Assessing the efficacy of 3D Dual-CycleGAN model for multi-contrast MRI synthesis

The Egyptian Journal of Radiology and Nuclear Medicine, Journal Year: 2024, Volume and Issue: 55(1)

Published: June 11, 2024

Abstract Background This research presents a novel methodology for synthesizing 3D multi-contrast MRI images utilizing the Dual-CycleGAN architecture. The performance of model is evaluated on different sequences, including T1-weighted (T1W), contrast-enhanced (T1c), T2-weighted (T2W), and FLAIR sequences. Results Our approach demonstrates proficient learning capabilities in transforming T1W into target modalities. proposed framework encompasses combination loss functions voxel-wise, gradient difference, perceptual, structural similarity losses. These components, along with adversarial dual cycle-consistency losses, contribute significantly to realistic accurate syntheses. Evaluation metrics MAE, PMAE, RMSE, PCC, PSNR, SSIM are employed assess fidelity synthesized compared their ground truth counterparts. Empirical results indicate effectiveness generating T1c from inputs minimal average discrepancies (MAE 2.8 ± 2.61) strong (SSIM 0.82 0.28). Furthermore, synthesis T2W yields promising outcomes, demonstrating acceptable 3.87 3.32 3.82 FLAIR) reasonable similarities 0.28 0.80 0.29 relative original images. Conclusions findings underscore efficacy high-fidelity images, significant implications diverse applications field medical imaging.

Language: Английский

---

A systematic literature review: deep learning techniques for synthetic medical image generation and their applications in radiotherapy

Frontiers in Radiology, Journal Year: 2024, Volume and Issue: 4

Published: March 27, 2024

The aim of this systematic review is to determine whether Deep Learning (DL) algorithms can provide a clinically feasible alternative classic for synthetic Computer Tomography (sCT). following categories are presented in study: ∙ MR-based treatment planning and CT generation techniques. id="IM2">∙ Generation images based on Cone Beam images. id="IM3">∙ Low-dose High-dose generation. id="IM4">∙ Attenuation correction PET To perform appropriate database searches, we reviewed journal articles published between January 2018 June 2023. Current methodology, study strategies, results with relevant clinical applications were analyzed as outlined the state-of-the-art deep learning approaches inter-modality intra-modality image synthesis. This was accomplished by contrasting provided methodologies traditional research approaches. key contributions each category highlighted, specific challenges identified, accomplishments summarized. As final step, statistics all cited works from various aspects analyzed, which revealed that DL-based sCTs have achieved considerable popularity, while also showing potential technology. In order assess readiness methods, examined current status sCT

Language: Английский

---

Challenges and opportunities in the development and clinical implementation of artificial intelligence based synthetic computed tomography for magnetic resonance only radiotherapy

Radiotherapy and Oncology, Journal Year: 2024, Volume and Issue: 198, P. 110387 - 110387

Published: June 15, 2024

Language: Английский

---

Region of interest focused MRI to synthetic CT translation using regression and segmentation multi-task network

Physics in Medicine and Biology, Journal Year: 2023, Volume and Issue: 68(19), P. 195003 - 195003

Published: Aug. 11, 2023

Abstract Objective . In MR-only clinical workflow, replacing CT with MR image is of advantage for workflow efficiency and reduces radiation to the patient. An important step required eliminate scan from generate information provided by via an image. this work, we aim demonstrate a method accurate synthetic (sCT) suit therapy (RT) treatment planning workflow. We show feasibility make way broader evaluation. Approach present machine learning sCT generation zero-echo-time (ZTE) MRI aimed at structural quantitative accuracies image, particular focus on bone density value prediction. The misestimation in path could lead unintended dose delivery target volume results suboptimal outcome. propose loss function that favors spatially sparse region harness ability multi-task network produce correlated outputs as framework enable localization interest (RoI) segmentation, emphasize regression values within RoI still retain overall accuracy global regression. optimized composite combines dedicated each task. Main have included 54 brain patient images study tested against reference subset 20 cases. A pilot evaluation was performed 9 test cases viability generated RT planning. average metrics produced proposed over set were—(a) mean absolute error (MAE) 70 ± 8.6 HU; (b) peak signal-to-noise ratio (PSNR) 29.4 2.8 dB; similarity metric (SSIM) 0.95 0.02; (d) Dice coefficient body 0.984 0. Significance generates resemble visual characteristics real has suits application. compare calculation setup based falls 0.5% error. presented here initial makes encouraging precursor different anatomical regions.

Language: Английский

---

Computed tomography synthesis from magnetic resonance imaging using cycle Generative Adversarial Networks with multicenter learning

Physics and Imaging in Radiation Oncology, Journal Year: 2023, Volume and Issue: 28, P. 100511 - 100511

Published: Oct. 1, 2023

Background and Purpose: Addressing the need for accurate dose calculation in MRI-only radiotherapy, generation of synthetic Computed Tomography (sCT) from MRI has emerged. Deep learning (DL) techniques, have shown promising results achieving high sCT accuracies. However, existing synthesis methods are often center-specific, posing a challenge to their generalizability. To overcome this limitation, recent studies proposed approaches, such as multicenter training . Material methods: The purpose work was propose by DL, using 2D cycle-GAN on 128 prostate cancer patients, four different centers. Four cases were compared: monocenter cases, test another center, trainings center not included with an test. Trainings performed 20 patients. accuracy evaluation Mean Absolute Error, Error Peak-Signal-to-Noise-Ratio. Dose assessed gamma index Volume Histogram comparison. Results: Qualitative, quantitative show that sCTs seen did differ significantly. when involved unseen quality inferior. Conclusions: aim generalizable MR-to-CT synthesis. It only few data one cohort allows equivalent study.

