Objective To ascertain utility of the model that combines serum markers and CT features in assessing the differentiation degree of hepatocellular carcinoma (HCC). Methods A total of 206 cases of HCC clinical and CT data were collected retrospectively and the patients were divided into training set (including 42 cases of low-differentiated HCC and 102 cases of middle-high differentiated HCC) and testing set (including 21 cases of low-differentiated HCC and 41 cases of middle-high differentiated HCC). The underlying differences between low-differentiated HCC group and middle-high differentiated HCC group in terms of clinical and CT features were meticulously compared. Applying multivariate Logistic regression, we isolated independent risk factors for HCC differentiation degree and construct the prediction models. Results Compared with medium-high differentiated HCC, low-differentiated HCC had statistically significant higher rate of AFP positivity (P=0.001), occurrences of hepatitis B (P=0.003), low-density ring sign (P=0.015), cancer thromboembolism (P=0.001), and lower CT values during plain scan (P=0.010). Further multivariate logistic regression analysis revealed that AFP (OR=0.269, P=0.027), low-density ring sign (OR=0.273, P=0.047), cancer thromboembolism (OR=0.191, P=0.005), and plain scan CT value of tumor (OR=1.091, P=0.009) act as risk factors for HCC differentiation degree. The optimal diagnostic performance was achieved by the model that integrated AFP, low-density ring sign, cancer thromboembolism, and CT value of tumor during plain scan, as demonstrated by the area under the curve of 0.780 and 0.620 in the training and testing set, respectively. Conclusion AFP, low-density ring sign, cancer thromboembolism, and CT value of tumor during plain scan are independent risk factors for the differentiation degree of HCC tissue. when amalgamated into the model, the joint model constructed based on these features can provide a high-accuracy diagnosis for HCC differentiation degree.
国家科技期刊平台
NSTL
万方数据
