Hybrid deep learning enables multi-institutional delineation of active bone marrow for gynecologic radiotherapy

Background and purpose: Pelvic radiotherapy for gynecologic cancer inevitably irradiates sensitive areas like iliac bones, lumbar vertebrae, and sacrum. Using 18F-FDG PET/CT as a reference, we developed a deep learning method to detect hematopoietic active bone marrow (ABM) on CT in gynecologic canc...

Full description

Saved in:
Bibliographic Details
Main Authors: Zhe Zhang, Xiao Lu, Sicheng He, Tao Huang, Shaobin Wang, Mingjun Lu, Xiaomin Zhang, Zhibo Tan, John Moraros, Lei Zhang, Xin Li, Zhan Li, Zihao Deng, Yimeng Zhang, Mengjie Dong, Shuihua Wang, Yajie Liu
Format: Article
Language:English
Published: Elsevier 2025-07-01
Series:Physics and Imaging in Radiation Oncology
Subjects:
Online Access:http://www.sciencedirect.com/science/article/pii/S2405631625001289
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849239519268700160
author Zhe Zhang
Xiao Lu
Sicheng He
Tao Huang
Shaobin Wang
Mingjun Lu
Xiaomin Zhang
Zhibo Tan
John Moraros
Lei Zhang
Xin Li
Zhan Li
Zihao Deng
Yimeng Zhang
Mengjie Dong
Shuihua Wang
Yajie Liu
author_facet Zhe Zhang
Xiao Lu
Sicheng He
Tao Huang
Shaobin Wang
Mingjun Lu
Xiaomin Zhang
Zhibo Tan
John Moraros
Lei Zhang
Xin Li
Zhan Li
Zihao Deng
Yimeng Zhang
Mengjie Dong
Shuihua Wang
Yajie Liu
author_sort Zhe Zhang
collection DOAJ
description Background and purpose: Pelvic radiotherapy for gynecologic cancer inevitably irradiates sensitive areas like iliac bones, lumbar vertebrae, and sacrum. Using 18F-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. Materials and methods: 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. Results: 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. Conclusion: This multi-institutional study confirms that the hybrid nnU-Net accurately segments ABM from CT images, showing potential for ABM-sparing radiotherapy.
format Article
id doaj-art-8a1266f048364df78ce5da3a3b14bcdf
institution Kabale University
issn 2405-6316
language English
publishDate 2025-07-01
publisher Elsevier
record_format Article
series Physics and Imaging in Radiation Oncology
spelling doaj-art-8a1266f048364df78ce5da3a3b14bcdf2025-08-20T04:00:55ZengElsevierPhysics and Imaging in Radiation Oncology2405-63162025-07-013510082310.1016/j.phro.2025.100823Hybrid deep learning enables multi-institutional delineation of active bone marrow for gynecologic radiotherapyZhe Zhang0Xiao Lu1Sicheng He2Tao Huang3Shaobin Wang4Mingjun Lu5Xiaomin Zhang6Zhibo Tan7John Moraros8Lei Zhang9Xin Li10Zhan Li11Zihao Deng12Yimeng Zhang13Mengjie Dong14Shuihua Wang15Yajie Liu16Department 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; Department of Computer Sciences, University of Liverpool, Liverpool, UKDepartment of Radiation Oncology, The First Affiliated Hospital of Soochow University, Suzhou, ChinaInterventional Diagnostic Centre, Zhongnan Hospital Wuhan University, Wuhan, ChinaDepartment of Biosciences and Bioinformatics, Suzhou Municipal Key Lab AI4Health, School of Science, Xi’an Jiaotong-Liverpool University, Suzhou, China; Department of Computer Sciences, University of Liverpool, Liverpool, UKDepartment of Biomedical Engineering, School of Medicine, Tsinghua University, Beijing, China; Department of Research and Development, MedMind Technology Co, Ltd., Beijing, ChinaDepartment of Radiology, Zhujiang Hospital Southern Medical University, Guangzhou, ChinaDepartment of Radiation Oncology, Peking University Shenzhen Hospital, Hong Kong University of Science and Technology Medical Center, Shenzhen, ChinaDepartment of Radiation Oncology, Proton and Heavy Ion Center, Heyou Hospital, Heyou International Health System, Foshan, Guangdong, ChinaDepartment of Biosciences and Bioinformatics, Suzhou Municipal Key Lab AI4Health, School of Science, Xi’an Jiaotong-Liverpool University, Suzhou, China; Department of Computer Sciences, University of Liverpool, Liverpool, UKDepartment of Radiation Oncology, Peking University Shenzhen Hospital, Hong Kong University of Science and Technology Medical Center, Shenzhen, ChinaDepartment of Radiation Oncology, Peking University Shenzhen Hospital, Hong Kong University of Science and Technology Medical Center, Shenzhen, ChinaDepartment of Nuclear Medicine, Peking University Shenzhen Hospital, Shenzhen, ChinaCollege of Life Sciences and Oceanography, Shenzhen University, Shenzhen, ChinaDepartment of Research and Development, MedMind Technology Co, Ltd., Beijing, ChinaDepartment of Nuclear Medicine, Peking University Shenzhen Hospital, Shenzhen, China; Corresponding author at: Department of Radiation Oncology, Peking University Shenzhen