基于超声造影的瘤内及瘤周影像组学鉴别乳腺原位癌与浸润性癌
回顾性纳入322例患者329个病灶的术前超声造影图像,提取瘤内及瘤周2mm、5mm、8mm影像组学特征构建逻辑回归模型。整合5mm瘤周特征、瘤内特征及临床数据的模型诊断效能最优,训练集和测试集AUC分别为0.927和0.911,优于仅用临床特征或其他影像组学配置的模型。
要点
- 联合瘤内与5mm瘤周超声造影影像组学及临床特征可鉴别乳腺原位癌与浸润性癌,测试集AUC为0.911。
- 5mm瘤周区域在鉴别病灶中表现最优,优于2mm和8mm瘤周配置。
- 模型诊断效能优于仅基于临床特征或其他影像组学配置的模型。
核心结论
- 基于超声造影的瘤内与瘤周影像组学联合模型对鉴别乳腺原位癌与浸润性癌具有显著潜力,整合5mm瘤周特征可明显提升诊断准确性。
研究设计与样本
- 回顾性连续收集经术后组织病理学证实的病例,包括2018年1月至2024年5月的143例原位癌和2022年5月至2024年5月的186例浸润性癌,共322例患者329个病灶,均有完整术前超声造影图像;在造影峰值相图像(灰度模式)上勾画瘤内ROI,并将瘤周ROI定义为超出肿瘤边缘2mm、5mm和8mm以提取影像组学特征;采用统计与机器学习方法筛选特征,使用逻辑回归分类器构建整合瘤内、瘤周及临床特征的影像组学模型,以AUC评估模型性能。
诊断标准与关键数值
- 整合5mm瘤周特征、瘤内特征及临床数据的模型诊断效能最优,训练集AUC为0.927,测试集AUC为0.911。
临床应用与阅片要点
- 原文未据此给出推荐。
局限性
- 原文未报告局限性。
以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。
英文原文(摘要)· 点击展开
RATIONALE AND OBJECTIVES: This study aimed to evaluate the efficacy of a diagnostic model integrating intratumoral and peritumoral radiomic features based on contrast-enhanced ultrasound (CEUS) for differentiation between carcinoma in situ (CIS) and invasive breast carcinoma (IBC).
METHODS: Consecutive cases confirmed by postoperative histopathological analysis were retrospectively gathered, comprising 143 cases of CIS from January 2018 to May 2024, and 186 cases of IBC from May 2022 to May 2024, totaling 322 patients with 329 lesion and complete preoperative CEUS imaging. Intratumoral regions of interest (ROI) were defined in CEUS peak-phase images deferring gray-scale mode, while peritumoral ROI were defined by expanding 2 mm, 5 mm, and 8 mm beyond the tumor margin for radiomic features extraction. Statistical and machine learning techniques were employed for feature selection. Logistic regression classifier was utilized to construct radiomic models integrating intratumoral, peritumoral, and clinical features. Model performance was assessed using the area under the curve (AUC).
RESULTS: The model incorporating 5 mm peritumoral features with intratumoral and clinical data exhibited superior diagnostic performance, achieving AUCs of 0.927 and 0.911 in the training and test sets, respectively. It outperformed models based only on clinical features or other radiomic configurations, with the 5 mm peritumoral region proving most effective for lesions discrimination.
CONCLUSION: This study highlights the significant potential of combined intratumoral and peritumoral CEUS radiomics for classifying CIS and IBC, with the integration of 5 mm peritumoral features notably enhancing diagnostic accuracy.
引用
- Zheng Y, Song Y, Wu T, et al. Contrast-Enhanced Ultrasound-Based Intratumoral and Peritumoral Radiomics for Discriminating Carcinoma In Situ and Invasive Carcinoma of the Breast. Acad Radiol. 2025. PMID: 40753025. doi:10.1016/j.acra.2025.07.021
供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。