A Novel Segmentation Method for Multiple Sequences Magnetic Resonance Imaging Based on Multiview Fuzzy Double Weighting Probability Clustering

摘要

In current clinical aided diagnosis, image-guided surgery and radiation therapy, the technology of medical image segmentation shows increasingly important clinical value. However, due to the small differences in greyscale, the ambiguity and complexity of images, and individual variability, the performance of classic algorithms in medical image segmentation still requires improvement. Conventional medical imaging techniques include magnetic resonance imaging, computed tomography imaging, positron emission computed tomography imaging, and ultrasound imaging, where MR imaging can also generate image modalities for a variety of different time parameter sequences. Effectively exploiting the knowledge of the imaging features of one patient is also a challenge for classic algorithms. In the field of machine learning, multi-view clustering (MV clustering) algorithms have been used to handle multi-view data ( MV data), which originate from the same data samples but are obtained from a variety of perspectives. Therefore, a novel MV clustering algorithm, which utilizes the knowledge of different perspectives, corresponds to different MR sequence images, and contributes each feature in one view, is applied to segment multiple MR sequence images. The experimental results demonstrate that the MV-FDW-PCM method achieves good performance in MRI segmentation.