Dataset Information

3D Unsupervised deep learning method for magnetic resonance imaging-to-computed tomography synthesis in prostate radiotherapy.

ABSTRACT:

Background and purpose

Magnetic resonance imaging (MRI)-to-computed tomography (CT) synthesis is essential in MRI-only radiotherapy workflows, particularly through deep learning techniques known for their accuracy. However, current supervised methods are limited to specific center's learnings and depend on registration precision. The aim of this study was to evaluate the accuracy of unsupervised and supervised approaches in the context of prostate MRI-to-CT generation for radiotherapy dose calculation.

Methods

CT/MRI image pairs from 99 prostate cancer patients across three different centers were used. A comparison between supervised and unsupervised conditional Generative Adversarial Networks (cGAN) was conducted. Unsupervised training incorporates a style transfer method with. Content and Style Representation for Enhanced Perceptual synthesis (CREPs) loss. For dose evaluation, the photon prescription dose was 60 Gy delivered in volumetric modulated arc therapy (VMAT). Imaging endpoint for sCT evaluation was Mean Absolute Error (MAE). Dosimetric endpoints included absolute dose differences and gamma analysis between CT and sCT dose calculations.

Results

The unsupervised paired network exhibited the highest accuracy for the body with a MAE at 33.6 HU, the highest MAE was 45.5 HU obtained with unsupervised unpaired learning. All architectures provided clinically acceptable results for dose calculation with gamma pass rates above 94 % (1 % 1 mm 10 %).

Conclusions

This study shows that multicenter data can produce accurate sCTs via unsupervised learning, eliminating CT-MRI registration. The sCTs not only matched HU values but also enabled precise dose calculations, suggesting their potential for wider use in MRI-only radiotherapy workflows.

SUBMITTER: Texier B

PROVIDER: S-EPMC11332181 | biostudies-literature | 2024 Jul

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

3D Unsupervised deep learning method for magnetic resonance imaging-to-computed tomography synthesis in prostate radiotherapy.

Texier Blanche B Hémon Cédric C Queffélec Adélie A Dowling Jason J Bessieres Igor I Greer Peter P Acosta Oscar O Boue-Rafle Adrien A de Crevoisier Renaud R Lafond Caroline C Castelli Joël J Barateau Anaïs A Nunes Jean-Claude JC

Physics and imaging in radiation oncology 20240719

<h4>Background and purpose</h4>Magnetic resonance imaging (MRI)-to-computed tomography (CT) synthesis is essential in MRI-only radiotherapy workflows, particularly through deep learning techniques known for their accuracy. However, current supervised methods are limited to specific center's learnings and depend on registration precision. The aim of this study was to evaluate the accuracy of unsupervised and supervised approaches in the context of prostate MRI-to-CT generation for radiotherapy do ...[more]

PMID: 39161728

Similar Datasets

Project description:ImportanceMagnetic resonance imaging (MRI) guidance offers multiple theoretical advantages in the context of stereotactic body radiotherapy (SBRT) for prostate cancer. However, to our knowledge, these advantages have yet to be demonstrated in a randomized clinical trial.ObjectiveTo determine whether aggressive margin reduction with MRI guidance significantly reduces acute grade 2 or greater genitourinary (GU) toxic effects after prostate SBRT compared with computed tomography (CT) guidance.Design, setting, and participantsThis phase 3 randomized clinical trial (MRI-Guided Stereotactic Body Radiotherapy for Prostate Cancer [MIRAGE]) enrolled men aged 18 years or older who were receiving SBRT for clinically localized prostate adenocarcinoma at a single center between May 5, 2020, and October 1, 2021. Data were analyzed from January 15, 2021, through May 15, 2022. All patients had 3 months or more of follow-up.InterventionsPatients were randomized 1:1 to SBRT with CT guidance (control arm) or MRI guidance. Planning margins of 4 mm (CT arm) and 2 mm (MRI arm) were used to deliver 40 Gy in 5 fractions.Main outcomes and measuresThe primary end point was the incidence of acute (≤90 days after SBRT) grade 2 or greater GU toxic effects (using Common Terminology Criteria for Adverse Events, version 4.03 [CTCAE v4.03]). Secondary outcomes included CTCAE v4.03-based gastrointestinal toxic effects and International Prostate Symptom Score (IPSS)-based and Expanded Prostate Cancer Index Composite-26 (EPIC-26)-based outcomes.ResultsBetween May 2020 and October 2021, 156 patients were randomized: 77 to CT (median age, 71 years [IQR, 67-77 years]) and 79 to MRI (median age, 71 years [IQR, 68-75 years]). A prespecified interim futility analysis conducted after 100 patients reached 90 or more days after SBRT was performed October 1, 2021, with the sample size reestimated to 154 patients. Thus, the trial was closed to accrual early. The incidence of acute grade 2 or greater GU toxic effects was significantly lower with MRI vs CT guidance (24.4% [95% CI, 15.4%-35.4%] vs 43.4% [95% CI, 32.1%-55.3%]; P = .01), as was the incidence of acute grade 2 or greater gastrointestinal toxic effects (0.0% [95% CI, 0.0%-4.6%] vs 10.5% [95% CI, 4.7%-19.7%]; P = .003). Magnetic resonance imaging guidance was associated with a significantly smaller percentage of patients with a 15-point or greater increase in IPSS at 1 month (6.8% [5 of 72] vs 19.4% [14 of 74]; P = .01) and a significantly reduced percentage of patients with a clinically significant (≥12-point) decrease in EPIC-26 bowel scores (25.0% [17 of 68] vs 50.0% [34 of 68]; P = .001) at 1 month.Conclusions and relevanceIn this randomized clinical trial, compared with CT-guidance, MRI-guided SBRT significantly reduced both moderate acute physician-scored toxic effects and decrements in patient-reported quality of life. Longer-term follow-up will confirm whether these notable benefits persist.Trial registrationClinicalTrials.gov Identifier: NCT04384770.

Dataset Information

3D Unsupervised deep learning method for magnetic resonance imaging-to-computed tomography synthesis in prostate radiotherapy.

Background and purpose

Methods

Results

Conclusions

Publications

3D Unsupervised deep learning method for magnetic resonance imaging-to-computed tomography synthesis in prostate radiotherapy.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets