Urine fluorescence spectroscopy combined with machine learning for screening of hepatocellular carcinoma and liver cirrhosis.

In this paper, we investigated the possibility of using urine fluorescence spectroscopy and machine learning method to identify hepatocellular carcinoma (HCC) and liver cirrhosis from healthy people. Urine fluorescence spectra of HCC (n = 62), liver cirrhosis (n = 65) and normal people (n = 60) were recorded at 405 nm excitation using a Fluorescent scan multimode reader. The normalized fluorescence spectra revealed endogenous metabolites differences associated with the disease, mainly the abnormal metabolism of porphyrin derivatives and bilirubin in the urine of patients with HCC and liver cirrhosis compared to normal people. The Support vector machine (SVM) algorithm was used to differentiate the urine fluorescence spectra of the HCC, liver cirrhosis and normal groups, and its overall diagnostic accuracy was 83.42%, the sensitivity for HCC and liver cirrhosis were 93.55% and 73.85%, and the specificity for HCC and liver cirrhosis were 88.00% and 89.34%, respectively. This exploratory work shown that the combination of urine fluorescence spectroscopy and SVM algorithm has great potential for the noninvasive screening of HCC and liver cirrhosis.