基于多模态MR神经影像的轻微型肝性脑病预测模型的研究

Minimal hepatic encephalopathy (MHE), which is characterized by subtle cognitive and psychomotor deficits, is the mildest form of hepatic encephalopathy (HE). Without immediate intervention, patients with MHE could convert to clinical HE. As a result, very early diagnosis and intervention are extremely important for MHE. So, an objective evaluation criterion can be complementary, and can avoid the subjectivity of normally used behavior diagnostic criteria. Our previous studies and preliminary studies have found that neuroimage evidences and image related predictive models can provide potential objective information to MHE behavior diagnostic criteria. Current neuroimage evidences have been shown that MHE patients are significantly different from controls in brain structure and function. But most studies applied statistical methods to analyze the numerical differences of brain structural parameters (e.g., volume, fractional anisotropy, and etc.) and functional parameters (e.g., function connectivity, regional homogeneity, and etc.). However, the spatial distribution of abnormality and how this spatial distribution is related to clinical variables were not addressed in these studies. Therefore, we plan to recruit 50 MHE subjects and 50 normal controls, and obtain their multimodal MR images, such as 3D structural, diffusion tensor and functional images. Then, we will use a Bayesian network based algorithm to detect a set of brain regions which characterize MHE, and model the associations among those regions and clinical neurological deficits. Finally, an objective MHE predictive model will be generated; and the relationship between each effective biomarker will be computed by combining clinical variables and features from multimodal MR images. Our MHE studies can be used in the clinical setting, and can effectively provide objective information to early diagnosis and intervention for MHE.

轻微型肝性脑病（MHE）是肝性脑病中最轻的一种形式，以轻度神经、运动障碍为特点，易发展成临床型肝性脑病。因此早期诊断干预就显得尤为重要，所以需辅以客观评估标准以避免常用的行为学诊断中存在的主观因素。我们前期研究和预实验初步发现：影像学证据及其预测模型可为MHE提供潜在的客观信息。目前大量影像学证据虽表明MHE患者的脑结构和功能存在显著变化，但大部分研究只通过统计学方法，研究脑结构参数(如体积等)和功能参数（如功能连接等）在数值上的差异，却忽视这些异常的空间分布及其与临床表征之间的联系。因此计划入组MHE患者50例，对照组50例，采集多模态MR图像，通过贝叶斯网络分析多模数据，获取MHE脑结构和功能差异的空间分布规律，及其与神经损伤间的联系；联合临床参数和多模态MR图像，共同为MHE建立客观的预测诊断模型，计算各个有效生物学标记之间的关系。我们的研究可为临床的早诊断早治疗提供有效客观依据。

轻微型肝性脑病（MHE）是肝性脑病中最轻的一种形式，以轻度神经、运动障碍为特点，易发展成临床型肝性脑病。因此早期诊断干预就显得尤为重要，所以需辅以客观评估标准以避免常用的行为学诊断中存在的主观因素。因此我们收集了无MHE的肝硬化患者20例，有严重HE病史的肝硬化患者20例，MHE 患者20 例，对照组20 例，采集多模态MR 图像，使用神经学量表，采集血样计算了child-pugh分级。通过数据分析，我们发现MHE患者存在比较显著的功能连接（局部一致性，低频振幅，以及默认网络）异常，而且存在一定程度的白质损伤（白质体积以及各向异性分数），并且这些功能和结构的异常与神经学量表之间有显著的相关性。在此基础上，通过计算各个脑区之间功能连接，通过基于统计学的主成分分析（kPCA）的降维，最大程度的保留原始信息，并以此为参数进行MHE预测模型的计算，得到了比较精确的预测模型；另外，我们还使用贝叶斯网络分析不同的脑网络（默认网络、腹侧注意网络和背侧注意网络），获取MHE脑网络功能差异的空间分布规律，为MHE 建立比较精确的早期诊断预测诊断模型。因此我们的研究证明了影像学证据MHE 提供潜在的客观信息，并且MHE预测模型可为临床的早诊断早治疗提供有效客观依据。