Preoperative Prediction The Grade Of Hepatocellular Carcinoma Based On Contrast-enhanced CT Radiomics Signature

Background: Hepatocellular carcinoma is one of the most common malignancies worldwide,and China accounts for more than half of the new cases each year.The grade of hepatocellular carcinoma is considered to be a significant predictor of patient prognosis.Compared with low-grade hepatocellular carcinoma,high-grade hepatocellular carcinoma has a higher rate of metastasis and recurrence and poor prognosis after liver resection and transplantation.Therefore,assessing the degree of hepatocellular carcinoma before surgery is helpful for the selection of clinical treatment strategies and the assessment of prognosis.The degree of hepatocellular carcinoma is usually assessed by pathology after surgery,and it is difficult to evaluate it before surgery.Radiomics can be used as a new method to predict the degree of hepatocellular carcinoma preoperatively because it can convert medical imaging into high-dimensional features that can be mined non-invasively before surgery.Objective: To explore the preoperative prediction the grade of hepatocellular carcinoma based on contrast-enhanced CT radiomics signatureMethods: The ethics committee approved this study.A retrospective analysis of 161 patients with hepatocellular carcinoma who underwent enhanced CT before surgery in our hospital.All the included patients ultimately had the portal venous phase CT images and clinical and pathological data.They were divided into 112 cases in the training datasets and 49 cases in the testing datasets.The portal venous phase CT image was used to extract the radiomics features,including Histogram,Formfactor,Gray-level Co-occurrence Matrix(GLCM),Run-length matrix(RLM),and Grey-level size zone matrix(GLSZM) features.Feature dimension reduction based on LASSO logistic regression method,select radiomics features related to the degree of hepatocellular carcinoma,and construct imaging radiomics signature.The optimal model parameters were selected,and the receiver operating characteristic curve was drawn to evaluate the diagnostic efficacy of radiomics signature to predict the degree of hepatocellular carcinoma before the operation,and verified in the testing datasets.Results: After image feature extraction and dimensionality reduction,the seven radiomics features are used to construct the radiomics signature.There were significant differences(P <0.01)on the portal venous phase CT based radiomics signature between low and high-grade hepatocellular carcinoma.The portal venous phase CT based radiomics signature has shown good efficacy in assessing the grade of hepatocellular carcinoma.The radiomics signature has an AUC of 0.890 in the training dataset and 0.938 in the testing dataset.Conclusions: Based on contrast-enhanced CT radiomics signature can be used to assess the grade of hepatocellular carcinoma differentiation before surgery,as a non-invasive auxiliary tool for the choice of treatment strategies for patients with hepatocellular carcinoma.