Machine-learning-based radiomics model for AKR1B10 prediction and prognosis in hepatocellular carcinoma

Abstract Background AKR1B10 plays a crucial role in the early diagnosis and prognosis of hepatocellular carcinoma (HCC). Our study aimed to develop a radiomics model that can effectively differentiate the expression of AKR1B10 in HCC patients, while also investigating its prognostic value. Methods The Cancer Genome Atlas (TCGA) database was used to investigate the differential expression of AKR1B10 and its prognostic value in HCC. We extracted computed tomography (CT) images from The Cancer Imaging Archive (TCIA) database and applied machine learning algorithms to extract radiomics features. The radiomics features were utilized to construct a logistic regression model for predicting AKR1B10 expression. The model underwent validation through cross-validation, and its performance was assessed using DCA and ROC curve analysis. Furthermore, we developed a nomogram utilizing both radiomics score (Rad_score) and clinical features to predict the prognosis of HCC patients. Results Higher AKR1B10 expression was identified as an independent risk factor for poor prognosis in patients with HCC. The radiomics features of HCC patients with high and low AKR1B10 expression were distinct. A radiomics-based prediction model for AKR1B10 expression was established, yielding an area under the ROC curve of 0.83 (95% CI, 0.69–0.97). Using Rad_score and clinical-pathological features, a nomogram was developed to predict 3-year survival in HCC patients. Conclusions AKR1B10 was an independent prognostic indicator in patients with HCC. Furthermore, a radiomics model based on CT could predict the AKR1B10 expression and prognosis in HCC patients.