Risk Estimation of Late Rectal Toxicity Using a Convolutional Neural Network-based Dose Prediction in Prostate Cancer Radiation Therapy

IF 2.2 Q3 ONCOLOGY

Advances in Radiation Oncology Pub Date : 2025-02-15 DOI:10.1016/j.adro.2025.101739

Seiya Takano MD, PhD , Natsuo Tomita MD, PhD , Taiki Takaoka MD, PhD , Machiko Ukai MD , Akane Matsuura MD , Masanosuke Oguri MD , Nozomi Kita MD, PhD , Akira Torii MD, PhD , Masanari Niwa MD, PhD , Dai Okazaki MD, PhD , Takahiro Yasui MD, PhD , Akio Hiwatashi MD, PhD

{"title":"Risk Estimation of Late Rectal Toxicity Using a Convolutional Neural Network-based Dose Prediction in Prostate Cancer Radiation Therapy","authors":"Seiya Takano MD, PhD , Natsuo Tomita MD, PhD , Taiki Takaoka MD, PhD , Machiko Ukai MD , Akane Matsuura MD , Masanosuke Oguri MD , Nozomi Kita MD, PhD , Akira Torii MD, PhD , Masanari Niwa MD, PhD , Dai Okazaki MD, PhD , Takahiro Yasui MD, PhD , Akio Hiwatashi MD, PhD","doi":"10.1016/j.adro.2025.101739","DOIUrl":null,"url":null,"abstract":"<div><h3>Purpose</h3><div>The present study investigated the feasibility of our automatic plan generation model based on a convolutional neural network (CNN) to estimate the baseline risk of grade ≥2 late rectal bleeding (G2-LRB) in volumetric modulated arc therapy for prostate cancer.</div></div><div><h3>Methods and Materials</h3><div>We built the 2-dimensional U-net model to predict dose distributions using the planning computed tomography and organs at risk masks as inputs. Seventy-five volumetric modulated arc therapy plans of prostate cancer, which were delivered at 74.8 Gy in 34 fractions with a uniform planning goal, were included: 60 for training and 5-fold cross-validation, and the remaining 15 for testing. Isodose volume dice similarity coefficient, dose-volume histogram, and normal tissue complication probability (NTCP) metrics between planned and CNN-predicted dose distributions were calculated. The primary endpoint was the goodness-of-fit, expressed as a coefficient of determination (<em>R</em><sup>2</sup>) value, in predicting the percentage of G2-LRB-Lyman-Kutcher-Burman-NTCP.</div></div><div><h3>Results</h3><div>In 15 test cases, 2-dimensional U-net predicted dose distributions with a mean isodose volume dice similarity coefficient value of 0.90 within the high-dose region (doses ≥ 50 Gy). Rectum V<sub>50Gy</sub>, V<sub>60Gy</sub>, and V<sub>70Gy</sub> were accurately predicted (<em>R</em><sup>2</sup> = 0.73, 0.82, and 0.87, respectively). Strong correlations were observed between planned and predicted G2-LRB-Lyman-Kutcher-Burman-NTCP (<em>R</em><sup>2</sup> = 0.80, <em>P</em> < .001), with a small percent mean absolute error (mean ± 1 standard deviation, 1.24% ± 1.42%).</div></div><div><h3>Conclusions</h3><div>A risk estimation of LRB using CNN-based automatic plan generation from anatomic information was feasible. These results will contribute to the development of a decision support system that identifies priority cases for preradiation therapy interventions, such as hydrogel spacer implantation.</div></div>","PeriodicalId":7390,"journal":{"name":"Advances in Radiation Oncology","volume":"10 4","pages":"Article 101739"},"PeriodicalIF":2.2000,"publicationDate":"2025-02-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Advances in Radiation Oncology","FirstCategoryId":"1085","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S2452109425000272","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q3","JCRName":"ONCOLOGY","Score":null,"Total":0}

引用次数: 0

Abstract

Purpose

The present study investigated the feasibility of our automatic plan generation model based on a convolutional neural network (CNN) to estimate the baseline risk of grade ≥2 late rectal bleeding (G2-LRB) in volumetric modulated arc therapy for prostate cancer.

Methods and Materials

We built the 2-dimensional U-net model to predict dose distributions using the planning computed tomography and organs at risk masks as inputs. Seventy-five volumetric modulated arc therapy plans of prostate cancer, which were delivered at 74.8 Gy in 34 fractions with a uniform planning goal, were included: 60 for training and 5-fold cross-validation, and the remaining 15 for testing. Isodose volume dice similarity coefficient, dose-volume histogram, and normal tissue complication probability (NTCP) metrics between planned and CNN-predicted dose distributions were calculated. The primary endpoint was the goodness-of-fit, expressed as a coefficient of determination (R²) value, in predicting the percentage of G2-LRB-Lyman-Kutcher-Burman-NTCP.

Results

In 15 test cases, 2-dimensional U-net predicted dose distributions with a mean isodose volume dice similarity coefficient value of 0.90 within the high-dose region (doses ≥ 50 Gy). Rectum V_50Gy, V_60Gy, and V_70Gy were accurately predicted (R² = 0.73, 0.82, and 0.87, respectively). Strong correlations were observed between planned and predicted G2-LRB-Lyman-Kutcher-Burman-NTCP (R² = 0.80, P < .001), with a small percent mean absolute error (mean ± 1 standard deviation, 1.24% ± 1.42%).

Conclusions

A risk estimation of LRB using CNN-based automatic plan generation from anatomic information was feasible. These results will contribute to the development of a decision support system that identifies priority cases for preradiation therapy interventions, such as hydrogel spacer implantation.

查看原文本刊更多论文

求助全文

约1分钟内获得全文求助全文

来源期刊

Advances in Radiation Oncology Medicine-Radiology, Nuclear Medicine and Imaging

CiteScore

4.60

自引率

4.30%

发文量

208

审稿时长

98 days

期刊介绍： The purpose of Advances is to provide information for clinicians who use radiation therapy by publishing: Clinical trial reports and reanalyses. Basic science original reports. Manuscripts examining health services research, comparative and cost effectiveness research, and systematic reviews. Case reports documenting unusual problems and solutions. High quality multi and single institutional series, as well as other novel retrospective hypothesis generating series. Timely critical reviews on important topics in radiation oncology, such as side effects. Articles reporting the natural history of disease and patterns of failure, particularly as they relate to treatment volume delineation. Articles on safety and quality in radiation therapy. Essays on clinical experience. Articles on practice transformation in radiation oncology, in particular: Aspects of health policy that may impact the future practice of radiation oncology. How information technology, such as data analytics and systems innovations, will change radiation oncology practice. Articles on imaging as they relate to radiation therapy treatment.