4) with the algorithm.Then we segment the lesion area through the results of gray expand, andmake TIC (Time Intensity Curve) assisting the doctor’s diagnosis on benignand malignant lesions.In the application of cerebral infarction lesions segmentation of brainMR images, choose T2and DWI images to fuse using FCM-DS algorithm.The brain image is divided into white matter, gray matter, infarction area andedema area totally four parts (C=4). through the reduction of thesegmentation results, bleeding area and edema area can also be divided.The results show that segmentation based on FCM-DS fusion algorithmwill be able to use gray information of the medical image. DS theorycalculates the membership degree according to the corresponding rules toreduce the uncertainty region. This algorithm can distinguish fuzzy boundarymore effectively, and may spread to other parts of the lesion division toprovide valuable diagnostic basis for the doctor.