Showing papers by "Jun Xia published in 2018"

Glioma Segmentation With a Unified Algorithm in Multimodal MRI Images

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.

Abnormalities of diffusional kurtosis imaging and regional homogeneity in idiopathic generalized epilepsy with generalized tonic-clonic seizures.

Abstract: Neuroimaging techniques have been used to investigate idiopathic generalized epilepsy with generalized tonic-clonic seizures (IGE-GTCS) and different studies employing these methods have produced varying results. However, there have been few studies exploring diffusional kurtosis imaging (DKI) and regional homogeneity (ReHo) techniques in patients with IGE-GTCS. In the current study, resting-state functional magnetic resonance imaging (fMRI) and DKI data were collected from 28 patients with IGE-GTCS and 28 healthy controls. The ReHo method and tract-based spatial statistical (TBSS) analysis were performed to compare differences between the groups. Compared with healthy controls, patients with IGE-GTCS exhibited markedly increased ReHo in the bilateral putamen, the thalamus, right pallidum, right supplementary motor area and the bilateral paracentral lobules. Compared with healthy controls, patients with IGE-GTCS also exhibited markedly decreased ReHo in the posterior cingulate/precuneus, left angular gyrus and dorsolateral prefrontal cortex. In patients with IGE-GTCS, DKI revealed lower fractional anisotropy in the left anterior/superior corona radiata, left superior longitudinal fasciculus and genu/body of the corpus callosum. Higher mean diffusivity was detected in the bilateral anterior corona radiata, left superior corona radiata, left cingulum, and genu/body/splenium of the corpus callosum. Furthermore, reduced mean kurtosis values were identified over the bilateral superior/posterior corona radiate, left anterior corona radiata, right superior longitudinal fasciculus, left posterior thalamic radiation and the genu/body/splenium of the corpus callosum. Therefore, the results of the current study revealed abnormalities in spontaneous activity in the gray and white matter tracts in patients with IGE-GTCS. These results suggest that novel MRI technology may be useful to help determine the pathogenesis of IGE-GTCS.