Language: Английский

---

Utilizing Pix2Pix conditional generative adversarial networks to recover missing data in preclinical PET scanner sinogram gaps

Physica Medica, Journal Year: 2025, Volume and Issue: 133, P. 104971 - 104971

Published: April 14, 2025

Language: Английский

---

Synthetic CT generation from CBCT and MRI using StarGAN in the Pelvic Region

Radiation Oncology, Journal Year: 2025, Volume and Issue: 20(1)

Published: Feb. 4, 2025

Abstract Rationale and objectives This study evaluated StarGAN, a deep learning model designed to generate synthetic computed tomography (sCT) images from magnetic resonance imaging (MRI) cone-beam (CBCT) data using single model. The goal was provide accurate Hounsfield unit (HU) for dose calculation enable MRI simulation adaptive radiation therapy (ART) CBCT or MRI. We also compared the performance benefits of StarGAN commonly used CycleGAN. Materials methods CycleGAN were employed in this study. dataset comprised 53 cases pelvic cancer. Evaluation involved qualitative quantitative analyses, focusing on image quality distribution calculation. Results For sCT generated CBCT, demonstrated superior anatomical preservation based evaluation. Quantitatively, exhibited lower mean absolute error (MAE) body (42.8 ± 4.3 HU) bone (138.2 20.3), whereas produced higher MAE (50.8 5.2 (153.4 27.7 HU). Dosimetric evaluation showed difference (DD) within 2% planning target volume (PTV) body, with gamma passing rate (GPR) > 90% under 2%/2 mm criteria. MRI, favored provided by StarGAN. recorded (79.8 14 HU 253.6 30.9 bone) (94.7 7.4 353.6 34.9 bone). Both models achieved DD PTV GPR 90%. Conclusion While metrics, better preservation, highlighting its potential generation radiotherapy.

Language: Английский

---

Patient-specific three-dimensional dose distribution prediction via deep learning for prostate cancer therapy: Improvement with the structure loss

Physica Medica, Journal Year: 2023, Volume and Issue: 107, P. 102544 - 102544

Published: Feb. 10, 2023

Language: Английский

---

A deep learning model to generate synthetic CT for prostate MR-only radiotherapy dose planning: a multicenter study

Frontiers in Oncology, Journal Year: 2023, Volume and Issue: 13

Published: Nov. 28, 2023

Introduction For radiotherapy based solely on magnetic resonance imaging (MRI), generating synthetic computed tomography scans (sCT) from MRI is essential for dose calculation. The use of deep learning (DL) methods to generate sCT has shown encouraging results if the images used training network and generation come same device. objective this study was create evaluate a generic DL model capable sCTs various devices prostate Materials In total, 90 patients three centers (30 CT-MR pairs/center) underwent treatment using volumetric modulated arc therapy cancer (PCa) (60 Gy in 20 fractions). T2 were acquired addition (CT) planning. 2D supervised conditional generative adversarial (Pix2Pix). Patient preprocessing steps, including nonrigid registration. Seven different models trained, incorporating one, two, or centers. Each trained 24 pairs. A all To compare CT, mean absolute error Hounsfield units calculated entire pelvis, prostate, bladder, rectum, bones. analysis, differences D 99% CTV, V 95% PTV, max rectum 3D gamma analysis (local, 1%/1 mm) CT sCT. Furthermore, Wilcoxon tests performed image obtained with those other models. Results Considering when data test comes center as training, not significantly model. Absolute less than 1 CTV every center. showed nonsignificant between monocentric Conclusion accuracy sCT, terms dose, equivalent whether are generated model, only eight MRI-CT pairs per center, offers robust generation, facilitating PCa MRI-only routine clinical use.

Language: Английский

---

Mapping confinement potentials and charge densities of interacting quantum systems using conditional generative adversarial networks

Machine Learning Science and Technology, Journal Year: 2023, Volume and Issue: 4(2), P. 025023 - 025023

Published: May 18, 2023

Abstract Accurate and efficient tools for calculating the ground state properties of interacting quantum systems are essential in design nanoelectronic devices. The exact diagonalization method fully accounts Coulomb interaction beyond mean field approximations it is regarded as gold-standard few electron systems. However, by increasing number instances to be solved, computational costs become prohibitive new approaches based on machine learning techniques can provide a significant reduction time resources, maintaining reasonable accuracy. Here, we employ pix2pix , general-purpose image-to-image translation conditional generative adversarial network (cGAN), predicting densities from randomly generated confinement potentials. Other mappings were also investigated, like potentials non-interacting densities. architecture cGAN was optimized with respect internal parameters generator discriminator. Moreover, inverse problem finding potential given density approached mapping, which an important step near-optimal solutions

Language: Английский

---