Application of CT Radiomics Method Based on Machine Learning Classifier in Hepatic Vascular Bifurcation Detection

Hepatic vascular tumors tend to occur at the bifurcation points of hepatic vessels, where the rupture of the vessel walls’ weaker sections can lead to life-threatening bleeding. As a result, quick and accurate detection of hepatic vascular bifurcations is essential to aid physicians in diagnosing hepatic vascular tumors and creating surgical plans. However, current bifurcation detection techniques face challenges in clinical detection, low detection efficiency, compromised detection accuracy, and insufficient data security due to individual factors such as vascular deformation and data noise. This study aims to address the limitations of current bifurcation detection techniques by combining radiomics feature engineering technology and existing machine learning methods to propose a classification pipeline that is suitable for hepatic vascular bifurcation detection. Multiple feature selection algorithms are employed to obtain optimal features, which are then used to evaluate the performance of different models in bifurcation detection. Experimental results demonstrate that the K-Nearest Neighbors (KNN) classifier achieves the best performance in hepatic vascular bifurcation detection, with the test set’s average accuracy, precision, recall, F1-score, and area under the curve (AUC) being 0.964, 0.942, 0.968, 0.954, and 0.988, respectively. These findings indicate that the proposed pipeline has significant practical value in the early diagnosis of hepatic vascular diseases.