尿液是传统医学观察疾病的重要窗口，然而其客观化与量化是难题。慢性心衰患者63%报告有2型心肾综合征，早期发现却是瓶颈问题。我们前期研究发现，心脏的代谢状态可某种程度上反映在尿液代谢物中，鉴定出18种差异代谢物可作为高度准确的预测因子。在采集样品的过程中还发现，慢性心衰患者尿液宏观性状有异常。由此提出假说：慢性心衰患者尿液的宏观性状、以及包括上述18种代谢物的微观指标可以区分慢性心衰与2型心肾综合征，并且与慢性心衰的寒热证相关。拟采用图像识别联合人工智能算法（包括卷积神经网络、支持向量机和随机森林等多种算法），以尿液宏观性状量化客观化为切入点，通过对慢性心衰及2型心肾综合征患者进行尿液代谢组学的分析，找出二者在宏观尿液和微观分子之间的区别，建立其与寒热证候的关系，从而在宏观性状与微观生物学指标间架起沟通的桥梁。对于尿诊的客观化与量化，指导中医寒热辨证，提高诊断的准确性和临床疗效都有重大意义。

Urine is an important window for traditional medicine to observe diseases, but its objectification and quantification has always been a difficult problem in academia. Type 2 cardiorenal syndrome is reported in 63% of patients with chronic heart failure, but early detection is a bottleneck problem. Our previous studies have found that the metabolic state of the heart can be reflected in urine metabolites to some extent, and 18 differential metabolites have been identified as highly accurate predictors. In the process of collecting samples, it was also found that the macroscopic characteristics of urine in patients with chronic heart failure were abnormal. It is hypothesized that the macroscopic characteristics of urine and the microscopic indexes including the 18 metabolites mentioned above can distinguish chronic heart failure from type 2 cardiorenal syndrome, and are related to the cold-heat syndrome of chronic heart failure. This paper intends to use image recognition combined with artificial intelligence (including support vector machine and random forest) to quantify and objectify the macroscopic characteristics of urine. Through the analysis of urine metabolomics in patients with chronic heart failure and type 2 cardiorenal syndrome, the difference between macroscopic urine and microscopic molecules is found, and the relationship between them and cold-heat syndrome is established. The bridge between the visual traits and the microbiological indicators is established. It is of great significance to the objectification and quantification of urine diagnosis, to guide the differentiation of cold and heat in TCM, and to improve the accuracy and clinical efficacy of diagnosis.