Development and evaluation of deep learning models for estimating the organ at-risk dose constraint from two-dimensional cine magnetic resonance imaging scans during irradiation.
Uložené v:
| Názov: | Development and evaluation of deep learning models for estimating the organ at-risk dose constraint from two-dimensional cine magnetic resonance imaging scans during irradiation. |
|---|---|
| Autori: | Tanaka S; Department of Radiation Oncology, Tohoku University Graduate School of Medicine, Sendai, Japan., Kadoya N; Department of Radiation Oncology, Tohoku University Graduate School of Medicine, Sendai, Japan., Lee W; Department of Radiation Oncology, Tohoku University Graduate School of Medicine, Sendai, Japan., Takagi H; Department of Radiological Technology, School of Health Sciences, Faculty of Medicine, Tohoku University, Sendai, Japan., Katsuta Y; Department of Radiation Oncology, Tohoku University Graduate School of Medicine, Sendai, Japan., Arai K; Department of Radiation Oncology, Tohoku University Graduate School of Medicine, Sendai, Japan., Xiao Y; Department of Radiation Oncology, Tohoku University Graduate School of Medicine, Sendai, Japan., Hoshino T; Department of Radiation Oncology, Tohoku University Graduate School of Medicine, Sendai, Japan., Takahashi N; Department of Radiation Oncology, Tohoku University Graduate School of Medicine, Sendai, Japan., Jingu K; Department of Radiation Oncology, Tohoku University Graduate School of Medicine, Sendai, Japan. |
| Zdroj: | Journal of applied clinical medical physics [J Appl Clin Med Phys] 2025 Dec; Vol. 26 (12), pp. e70403. |
| Spôsob vydávania: | Journal Article; Evaluation Study |
| Jazyk: | English |
| Informácie o časopise: | Publisher: Wiley on behalf of American Association of Physicists in Medicine Country of Publication: United States NLM ID: 101089176 Publication Model: Print Cited Medium: Internet ISSN: 1526-9914 (Electronic) Linking ISSN: 15269914 NLM ISO Abbreviation: J Appl Clin Med Phys Subsets: MEDLINE |
| Imprint Name(s): | Publication: 2017- : Malden, MA : Wiley on behalf of American Association of Physicists in Medicine Original Publication: Reston, VA : American College of Medical Physics, c2000- |
| Výrazy zo slovníka MeSH: | Deep Learning* , Prostatic Neoplasms*/radiotherapy , Prostatic Neoplasms*/diagnostic imaging , Organs at Risk*/radiation effects , Organs at Risk*/diagnostic imaging , Magnetic Resonance Imaging, Cine*/methods , Radiotherapy Planning, Computer-Assisted*/methods , Urinary Bladder*/radiation effects , Urinary Bladder*/diagnostic imaging , Image Processing, Computer-Assisted*/methods, Humans ; Male ; Radiotherapy Dosage ; Radiotherapy, Intensity-Modulated/methods ; Particle Accelerators ; Aged |
| Abstrakt: | Purpose: Two-dimensional (2D) cine magnetic resonance imaging (MRI), available with a MR-linear accelerator (MR-Linac), allows real-time visualization of anatomical information during irradiation. The present study aimed to develop and evaluate a deep learning model that can estimate the organ-at-risk (OAR) dose constraints (mainly bladder V37Gy) from 2D cine MRI. Methods: The present study enrolled 91 prostate cancer patients treated with MR-Linac. From 381 treatment fractions, sagittal images at the start and end of the 2D cine MRI were extracted. Additionally, 3D MRI data acquired pre- and post-irradiation were collected, from which bladder V37Gy was calculated. We designed the deep learning model to predict the end-of-irradiation bladder V37Gy value based on the bladder image on the end-of-irradiation 2D cine MRI. The model inputs included the start and end 2D cine MR images, a difference image between them, and the pre-irradiation bladder V37Gy. The model output was the post-irradiation bladder V37Gy. We utilized a five-fold cross-validation for model training and evaluated the performance using a test dataset. For reference, we also evaluated the predictions made using only the pre-irradiation bladder V37Gy. Results: In the test dataset, the model-predicted and true bladder V37Gy values showed a strong correlation (r = 0.89), with a mean absolute error (MAE) of 1.40 cm 3 . Using only the pre-irradiation bladder V37Gy value yielded an r of 0.79 and an MAE of 2.02 cm 3 . Our model also achieved an area under the curve, sensitivity, and specificity values of 0.98, 0.91, and 0.95, respectively, in detecting dose constraint violations (bladder V37Gy of > 10 cm 3 ). Conclusions: Our results demonstrated that deep learning can effectively predict the OAR dose constraints during irradiation. However, it is noteworthy that these results show only a limited improvement and are constrained by several limitations. (© 2025 The Author(s). Journal of Applied Clinical Medical Physics published by Wiley Periodicals, LLC on behalf of The American Association of Physicists in Medicine.) |