Hospital, Shenzhen, Guangdong, ChinaDepartment of Biosciences and Bioinformatics, Suzhou Municipal Key Lab AI4Health, School of Science, Xi’an Jiaotong-Liverpool University, Suzhou, China; Department of Computer Sciences, University of Liverpool, Liverpool, UK; Corresponding author: Department of Biosciences and Bioinformatics, Xi’an Jiaotong-Liverpool University, Suzhou, Jiangsu, ChinaDepartment of Radiation Oncology, Peking University Shenzhen Hospital, Hong Kong University of Science and Technology Medical Center, Shenzhen, China; Corresponding author: Department of Nuclear Medicine, Peking University Shenzhen Hospital, Shenzhen, Guangdong, China.Background and purpose: Pelvic radiotherapy for gynecologic cancer inevitably irradiates sensitive areas like iliac bones, lumbar vertebrae, and sacrum. Using 18F-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. Materials and methods: 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. Results: 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. Conclusion: This multi-institutional study confirms that the hybrid nnU-Net accurately segments ABM from CT images, showing potential for ABM-sparing radiotherapy.http://www.sciencedirect.com/science/article/pii/S2405631625001289Active bone marrowGynecologic radiotherapyHybrid deep learning segmentationMulti-institutional studyHematologic toxicity
spellingShingle Zhe Zhang
Xiao Lu
Sicheng He
Tao Huang
Shaobin Wang
Mingjun Lu
Xiaomin Zhang
Zhibo Tan
John Moraros
Lei Zhang
Xin Li
Zhan Li
Zihao Deng
Yimeng Zhang
Mengjie Dong
Shuihua Wang
Yajie Liu
Hybrid deep learning enables multi-institutional delineation of active bone marrow for gynecologic radiotherapy
Physics and Imaging in Radiation Oncology
Active bone marrow
Gynecologic radiotherapy
Hybrid deep learning segmentation
Multi-institutional study
Hematologic toxicity
title Hybrid deep learning enables multi-institutional delineation of active bone marrow for gynecologic radiotherapy
title_full Hybrid deep learning enables multi-institutional delineation of active bone marrow for gynecologic radiotherapy
title_fullStr Hybrid deep learning enables multi-institutional delineation of active bone marrow for gynecologic radiotherapy
title_full_unstemmed Hybrid deep learning enables multi-institutional delineation of active bone marrow for gynecologic radiotherapy
title_short Hybrid deep learning enables multi-institutional delineation of active bone marrow for gynecologic radiotherapy
title_sort hybrid deep learning enables multi institutional delineation of active bone marrow for gynecologic radiotherapy
topic Active bone marrow
Gynecologic radiotherapy
Hybrid deep learning segmentation
Multi-institutional study
Hematologic toxicity
url http://www.sciencedirect.com/science/article/pii/S2405631625001289
work_keys_str_mv AT zhezhang hybriddeeplearningenablesmultiinstitutionaldelineationofactivebonemarrowforgynecologicradiotherapy
AT xiaolu hybriddeeplearningenablesmultiinstitutionaldelineationofactivebonemarrowforgynecologicradiotherapy
AT sichenghe hybriddeeplearningenablesmultiinstitutionaldelineationofactivebonemarrowforgynecologicradiotherapy
AT taohuang hybriddeeplearningenablesmultiinstitutionaldelineationofactivebonemarrowforgynecologicradiotherapy
AT shaobinwang hybriddeeplearningenablesmultiinstitutionaldelineationofactivebonemarrowforgynecologicradiotherapy
AT mingjunlu hybriddeeplearningenablesmultiinstitutionaldelineationofactivebonemarrowforgynecologicradiotherapy
AT xiaominzhang hybriddeeplearningenablesmultiinstitutionaldelineationofactivebonemarrowforgynecologicradiotherapy
AT zhibotan hybriddeeplearningenablesmultiinstitutionaldelineationofactivebonemarrowforgynecologicradiotherapy
AT johnmoraros hybriddeeplearningenablesmultiinstitutionaldelineationofactivebonemarrowforgynecologicradiotherapy
AT leizhang hybriddeeplearningenablesmultiinstitutionaldelineationofactivebonemarrowforgynecologicradiotherapy
AT xinli hybriddeeplearningenablesmultiinstitutionaldelineationofactivebonemarrowforgynecologicradiotherapy
AT zhanli hybriddeeplearningenablesmultiinstitutionaldelineationofactivebonemarrowforgynecologicradiotherapy
AT zihaodeng hybriddeeplearningenablesmultiinstitutionaldelineationofactivebonemarrowforgynecologicradiotherapy
AT yimengzhang hybriddeeplearningenablesmultiinstitutionaldelineationofactivebonemarrowforgynecologicradiotherapy
AT mengjiedong hybriddeeplearningenablesmultiinstitutionaldelineationofactivebonemarrowforgynecologicradiotherapy
AT shuihuawang hybriddeeplearningenablesmultiinstitutionaldelineationofactivebonemarrowforgynecologicradiotherapy
AT yajieliu hybriddeeplearningenablesmultiinstitutionaldelineationofactivebonemarrowforgynecologicradiotherapy