颞叶癫痫病患者的脑皮层多特征分类

发布时间：2019-01-01 15:02

【摘要】：为了自动识别左、右侧颞叶癫痫病患者,提出了一种基于脑皮层结构特征的分类方法.采集了21例左侧TLE患者(LTLE)、18例右侧TLE患者(RTLE)与28例健康正常人(NC)的脑结构磁共振图像,提取与分析其脑皮层形态结构特征.为了降低特征维数,利用递归特征消除法获取对分类有效的主要特征.然后,采用支持向量机对3组人群进行分类.结果表明,大脑皮层表面积为最佳分类特征,灰质体积也具有突出的分类能力,皮层厚度识别不同侧TLE的能力稍差,平均曲率对右侧TLE与健康人群的区分能力相对较弱.对分类有重要影响的脑区主要分布于颞叶、额叶与顶叶,尤其是大脑左侧脑区.由此证明利用脑皮层特征可以有效地区分不同侧TLE患者和健康正常人,TLE患者的左脑更易受损.研究结果有助于理解TLE患者的病理机制与进展规律、确定病灶位置及实现自动诊断.
[Abstract]:In order to automatically recognize the left and right temporal lobe epilepsy patients, a classification method based on cortical structure features was proposed. Brain structure MRI images of 21 patients with left TLE (LTLE), 18 patients with right TLE and 28 healthy people with (NC) were collected and the morphological features of cerebral cortex were extracted and analyzed. In order to reduce the feature dimension, the recursive feature elimination method is used to obtain the main features that are effective for classification. Then, support vector machine was used to classify three groups of people. The results showed that the surface area of cerebral cortex was the best classification feature, and the volume of gray matter also had prominent classification ability. The ability of cortex thickness to recognize TLE in different sides was slightly poor, and the ability of mean curvature to distinguish the right side of TLE from the healthy population was relatively weak. The brain regions which have important influence on classification are mainly distributed in the temporal lobe, frontal lobe and parietal lobe, especially in the left brain area of the brain. It is proved that the cortical features can be used to distinguish the different sides of TLE patients from healthy people, and the left brain of TLE patients is more easily damaged. The results are helpful to understand the pathological mechanism and progression of TLE patients, to determine the location of lesions and to realize automatic diagnosis.
【作者单位】：华南理工大学材料科学与工程学院;广东三九脑科医院影像诊断中心;
【基金】：国家自然科学基金资助项目(31371008;81171179) 广东省科技计划资助项目(503253185076)
【分类号】：R742.1;R445.2;TP391.41

【相似文献】