| References: | Radiat Oncol J. 2025 Mar;43(1):40-48. (PMID: 40200656) Med Phys. 2025 May;52(5):2785-2795. (PMID: 39904621) Phys Med. 2021 Mar;83:101-107. (PMID: 33756222) J Appl Clin Med Phys. 2023 Dec;24(12):e14122. (PMID: 37559561) Int J Radiat Oncol Biol Phys. 2018 Feb 1;100(2):361-373. (PMID: 29353654) Radiother Oncol. 2022 Jun;171:182-188. (PMID: 35489444) Cancers (Basel). 2025 Jun 17;17(12):. (PMID: 40563661) J Radiat Res. 2023 Sep 22;64(5):783-794. (PMID: 37466450) Med Phys. 2018 Jul 31;:. (PMID: 30066388) Radiother Oncol. 2023 May;182:109506. (PMID: 36736589) Transl Cancer Res. 2024 Feb 29;13(2):1131-1138. (PMID: 38482421) Radiother Oncol. 2022 Nov;176:25-30. (PMID: 36113777) Phys Imaging Radiat Oncol. 2020 Jul 13;15:23-29. (PMID: 33458322) Int J Radiat Oncol Biol Phys. 2004 Jun 1;59(2):436-44. (PMID: 15145160) Phys Med Biol. 2018 Jul 24;63(15):155003. (PMID: 29952320) Sci Rep. 2022 May 27;12(1):8899. (PMID: 35624113) PLoS One. 2024 Oct 30;19(10):e0312032. (PMID: 39475854) Radiother Oncol. 2020 Oct;151:88-94. (PMID: 32622779) Lancet Oncol. 2022 Oct;23(10):1308-1320. (PMID: 36113498) J Appl Clin Med Phys. 2025 Apr;26(4):e70005. (PMID: 39955657) Phys Med Biol. 2019 Dec 05;64(23):235008. (PMID: 31698351) J Appl Clin Med Phys. 2025 Feb;26(2):e14566. (PMID: 39540669) Radiat Oncol. 2020 Jul 10;15(1):168. (PMID: 32650811) Radiother Oncol. 2020 Apr;145:88-94. (PMID: 31931291) Radiother Oncol. 2023 Dec;189:109932. (PMID: 37778533) Radiother Oncol. 2021 Apr;157:197-202. (PMID: 33545251) Biomed Phys Eng Express. 2021 Aug 18;7(5):. (PMID: 34375963) J Appl Clin Med Phys. 2023 Oct;24(10):e14055. (PMID: 37261720) J Appl Clin Med Phys. 2025 Dec;26(12):e70403. (PMID: 41310917) Med Phys. 2021 Sep;48(9):4769-4783. (PMID: 34101848) |
| Grant Information: | JP23K07077 JSPS |
| Contributed Indexing: | Keywords: 2D cine MRI; MR‐Linac; deep learning; dose prediction; radiotherapy |
| Entry Date(s): | Date Created: 20251128 Date Completed: 20251128 Latest Revision: 20251201 |
| Update Code: | 20251201 |
| PubMed Central ID: | PMC12660054 |
| DOI: | 10.1002/acm2.70403 |
| PMID: | 41310917 |
| Databáza: | MEDLINE |
Full Text 
Full Text Finder 
Nájsť tento článok vo Web of Science | Abstrakt: | Purpose: Two-dimensional (2D) cine magnetic resonance imaging (MRI), available with a MR-linear accelerator (MR-Linac), allows real-time visualization of anatomical information during irradiation. The present study aimed to develop and evaluate a deep learning model that can estimate the organ-at-risk (OAR) dose constraints (mainly bladder V37Gy) from 2D cine MRI.<br />Methods: The present study enrolled 91 prostate cancer patients treated with MR-Linac. From 381 treatment fractions, sagittal images at the start and end of the 2D cine MRI were extracted. Additionally, 3D MRI data acquired pre- and post-irradiation were collected, from which bladder V37Gy was calculated. We designed the deep learning model to predict the end-of-irradiation bladder V37Gy value based on the bladder image on the end-of-irradiation 2D cine MRI. The model inputs included the start and end 2D cine MR images, a difference image between them, and the pre-irradiation bladder V37Gy. The model output was the post-irradiation bladder V37Gy. We utilized a five-fold cross-validation for model training and evaluated the performance using a test dataset. For reference, we also evaluated the predictions made using only the pre-irradiation bladder V37Gy.<br />Results: In the test dataset, the model-predicted and true bladder V37Gy values showed a strong correlation (r = 0.89), with a mean absolute error (MAE) of 1.40 cm <sup>3</sup> . Using only the pre-irradiation bladder V37Gy value yielded an r of 0.79 and an MAE of 2.02 cm <sup>3</sup> . Our model also achieved an area under the curve, sensitivity, and specificity values of 0.98, 0.91, and 0.95, respectively, in detecting dose constraint violations (bladder V37Gy of > 10 cm <sup>3</sup> ).<br />Conclusions: Our results demonstrated that deep learning can effectively predict the OAR dose constraints during irradiation. However, it is noteworthy that these results show only a limited improvement and are constrained by several limitations.<br /> (© 2025 The Author(s). Journal of Applied Clinical Medical Physics published by Wiley Periodicals, LLC on behalf of The American Association of Physicists in Medicine.) |
|---|---|
| ISSN: | 1526-9914 |
| DOI: | 10.1002/acm2.70403 |