Supervoxel-based brain tumor segmentation with multimodal MRI images

Magnetic resonance imaging (MRI) has high accuracy, which does not only causes less damage to the patients, but also has a high spatial resolution. Besides, MRI is often used in brain tumor detection. With the gradual improvement of two-dimensional image segmentation research, people’s demand for higher-level image segmentation technology is gradually increasing. Therefore, three-dimensional (3D) image segmentation has become an important part of medical image segmentation. In this paper, a novel 3D supervoxel segmentation method is proposed for the brain tumor in multimodal MRI images. The method can directly process a 3D image, divide the image into several voxel blocks of the same size, and find the minimum distance according to the image features to generate the supervoxel segmentation boundary. The experimental results demonstrate the usefulness of the proposed method. Some comparisons also show that the performance of the proposed method is more competitive than other state-of-the-art methods.