Zhang Zhe, Lu Xiao, He Sicheng, Huang Tao, Wang Shaobin, Lu Mingjun, Zhang Xiaomin, Tan Zhibo, Moraros John, Zhang Lei, Li Xin, Li Zhan, Deng Zihao, Zhang Yimeng, Dong Mengjie, Wang Shuihua, Liu Yajie
Department of Radiation Oncology, Peking University Shenzhen Hospital, Hong Kong University of Science and Technology Medical Center, Shenzhen, China.
Department of Biosciences and Bioinformatics, Suzhou Municipal Key Lab AI4Health, School of Science, Xi'an Jiaotong-Liverpool University, Suzhou, China.
Phys Imaging Radiat Oncol. 2025 Aug 8;35:100823. doi: 10.1016/j.phro.2025.100823. eCollection 2025 Jul.
Pelvic radiotherapy for gynecologic cancer inevitably irradiates sensitive areas like iliac bones, lumbar vertebrae, and sacrum. Using F-FDG PET/CT as a reference, we developed a deep learning method to detect hematopoietic active bone marrow (ABM) on CT in gynecologic cancer patients and assess clinical benefits.
We analyzed 319 patients from five institutions retrospectively. ABM was divided into three 18F-FDG PET/CT-defined subregions: active iliac bone marrow (A_IBM), active sacral bone marrow (A_SBM), and active lumbar vertebrae bone marrow (A_LVBM), defined as areas with standardized uptake values exceeding subregional means. Six deep learning models were trained: hybrid nnU-Net, U-Net, V-Net, ResU-Net, nnU-Net, and UNETR. The hybrid nnU-Net approach integrated nnU-Net predictions with anatomical bone structures via Boolean operations, providing a post-processing strategy. The dataset was split into 290 cases for training and 29 for independent testing. Performance was evaluated using Dice similarity coefficients (DSCs) and 95th percentile Hausdorff distance (HD95). Two clinical cases were prospectively evaluated for ABM-sparing radiotherapy with hematologic monitoring.
The hybrid nnU-Net achieved the highest DSCs for A_IBM (0.74 ± 0.06), A_LVBM (0.79 ± 0.07), and A_SBM (0.75 ± 0.06), with significant improvements over most models (p < 0.001), except nnU-Net. Despite ResU-Net's lower HD95 in two subregions, hybrid nnU-Net showed superior accuracy. No grade ≥2 hematologic toxicity occurred in prospective cases.
This multi-institutional study confirms that the hybrid nnU-Net accurately segments ABM from CT images, showing potential for ABM-sparing radiotherapy.
妇科癌症盆腔放疗不可避免地会照射髂骨、腰椎和骶骨等敏感区域。以F-FDG PET/CT为参考，我们开发了一种深度学习方法，用于在CT上检测妇科癌症患者的造血活性骨髓（ABM）并评估临床益处。
我们回顾性分析了来自五个机构的319例患者。ABM被分为三个由18F-FDG PET/CT定义的子区域：活性髂骨骨髓（A_IBM）、活性骶骨骨髓（A_SBM）和活性腰椎骨髓（A_LVBM），定义为标准化摄取值超过子区域均值的区域。训练了六个深度学习模型：混合nnU-Net、U-Net、V-Net、ResU-Net、nnU-Net和UNETR。混合nnU-Net方法通过布尔运算将nnU-Net预测结果与解剖学骨结构相结合，提供了一种后处理策略。数据集分为290例用于训练，29例用于独立测试。使用Dice相似系数（DSC）和第95百分位数豪斯多夫距离（HD95）评估性能。前瞻性评估了两例临床病例的ABM保留放疗及血液学监测情况。
混合nnU-Net在A_IBM（0.74±0.06）、A_LVBM（0.79±0.07）和A_SBM（0.75±0.06）方面获得了最高的DSC，与大多数模型相比有显著提高（p<0.001），nnU-Net除外。尽管ResU-Net在两个子区域的HD95较低，但混合nnU-Net显示出更高的准确性。前瞻性病例中未发生≥2级血液学毒性。
这项多机构研究证实，混合nnU-Net能准确地从CT图像中分割出ABM，显示出在ABM保留放疗中的潜力。
