Glioma Segmentation Using a Novel Unified Algorithm in Multimodal MRI Images
Qingneng Li,Zhifan Gao,Qiuyu Wang,Jun Xia,Heye Zhang,Huailing Zhang,Huafeng Liu,Shuo Li
DOI: https://doi.org/10.1109/access.2018.2807698
IF: 3.9
2018-01-01
IEEE Access
Abstract:To achieve the better segmentation performance, we propose a unified algorithm for automatic glioma segmentation. In this paper, we first use spatial fuzzy c-mean clustering to estimate region-of-interest in multimodal MRI images, and then extract some seed points from there for region growing based on a new notion "affinity". In the end, we design a two-step strategy to refine the glioma border with region merging and improved distance regularization level set method. In BRATS 2015 database, we evaluate the accuracy and robustness of our method with performance scores, including dice, positive predictive value (PPV), and sensitivity metrics, as well as Hausdorff and Euclidean distance (HD&ED). The high metric values (dice = 0.86, PPV = 0.90, and sensitivity = 0.84) and small distance errors (HD = 14.39 mm and ED = 3.31 mm) indicate a remarkable accuracy. Also, we observe the ranking is No.1 in terms of dice and PPV, comparing with the state-of-the-art methods. In addition, the robustness is also at a high-level due to the refinement structure. And Spearman's rank coefficient test verities a significant correlation between the high-grade gliomas and low-grade gliomas. Overall, the proposed method is effective in segmenting gliomas in multimodal images or flair images, and has the potential in routine examinations of gliomas in daily clinical practice.