多模态模型在子宫内膜癌诊疗中的应用进展

doi:10.12280/gjfckx.20260018

摘要/Abstract

摘要：

子宫内膜癌是最常见的妇科恶性肿瘤之一，具有明显的临床病理及分子异质性，传统单一模态信息难以全面反映肿瘤生物学行为及患者预后。随着人工智能技术的快速发展，多模态模型通过整合影像学、病理学、临床资料及基因组等不同类型的数据，并结合先进的算法进行联合分析，为子宫内膜癌精准诊疗提供了新的方向。对多模态模型的基本概念、开发流程、常见融合策略及不同模态数据的预处理方法进行了概述，重点总结了其在子宫内膜癌术前评估和术后预后预测中的应用进展，包括术前良恶性鉴别、早期分期、肌层浸润评估以及术后复发风险分层等。然而，该领域仍面临数据异质性大、外部验证不足、可解释性有限等问题，未来仍需开展多中心、前瞻性研究，并加强标准化建设，以推动多模态模型在子宫内膜癌精准诊疗中的规范应用。

关键词: 子宫内膜肿瘤, 癌, 多模态模型, 人工智能, 精准医学, 预后, 治疗

Abstract:

Endometrial cancer is one of the most common gynecological malignancies, characterized by significant clinicopathological and molecular heterogeneity. Traditional single-modality information struggles to comprehensively reflect tumor biology and patient prognosis. With the rapid development of artificial intelligence technology, multimodal models provide a novel direction for precision diagnosis and treatment of endometrial cancer by integrating different types of data such as imaging, pathology, clinical information, and genomics, and performing joint analysis with advanced algorithms. This article provides an overview of the basic concepts, development processes, common fusion strategies, and data preprocessing methods for different modalities in multimodal models. It focuses on summarizing their application advances in preoperative evaluation and postoperative prognosis prediction for endometrial cancer, including preoperative differentiation of benign from malignant lesions, early-stage diagnosis, myometrial invasion assessment, and postoperative recurrence risk stratification. However, this field still faces challenges, including significant data heterogeneity, insufficient external validation, and limited interpretability. Future efforts should involve multicenter, prospective studies and enhanced standardization to promote the standardized application of multimodal models in the precise diagnosis and treatment of endometrial cancer.

Key words: Endometrial neoplasms, Carcinoma, Multimodal model, Artificial intelligence, Precision medicine, Prognosis, Therapy

邓园园, 阮晓红, 张鑫, 魏继红. 多模态模型在子宫内膜癌诊疗中的应用进展[J]. 国际妇产科学杂志, 2026, 53(4): 374-378.

DENG Yuan-yuan, RUAN Xiao-hong, ZHANG Xin, WEI Ji-hong. Advances in the Application of Multimodal Models in the Diagnosis and Treatment of Endometrial Cancer[J]. Journal of International Obstetrics and Gynecology, 2026, 53(4): 374-378.

