文献速递胸部乳腺技术与安全CT

CT中用于个性化乳腺辐射剂量学的机器学习方法:整合影像组学与深度神经网络

A machine learning approach for personalized breast radiation dosimetry in CT: Integrating radiomics and deep neural networks.
PubMed 文献 · Eur J Radiol · 2025-06-11

本研究回顾性收集286例胸部CT及左右乳腺轮廓,开发并验证乳腺分割U-Net,并用蒙特卡洛模拟计算乳腺剂量,再以X射线管电流和影像组学特征训练深度神经网络预测乳腺剂量;分割平均Dice系数0.92,左右乳腺预测剂量与真实值接近,平均绝对百分比误差右乳4.01%、左乳4.82%,预测与真实剂量无显著差异(p≥0.07)。

要点

  • 乳腺分割U-Net平均Dice系数0.92,左右乳腺精确率与敏感度均高于0.90。
  • 深度神经网络预测剂量与蒙特卡洛真实值接近,右乳平均预测5.05±0.50 mGy vs 真实5.03±0.57 mGy。
  • 左乳平均预测5.06±0.55 mGy vs 真实5.02±0.61 mGy,平均绝对百分比误差4.82%(范围4.56%-5.11%)。

核心结论

  • 本研究提出一种自动化、基于机器学习的CT乳腺辐射剂量学工作流,整合分割与剂量预测模型。

研究设计与样本

  • 回顾性收集本机构放疗科286例胸部CT及对应左右乳腺轮廓,开发并验证乳腺分割U-Net,并对每例CT行蒙特卡洛模拟确定乳腺辐射剂量。

诊断标准与关键数值

  • 乳腺分割模型平均Dice相似系数0.92,左右乳腺精确率与敏感度均高于0.90。
  • 右乳平均预测剂量5.05±0.50 mGy vs 真实值5.03±0.57 mGy,平均绝对百分比误差4.01%(范围3.90%-4.12%)。
  • 左乳平均预测剂量5.06±0.55 mGy vs 真实值5.02±0.61 mGy,平均绝对百分比误差4.82%(范围4.56%-5.11%)。
  • 预测与真实剂量统计比较无显著差异(p≥0.07)。
  • 平均推理时间30.2±4.3 s。

临床应用与阅片要点

  • 作者指出该工作流可自动化完成CT乳腺辐射剂量学评估,整合分割与剂量预测模型。
  • 模型与代码公开于 https://github.com/eltzanis/ML-based-Breast-Radiation-Dosimetry-in-CT。

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

PURPOSE: To develop a machine learning-based workflow for patient-specific breast radiation dosimetry in CT.

MATERIALS AND METHODS: Two hundred eighty-six chest CT examinations, with corresponding right and left breast contours, were retrospectively collected from the radiotherapy department at our institution to develop and validate breast segmentation U-Nets. Additionally, Monte Carlo simulations were performed for each CT scan to determine radiation doses to the breasts. The derived breast doses, along with predictors such as X-ray tube current and radiomic features, were then used to train deep neural networks (DNNs) for breast dose prediction.

RESULTS: The breast segmentation models achieved a mean dice similarity coefficient of 0.92, with precision and sensitivity scores above 0.90 for both breasts, indicating high segmentation accuracy. The DNNs demonstrated close alignment with ground truth values, with mean predicted doses of 5.05 ± 0.50 mGy for the right breast and 5.06 ± 0.55 mGy for the left breast, compared to ground truth values of 5.03 ± 0.57 mGy and 5.02 ± 0.61 mGy, respectively. The mean absolute percentage errors were 4.01 % (range: 3.90 %-4.12 %) for the right breast and 4.82 % (range: 4.56 %-5.11 %) for the left breast. The mean inference time was 30.2 ± 4.3 s. Statistical analysis showed no significant differences between predicted and actual doses (p ≥ 0.07).

CONCLUSION: This study presents an automated, machine learning-based workflow for breast radiation dosimetry in CT, integrating segmentation and dose prediction models. The models and code are available at: https://github.com/eltzanis/ML-based-Breast-Radiation-Dosimetry-in-CT.

来源:Eur J Radiol,查看原文。版权归原作者所有。

引用

  1. Tzanis E, Stratakis J, Damilakis J, et al. A machine learning approach for personalized breast radiation dosimetry in CT: Integrating radiomics and deep neural networks. Eur J Radiol. 2025. PMID: 40516502. doi:10.1016/j.ejrad.2025.112236

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。