优化左侧乳腺癌放疗心脏剂量预测:识别深吸气屏气技术获益者的临床策略
本研究评估卷积神经网络(CNN)模型在左侧乳腺癌全乳放疗中预测自由呼吸(FB)与深吸气屏气(DIBH)下心脏剂量的性能,并基于心脏剂量学识别最可能从DIBH获益的患者。共纳入265例患者(200例回顾性训练集,65例前瞻性测试集)。模型预测心脏剂量参数相关系数达0.84–0.99,基于FB心脏平均剂量及其DIBH下降低值的分类正确识别90.8%的DIBH获益者(kappa=0.876);以ΔHeart Dmean≥1 Gy为阈值时,模型识别63.1%患者有显著DIBH获益,预测与临床分类一致率96.9%。
要点
- CNN模型可高精度预测左侧乳腺癌FB和DIBH下心脏及亚结构剂量参数。
- 基于心脏平均剂量及其DIBH降低值的分类正确识别90.8%的DIBH获益者。
- ΔHeart Dmean≥1 Gy阈值下模型识别63.1%患者获益,一致率96.9%。
核心结论
- 基于CNN的模型为预测心脏剂量和识别最可能从DIBH获益的患者提供了高效准确的框架。
研究设计与样本
- 共纳入265例接受全乳放疗的左侧乳腺癌患者,200例回顾性分配至训练集,65例前瞻性分配至测试集。
- CNN模型纳入器官结构和距靶区距离体积图等解剖数据,预测三维剂量分布。
主要发现
- 模型预测心脏、左前降支冠状动脉和心室在FB和DIBH下的平均剂量(Dmean)和D2%的相关系数为0.84–0.99。
- 模型准确预测上述结构的剂量体积直方图,临床值与预测值间无显著差异。
- 基于FB心脏Dmean及其DIBH下降低值的分类方法,模型正确识别90.8%的患者为DIBH获益者(kappa值0.876)。
- 应用ΔHeart Dmean≥1 Gy阈值时,模型识别63.1%的患者有显著DIBH获益,预测与临床分类一致率96.9%。
关键数值与推荐
- DIBH获益分类基于FB下心脏Dmean及其通过DIBH的降低值。
- ΔHeart Dmean≥1 Gy阈值下,63.1%患者被模型识别为有显著DIBH获益。
临床怎么用
- 可利用CNN模型预测FB和DIBH下心脏剂量,辅助筛选最可能从DIBH获益的左侧乳腺癌放疗患者。
局限性
- 未来研究需探索该模型在更广泛放疗场景中的适用性,并评估其对心脏结局的长期影响。
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)。
原文
PURPOSE: Deep inspiration breath-hold (DIBH) is widely used in breast cancer radiation therapy to reduce cardiac radiation exposure. However, not all patients benefit from it. This study evaluated the performance of a convolutional neural network (CNN) model to predict cardiac doses under free-breathing (FB) and DIBH conditions for left-sided breast cancer radiation therapy, aiming to identify patients most likely to benefit from DIBH based on cardiac dosimetry. METHODS AND MATERIALS: A total of 265 left-sided breast cancer patients undergoing whole-breast irradiation were included, with 200 retrospectively assigned to the training set and 65 prospectively assigned to the test set. The CNN model incorporated anatomic data, including organ structures and distance-to-target volume maps, to predict 3-dimensional dose distributions. Predicted dosimetric parameters were compared with clinical data to assess accuracy, and agreement between model-based and clinical classifications of DIBH benefit was evaluated using kappa statistics. RESULTS: The CNN model demonstrated high accuracy in predicting cardiac dosimetric parameters, with correlation coefficients ranging from 0.84 to 0.99 for mean dose (Dmean) and D2% in the heart, left anterior descending coronary artery, and ventricles under both FB and DIBH conditions. The model also accurately predicted dose-volume histograms for these structures, with no significant differences between clinical and predicted values. Using a classification approach based on heart Dmean in FB and its reduction via DIBH, the model correctly identified 90.8% of patients as DIBH beneficiaries (kappa value, 0.876). When applying a threshold of ΔHeart Dmean ≥1 Gy, the model identified significant DIBH benefits in 63.1% of patients, with 96.9% agreement between predicted and clinical classifications. CONCLUSIONS: The CNN-based model provides an efficient and accurate framework for predicting cardiac dose and identifying patients most likely to benefit from DIBH. Future studies should explore its applicability across broader radiation therapy scenarios and evaluate its long-term impact on cardiac outcomes.
引用
- Chen SY, Wang Y, Tang Y, et al. Optimizing Cardiac Dose Prediction in Left-Sided Breast Cancer Radiation Therapy: Clinical Strategies for Identifying Beneficiaries of the Deep Inspiration Breath-Hold Technique. Adv Radiat Oncol. 2026. PMID: 42199527. doi:10.1016/j.adro.2026.102057
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。