图/表 1

参考文献 36

[1]	Anca-Stanciu MB, Manu A, Olinca MV, et al. Comprehensive Review of Endometrial Cancer: New Molecular and FIGO Classification and Recent Treatment Changes[J]. J Clin Med, 2025, 14(4):1385. doi: 10.3390/jcm14041385.
[2]	Kumari A, Kumar H, Harvey SE, et al. Molecular Classification of Endometrial Carcinomas: Review and Recent Updates[J]. Cancers(Basel), 2025, 18(1):51. doi: 10.3390/cancers18010051.
[3]	Banerji C, Bhardwaj Shah A, Dabson B, et al. Clinicians must participate in the development of multimodal AI[J]. EClinicalMedicine, 2025, 84:103252. doi: 10.1016/j.eclinm.2025.103252.
[4]	Krones F, Marikkar U, Parsons G, et al. Review of multimodal machine learning approaches in healthcare[J]. Inf Fusion, 2025, 114:None. doi: 10.1016/j.inffus.2024.102690.
[5]	Lu Y, Liu F, Yu Y, et al. AI-enabled molecular phenotyping and prognostic predictions in lung cancer through multimodal clinical information integration[J]. Cell Rep Med, 2025, 6(7):102216. doi: 10.1016/j.xcrm.2025.102216.
[6]	Duran J, Zou Y, Vallières M, et al. Beyond single-run metrics with CP-fuse: A rigorous multi-cohort evaluation of clinico-pathological fusion for improved survival prediction in TCGA[J]. Mach Learn Appl, 2025, 22:100789. doi: 10.1016/j.mlwa.2025.100789.
[7]	Captier N, Lerousseau M, Orlhac F, et al. Integration of clinical, pathological, radiological, and transcriptomic data improves prediction for first-line immunotherapy outcome in metastatic non-small cell lung cancer[J]. Nat Commun, 2025, 16(1):614. doi: 10.1038/s41467-025-55847-5.
[8]	Xu X, Li J, Zhu Z, et al. A Comprehensive Review on Synergy of Multi-Modal Data and AI Technologies in Medical Diagnosis[J]. Bioengineering(Basel), 2024, 11(3):51. doi: 10.3390/bioengineering11030219.
[9]	Chen Q, Liu A, Zhang J, et al. Foundation models in medical imaging: A review[J]. Eng Medicine, 2026, 3(2):100123.
[10]	AlSaad R, Abd-Alrazaq A, Boughorbel S, et al. Multimodal Large Language Models in Health Care: Applications, Challenges, and Future Outlook[J]. J Med Internet Res, 2024, 26:e59505. doi: 10.2196/59505.
[11]	Lekadir K, Frangi AF, Porras AR, et al. FUTURE-AI: international consensus guideline for trustworthy and deployable artificial intelligence in healthcare[J]. BMJ, 2025, 388:e081554. doi: 10.1136/bmj-2024-081554.
[12]	Jiao T, Guo C, Feng X, et al. A Comprehensive Survey on Deep Learning Multi-Modal Fusion: Methods, Technologies and Applications[J]. Comput Mater Contin, 2024, 80(1):1-35. doi: 10.32604/cmc.2024.053204.
[13]	Teoh JR, Dong J, Zuo X, et al. Advancing healthcare through multimodal data fusion: a comprehensive review of techniques and applications[J]. PeerJ Comput Sci, 2024, 10:e2298. doi: 10.7717/peerj-cs.2298. pmid: 39650483
[14]	Chen C, Mat Isa NA, Liu X. A review of convolutional neural network based methods for medical image classification[J]. Comput Biol Med, 2025, 185:109507. doi: 10.1016/j.compbiomed.2024.109507.
[15]	Huang Q, Wu S, Ou Z, et al. Computational pathology: A comprehensive review of recent developments in digital and intelligent pathology[J]. Intel Oncol, 2025, 1(2):139-159. doi: 10.1016/j.intonc.2025.03.004.
[16]	Salih AM. Common Steps in Machine Learning Might Hinder The Explainability Aims in Medicine[J]. ArXiv[Preprint], 2024,arXiv:2409.00155.
[17]	Elvas LB, Almeida A, Ferreira JC. Natural language processing in medical text processing: A scoping literature review[J]. Int J Med Inform, 2025, 204:106049. doi: 10.1016/j.ijmedinf.2025.106049.
[18]	Wekesa JS, Kimwele M. A review of multi-omics data integration through deep learning approaches for disease diagnosis, prognosis, and treatment[J]. Front Genet, 2023, 14:1199087. doi: 10.3389/fgene.2023.1199087.
[19]	Tripathi A, Waqas A, Schabath MB, et al. HONeYBEE: enabling scalable multimodal AI in oncology through foundation model-driven embeddings[J]. NPJ Digit Med, 2025, 8(1):622. doi: 10.1038/s41746-025-02003-4. pmid: 41131352
[20]	Zhang J, Zhang Q, Wang T, et al. Multimodal MRI-Based Radiomics-Clinical Model for Preoperatively Differentiating Concurrent Endometrial Carcinoma From Atypical Endometrial Hyperplasia[J]. Front Oncol, 2022, 12:887546. doi: 10.3389/fonc.2022.887546.
[21]	Zheng Z, Liu Y, Feng L, et al. Multimodal MRI Image Fusion for Early Automatic Staging of Endometrial Cancer[J]. Sensors(Basel), 2025, 25(9):2932. doi: 10.3390/s25092932.
[22]	Chen Y, Ruan X, Wang X, et al. Preoperative discrimination of absence or presence of myometrial invasion in endometrial cancer with an MRI-based multimodal deep learning radiomics model[J]. Abdom Radiol(NY), 2025, 50(7):3335-3346. doi: 10.1007/s00261-024-04766-y.
[23]	Li P, Wang X, Chen Y, et al. Preoperative identification of deep myometrial invasion in endometrial cancer: a multicenter MRI study with a vision foundation model-enhanced multimodal deep learning framework[J]. Eur J Obstet Gynecol Reprod Biol, 2026, 321:115018. doi: 10.1016/j.ejogrb.2026.115018.
[24]	Li J, Liao L, Jia M, et al. Latent relation shared learning for endometrial cancer diagnosis with incomplete multi-modality medical images[J]. iScience, 2024, 27(8):110509. doi: 10.1016/j.isci.2024.110509.
[25]	Sumitha BS, Tajuddin M. Exploring The Efficacy of Multi-modal LLM In Endometrial Cancer Diagnosis[C]// 2024 International Conference on Innovation and Novelty in Engineering and Technology (INNOVA),Vijayapura, India:IEEE,2024:1-6. doi: 10.1109/innova63080.2024.10847017.
[26]	Lin C, Chen W, Lai J, et al. Integrating deep learning and clinical characteristics for early prediction of endometrial cancer using multimodal ultrasound imaging: a multicenter study[J]. Front Oncol, 2025, 15:1600242. doi: 10.3389/fonc.2025.1600242.
[27]	Lin D, Zhao L, Zhu Y, et al. Combination IETA Ultrasonographic Characteristics Simple Scoring Method With Tumor Biomarkers Effectively Improves the Differentiation Ability of Benign and Malignant Lesions in Endometrium and Uterine Cavity[J]. Front Oncol, 2021, 11:605847. doi: 10.3389/fonc.2021.605847.
[28]	Hoivik EA, Hodneland E, Dybvik JA, et al. A radiogenomics application for prognostic profiling of endometrial cancer[J]. Commun Biol, 2021, 4(1):1363. doi: 10.1038/s42003-021-02894-5. pmid: 34873276
[29]	Volinsky-Fremond S, Horeweg N, Andani S, et al. Prediction of recurrence risk in endometrial cancer with multimodal deep learning[J]. Nat Med, 2024, 30(7):1962-1973. doi: 10.1038/s41591-024-02993-w. pmid: 38789645
[30]	Li Y, Du P, Zeng H, et al. Integrative models of histopathological images and multi-omics data predict prognosis in endometrial carcinoma[J]. PeerJ, 2023, 11:e15674. doi: 10.7717/peerj.15674.
[31]	Zhang J, Zhou Y, Huang Y, et al. A Clinicopathology and Simplified Molecular Marker-Based Risk Stratification Model for Predicting 3-Year Recurrence in Endometrial Cancer[J]. Clin Med Insights Oncol, 2026, 20:11795549261428372. doi: 10.1177/11795549261428372.
[32]	Guan H, Yap PT, Bozoki A, et al. Federated learning for medical image analysis: A survey[J]. Pattern Recognit, 2024, 151:110424. doi: 10.1016/j.patcog.2024.110424.
[33]	Gaffney H, Mirza KM. Pathology in the artificial intelligence era: Guiding innovation and implementation to preserve human insight[J]. Acad Pathol, 2025, 12(1):100166. doi: 10.1016/j.acpath.2025.100166.
[34]	Abbas Q, Jeong W, Lee SW. Explainable AI in Clinical Decision Support Systems: A Meta-Analysis of Methods, Applications, and Usability Challenges[J]. Healthcare(Basel), 2025, 13(17):2154. doi: 10.3390/healthcare13172154.
[35]	Momani A. Implications of Artificial Intelligence on Health Data Privacy and Confidentiality[J]. ArXiv[Preprint], 2025,arXiv:2501. 01639.
[36]	Russ P, Mross PM, Kräling G, et al. Feasibility of a multimodal AI-based clinical assessment platform in emergency care: an exploratory pilot study[J]. Front Digit Health, 2025, 7:1657583. doi: 10.3389/fdgth.2025.1657583.

