多模态影像对非酒精性脂肪肝肝脏脂肪含量精准定量的实验研究

Nonalcoholic fatty liver disease has emerged as one of main causes of chronic liver disease, which could progress to non-alcoholic steatohepatitis, cirrhosis and hepatocellular carcinoma; what is more, NAFLD is association with many metabolic diseases. Currently, there is no specific serum biomarkers for hepatic fat accurate quantification. Liver biopsy is considered as the gold standard for evaluation of NAFLD, however, liver biopsy could not been available widely because of its invasive characteristic. Therefore, it is very important for clinical application to evaluate accurate quantification of liver fat content noninvasively. The purpose of this study is to evaluate the dynamic changes of liver fat content in the rabbit model of NAFLD by multimodality imaging of ultrasound real-time shear wave elastography, conventional computed tomography (CT), quantitative CT, dual energy CT, magnetic resonance spectroscopy and the newest mDixon Quant technique. The diagnostic efficiency of each modality for assessment of liver fat content is analyzed and compared by Receiver operating characteristic (ROC) curve， and advantage of each modality is also integrated by means of mathematical models for accurate quantification of liver fat content. The significant of this project is to evaluate the liver fat content of NAFLD by multimodality imaging for accurate quantification, which could offer the appropriate modality for intervention, surveillance and curative effect evaluation of NAFLD.

NAFLD已经成为肝脏慢性病变主要原因之一，其可发展成为非酒精性脂肪肝炎、肝纤维化甚至发展成肝癌，并与多种代谢性病变有密切关系。目前，没有特异的生化检查能精准定量NAFLD的肝脏脂肪含量，其诊断的“金标准”仍然是肝脏活检，但由于肝脏病理活检具有创伤性，不适合广泛应用，因此，无创地评估肝脏脂肪含量就显得很重要。本研究拟通过建立家兔NAFLD的动物模型，探索NAFLD发生发展过程中肝脏脂肪含量的动态变化，利用超声实时弹性成像、常规CT、定量CT、双能量CT、磁共振波谱及最新的mDixon Quant技术对家兔NAFLD模型进行多模态影像评估，并与病理结果作比较，分析及评估多模态影像的诊断效能，并建立数学模型整合多模态影像的优势来精准定量肝脏脂肪含量。本项目的意义在于通过多模态影像精准定量NAFLD的肝脏脂肪含量，为指导临床大夫及时对NAFLD患者进行干预、随访及疗效评估提供合适的定量手段。

非酒精性脂肪肝病是最常见的慢性肝病原因，影响着全球四分之一人口，目前尚缺乏特征性的生化检查和影像手段精准地诊断NAFLD的严重程度。本项目通过建立脂肪肝家兔动物模型，基于超声、CT 和MR 新技术，以病理结果作为金标准，同时基于深度学习对肝脏病理图片进行病理切片的定量分析，最后通过建立数学模型进行拟合多种影像学技术，从而实现精准定量非酒精性脂肪肝病肝脏的脂肪含量。研究成果如下：（1）通过高脂高胆固醇饮食，可以建立不同严重程度的家兔NAFLD动物模型，且随着造模时间的延长肝实质出现不同的病理特点。（2）基于深度学习肝脏病理图片进行病理切片的定量分析是可行，具有较高的准确性及稳定性。（3）双能量CT虚拟单能量图像对NAFLD的定性及定量诊断具有一定的价值。对于正常肝实质，随着keV能级的升高，肝实质的CT值逐渐降低；而对于非酒精性脂肪肝病的肝实质，随着keV能级的升高，肝实质的CT值逐渐升高。在NAFLD肝实质信噪比方面，我们推荐使用60keV。（4）单模态影像鉴别正常肝与轻度脂肪变性以上、鉴别轻度脂肪变性以下与中度脂肪变性以上、鉴别中度脂肪变性以下与重度脂肪变性的ROC曲线下面积范围为0.551-0.941；而联合超声、CT及MRI多模态影像诊断，ROC曲线下面积范围为0.937-1.000。单模态影像评价NAFLD脂肪含量的严重程度均具有一定的准确性，联合合适的多模态影像可以提高诊断效能，这为临床及时对NAFLD患者进行诊断、干预、及疗效评估提供合适的定量手段。本项目执行期间已发表论文8篇，其中7篇被SCI检索，指导3名研究生毕业论文。