Screening Value Of Decision Tree And Present Situation Of Early Diagnosis Of Primary Liver Cancer

Objective:To understand the current situation of early diagnosis of primary liver cancer and to explore the application value of decision tree in early screening of primary liver cancer.Methods:The data of patients with primary liver cancer diagnosed for the first time and high risk population such as chronic hepatitis and cirrhosis in our hospital from January 2014 to January 2019 were retrospectively collected.The clinical features of patients with primary liver cancer were analyzed,and the predictive values of serum AFP and/or ultrasound for primary liver cancer were compared.The specificity,sensitivity and accuracy of decision tree were calculated,and compared to Logistic regression analysis by using ROC curves.Results:1.There were 372 patients with primary liver cancer.Stages Ia,Ib,II,IIIa,IIIb and IV accounted for 5.4%,17.5%,14.8%,11.6%,34.4% and 16.4%,respectively.The incidence ofvascular invasion was 26.3%,28.2%,29.7% and 29.2% in patients with maximum diameter ? 3cm,? 5cm,? 10 cm and >10cm,respectively,and there was no significant difference between the two groups.The incidence of extrahepatic metastasis was34.2%,40.4%,44.0% and 36.1%,respectively,and there was no statistical difference.The patients' symptoms was not specific.More than 87.4% of the patients had definite HBV infection,but the proportion of patients with occult hepatitis B was high,the proportion of antiviral therapy was low,and the virological response rate was even lower.2.In the high risk population such as primary liver cancer and chronic liver disease,the proportion of patients with serum AFP and ultrasound results simultaneously was low.The sensitivity of serum AFP in predicting for primary liver cancer was 83.0%,the negative predictive value was 92.3%,and the positive predictive value was 53.3%.The predictive value of ultrasound is higher than that of serum AFP.The sensitivity and positive predictive value of serum AFP combined with upper abdominal ultrasound were 99.2% and 49.7%,if any of the two indicators was positive,the prediction could be considered as positive.But only if both indicators were positive at thesame time,the prediction should be considered as positive,the sensitivity was 71.4% and the positive predictive value was 92.0%.3.The specificity,sensitivity and accuracy of decision tree for predicting primary liver cancer in patients with cirrhosis were 95.4%,85.6%,and 92.1%,respectively.While the specificity,sensitivity and accuracy in patients with chronic hepatitis were 98.4%,86.5% and 96.5%,respectively.The areas under the ROC curve of patients with liver cirrhosis by decision tree and Logistic regression analysis were 0.97 and 0.956,respectively.While the patients with chronic hepatitis were 0.987 and 0.977,respectively.Conclusions:1.The low rate of early diagnosis of primary liver cancer is due to the easy metastasis of lesions,the increase of atypical cases,lower patient standard treatment ratio;screening program is not perfect.2.Decision tree can improve the specificity and sensitivity of primary liver cancer early screening.