模态数据	数据来源	主要预处理步骤	常用编码模型
临床与文本数据	结构化数据（人口统计学、肿瘤分期等）、非结构化文本（病理报告、放射学报告等）	缺失值处理、数据标准化、异常值校正、文本清洗与分词、向量化表示	GatorTron、Qwen3、Med-Gemma、Llama-3.2
病理学数据	病理WSI	染色归一化、图像块切分、背景与伪影剔除、特征提取、多图像块聚合	UNI、UNI2-h、Virchow2
影像学数据	CT、MRI、B超等影像	空间标准化与重采样、分割、强度标准化、去噪	RadImageNet
基因组数据	基因表达、变异等基因组特征	质量控制、低表达或低变异过滤、缺失值处理、归一化、降维	SeNMo

模态数据	数据来源	主要预处理步骤	常用编码模型
临床与文本数据	结构化数据（人口统计学、肿瘤分期等）、非结构化文本（病理报告、放射学报告等）	缺失值处理、数据标准化、异常值校正、文本清洗与分词、向量化表示	GatorTron、Qwen3、Med-Gemma、Llama-3.2
病理学数据	病理WSI	染色归一化、图像块切分、背景与伪影剔除、特征提取、多图像块聚合	UNI、UNI2-h、Virchow2
影像学数据	CT、MRI、B超等影像	空间标准化与重采样、分割、强度标准化、去噪	RadImageNet
基因组数据	基因表达、变异等基因组特征	质量控制、低表达或低变异过滤、缺失值处理、归一化、降维	SeNMo