多模态影像组学在预测直肠癌免疫治疗相关蛋白表达中的研究进展

Research Progress of Multimodal Radiomics in Predicting Protein Expression Associated with Rectal Cancer Immunotherapy

摘要: 近年来，多模态影像组学在直肠癌精准诊疗中展现出重要潜力，尤其在预测免疫治疗相关蛋白(如PD-L1、KI-67等)表达方面取得了显著进展。本文综述了基于CT、MRI及PET等多模态影像的组学特征提取与分析方法，探讨了其与免疫检查点蛋白表达的关联性。研究表明，影像组学可通过量化肿瘤异质性、功能代谢及灌注特征，无创预测关键生物标志物状态，为免疫治疗患者筛选提供新思路。然而，当前研究仍面临多中心数据标准化不足、模型泛化性有限等挑战。未来需结合深度学习与多组学数据，以进一步优化预测效能，推动个体化免疫治疗决策体系的建立。

Abstract: In recent years, multimodal radiomics has shown important potential in the precise diagnosis and treatment of rectal cancer, especially in predicting the expression of immunotherapy-related proteins (such as PD-L1, KI-67, etc.). In this paper, the omics feature extraction and analysis methods based on multimodal images such as CT, MRI and PET were reviewed, and their association with immune checkpoint protein expression was discussed. Studies have shown that radiomics can non-invasively predict the status of key biomarkers by quantifying tumor heterogeneity, functional metabolism and perfusion characteristics, providing new ideas for immunotherapy patient screening. However, current research still faces challenges such as insufficient standardization of multicenter data and limited generalization of models. In the future, it is necessary to combine deep learning and multi-omics data to further optimize the prediction efficiency and promote the establishment of an individualized immunotherapy decision-making system.