Normalized Cuts and Image Segmentation
IEEE Transactions on Pattern Analysis and Machine Intelligence
Hi-index | 0.00 |
In the clinical diagnosis of thyroid tumor, ultrasound image may provide anatomical detail of the tumor, and radionuclide image may provide functional information about activity distribution of the tumor. Fusion of the two-modality medical image doesn't only supply more abundant and comprehensive pathology information for clinic diagnosis, but also reduce the radioactive hazard from ionizing radiation because of multiple scans of x-rays of Computed Tomography. In order to realize the registration and fusion of the two modality images, we must segment the thyroid and surrounding tissues. Most of original medical images are poor contrast and intensity inhomogeneous. Hence, it is very difficult to segment using traditional segmental methods. A novel normalized cut segmentation method based on fractional derivatives is proposed and applied into thyroid tumor images in this paper. In our proposed method, the thought of fractional derivatives is introduced to implement normalized cut, which enhances thyroid tumor images by adjusting the fractional derivatives parameters, marking edge and raising the accuracy. The results of experiments show feasibility and effectiveness of proposed method.