Zhan Feng, He Lidan, Qin Shilong, Guo Yina
School of Electronic Information Engineering, Taiyuan University of Science and Technology, Taiyuan, 030024, Shanxi, China.
College of Engineering, Fujian Jiangxia University, Fuzhou, 350108, Fujian, China.
Clin Transl Oncol. 2025 Jun 19. doi: 10.1007/s12094-025-03967-1.
Ovarian cancer represents one of the most lethal gynecologic malignancies, characterized by low early detection rates and challenging prognostic assessment. Conventional diagnostic modalities demonstrate limited sensitivity and specificity for early-stage disease identification. Recent research has begun to explore causal inference methodologies as complementary approaches that may enhance diagnostic precision and prognostic capability. This systematic review evaluates the current state and future prospects of causal inference methodologies in enhancing ovarian cancer diagnosis and prognosis. We performed a comprehensive systematic review focusing on causal inference methodologies applied to ovarian cancer research. The analysis encompassed biomarker identification, pathogenic mechanism elucidation, and multimodal data integration. Additionally, we analyzed the synergistic combination of causal inference with machine learning approaches across genomic, transcriptomic, proteomic, and imaging datasets. Causal inference methods have shown effectiveness in identifying crucial biomarkers and revealing underlying pathogenic mechanisms of ovarian cancer. The integration of machine learning with causal inference has enhanced model interpretability, clinical applicability, and diagnostic-prognostic accuracy. These approaches have achieved improved predictions of disease progression and optimization of treatment strategies by leveraging clinical, genetic, and imaging data. Causal inference shows considerable potential in advancing precision medicine for ovarian cancer, offering robust frameworks for addressing confounding factors and establishing causal relationships. As these methodologies evolve and data volumes expand, their application may become increasingly valuable in oncology practice.
卵巢癌是最致命的妇科恶性肿瘤之一，其特点是早期检测率低且预后评估具有挑战性。传统的诊断方法对早期疾病识别的敏感性和特异性有限。最近的研究已开始探索因果推断方法作为补充手段，以提高诊断准确性和预后能力。本系统评价评估了因果推断方法在增强卵巢癌诊断和预后方面的现状和未来前景。我们进行了一项全面的系统评价，重点关注应用于卵巢癌研究的因果推断方法。分析涵盖了生物标志物识别、致病机制阐明和多模态数据整合。此外，我们分析了因果推断与机器学习方法在基因组、转录组、蛋白质组和成像数据集方面的协同结合。因果推断方法在识别关键生物标志物和揭示卵巢癌潜在致病机制方面已显示出有效性。机器学习与因果推断的整合提高了模型的可解释性、临床适用性以及诊断和预后准确性。这些方法通过利用临床、遗传和成像数据，在疾病进展预测和治疗策略优化方面取得了进展。因果推断在推进卵巢癌精准医学方面显示出巨大潜力，为解决混杂因素和建立因果关系提供了有力框架。随着这些方法的不断发展和数据量的增加，它们在肿瘤学实践中的应用可能会变得越来越有价值。
