Cai Chengfeng, Hu Wenhui, Zhou Haimei, Zhang Xian, Ren Rongfei, Liu Yilin, Ye Facui
Department of Ultrasound, Liangshan Yi Autonomous Prefecture Hospital of Integrated Traditional Chinese and Western Medicine, Xichang, China.
Department of Reproductive, Chengdu Jinjiang District Maternal and Child Health Hospital, Chengdu, China.
Front Oncol. 2025 Jun 24;15:1621642. doi: 10.3389/fonc.2025.1621642. eCollection 2025.
Uterine fibroids (leiomyomas) are the most common benign uterine tumours, affecting a significant portion of women, and often present with symptoms similar to malignant tumours, such as leiomyosarcoma or endometrial carcinoma, particularly in patients with cancer-related pelvic pain. Conventional imaging modalities, including ultrasound, CT, and MRI, struggle to differentiate between these benign and malignant conditions, often leading to misdiagnoses with potentially severe consequences, such as unnecessary hysterectomies or inadequate treatment for malignancy. Recent advances in artificial intelligence (AI) have begun to address these challenges by enhancing diagnostic accuracy and workflow efficiency. AI-assisted imaging, encompassing techniques like radiomics, convolutional neural networks (CNNs), and multimodal fusion, has demonstrated substantial improvements in distinguishing between uterine fibroids and malignant smooth-muscle tumours. Furthermore, AI has streamlined clinical workflows, enabling faster, more accurate segmentation, and automating decision-making processes, which significantly benefits patients presenting with acute cancer-related pain. Throughout this article the term radiation imaging is used as an umbrella for ionising-based modalities (CT, PET/CT) and non-ionising, radiation-planned modalities such as MRI and diagnostic ultrasound that feed the same radiotherapy or interventional planning pipelines; with that definition clarified, the review synthesizes current developments in AI-assisted radiation imaging for differentiating uterine fibroids from malignant lesions, exploring diagnostic gaps, emerging AI frameworks, and their integration into clinical workflows. By addressing the technical, regulatory, and operational aspects of AI deployment in pelvic-pain management, this review aims to provide a comprehensive roadmap for incorporating AI into personalized, efficient, and equitable oncologic care for women.
子宫肌瘤（平滑肌瘤）是最常见的子宫良性肿瘤，影响着相当一部分女性，并且常常表现出与恶性肿瘤相似的症状，如平滑肌肉瘤或子宫内膜癌，尤其是在患有癌症相关盆腔疼痛的患者中。包括超声、CT和MRI在内的传统成像方式难以区分这些良性和恶性情况，常常导致误诊，可能产生严重后果，如不必要的子宫切除术或对恶性肿瘤的治疗不足。人工智能（AI）的最新进展已开始通过提高诊断准确性和工作流程效率来应对这些挑战。AI辅助成像，包括放射组学、卷积神经网络（CNN）和多模态融合等技术，在区分子宫肌瘤和恶性平滑肌肿瘤方面已显示出显著改善。此外，AI简化了临床工作流程，实现了更快、更准确的分割，并使决策过程自动化，这对患有急性癌症相关疼痛的患者有显著益处。在本文中，术语“放射成像”用作基于电离的成像方式（CT、PET/CT）以及为相同放疗或介入规划流程提供数据的非电离、放射规划成像方式（如MRI和诊断超声）的统称；明确该定义后，本综述综合了AI辅助放射成像在区分子宫肌瘤与恶性病变方面的当前进展，探讨诊断差距、新兴的AI框架及其融入临床工作流程的情况。通过解决AI在盆腔疼痛管理中部署的技术、监管和操作方面的问题，本综述旨在提供一个全面的路线图，以将AI纳入针对女性的个性化、高效且公平的肿瘤护理中。
