Unsupervised Three-Dimensional Tubular Structure Segmentation via Filter Combination

Abstract Tubular structure enhancement plays an utmost role in medical image segmentation as a pre-processing technique. In this work, unsupervised 3D tubular technique is developed, which mainly inspired by the idea of filter combination. Three well-known vessel filters, Frangi’s filter, modified and Multiscale Fractional Anisotropic Tensor (MFAT) separately enhance original images. Next, enhanced images obtained using three different filters are combined. Different categories have ability complementarity, main motivation combining these advanced filters. The combination them ensures high diversity enhancing results. Weighted mean median ranking methods used to conduct operation Based on optimized weights for all individual fuzzy C-means method then applied segment structures. proposed tested public DRIVE STARE datasets, synthetic vascular models (2011 2013 VascuSynth Sample), real-patient Coronary Computed Tomography Angiography (CCTA) datasets. Experimental results demonstrate that outperforms state-of-the-art combination-based methods. Moreover, our able yield better than each exhibits superiority method. conclusion, can be further facilitate practice.