Evaluation of a cone-beam computed tomography system calibrated for accurate radiotherapy dose calculation

Physics and Imaging in Radiation Oncology, Journal Year: 2024, Volume and Issue: 29, P. 100566 - 100566

Published: Jan. 1, 2024

Language: Английский

---

Transformer CycleGAN with uncertainty estimation for CBCT based synthetic CT in adaptive radiotherapy

Physics in Medicine and Biology, Journal Year: 2024, Volume and Issue: 69(3), P. 035014 - 035014

Published: Jan. 10, 2024

Abstract Objective . Clinical implementation of synthetic CT (sCT) from cone-beam (CBCT) for adaptive radiotherapy necessitates a high degree anatomical integrity, Hounsfield unit (HU) accuracy, and image quality. To achieve these goals, vision-transformer anatomically sensitive loss functions are described. Better quantification quality is achieved using the alignment-invariant Fréchet inception distance (FID), uncertainty estimation sCT risk prediction implemented in scalable plug-and-play manner. Approach Baseline U-Net, generative adversarial network (GAN), CycleGAN models were trained to identify shortcomings each approach. The proposed CycleGAN-Best model was empirically optimized based on large ablation study evaluated classical metrics, FID, gamma index, segmentation analysis. Two methods, Monte-Carlo Dropout (MCD) test-time augmentation (TTA), introduced epistemic aleatoric uncertainty. Main results FID correlated blind observer scores with Correlation Coefficient −0.83, validating metric as an accurate quantifier perceived mean absolute error (MAE) 42.11 ± 5.99 25.00 1.97 HU, compared 63.42 15.45 31.80 HU CycleGAN-Baseline, 144.32 20.91 68.00 5.06 CBCT, respectively. Gamma 1%/1 mm pass rates 98.66 0.54% CycleGAN-Best, 86.72 2.55% CBCT. TTA MCD-based maps well spatially poor synthesis outputs. Significance Anatomical accuracy by suppressing CycleGAN-related artefacts. better discriminated quality, where alignment-based metrics such MAE erroneously suggest poorer outputs perform better. Uncertainty shown correlate has clinical relevancy toward assessment assurance. accompanying evaluation tools necessary additions clinically robust generation models.

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

---

Assessment of bias in scoring of AI-based radiotherapy segmentation and planning studies using modified TRIPOD and PROBAST guidelines as an example

Radiotherapy and Oncology, Journal Year: 2024, Volume and Issue: 194, P. 110196 - 110196

Published: March 2, 2024

Background and purposeStudies investigating the application of Artificial Intelligence (AI) in field radiotherapy exhibit substantial variations terms quality. The goal this study was to assess amount transparency bias scoring articles with a specific focus on AI based segmentation treatment planning, using modified PROBAST TRIPOD checklists, order provide recommendations for future guideline developers reviewers.Materials methodsThe checklist items were discussed Delphi process. After consensus reached, 2 groups 3 co-authors scored evaluate usability further optimize adapted checklists. Finally, 10 by all co-authors. Fleiss' kappa calculated reliability agreement between observers.ResultsThree 37 5 32 deemed irrelevant. General terminology (e.g., multivariable prediction model, predictors) align AI-specific terms. first round, improvements formulated, e.g., preventing use sub-questions or subjective words adding clarifications how score an item. Using final list articles, only out 61 resulted statistically significant 0.4 more demonstrating agreement. For 41 no obtained indicating that level among multiple observers is due chance alone.ConclusionOur showed low scores Although such checklists have shown great value during development reporting, raises concerns about applicability objectively scientific applications. When developing revising guidelines, it essential consider their without introducing bias.

Language: Английский

---

Uncertainty estimation in female pelvic synthetic computed tomography generated from iterative reconstructed cone-beam computed tomography

Physics and Imaging in Radiation Oncology, Journal Year: 2025, Volume and Issue: 33, P. 100743 - 100743

Published: Jan. 1, 2025

Language: Английский

---

Physics-based data augmentation for improved training of cone-beam computed tomography auto-segmentation of the female pelvis

Physics and Imaging in Radiation Oncology, Journal Year: 2025, Volume and Issue: 34, P. 100744 - 100744

Published: March 7, 2025

Labeling cone-beam computed tomography (CBCT) images is challenging due to poor image quality. Training auto-segmentation models without labelled data often involves deep-learning generate synthetic CBCTs (sCBCT) from planning CTs (pCT), which can result in anatomical mismatches and inaccurate labels. To prevent this issue, study assesses an model for female pelvic CBCT scans exclusively trained on delineated pCTs, were transformed into sCBCT using a physics-driven approach. replicate noise artefacts, (Ph-sCBCT) was synthesized pCT water-phantom scans. A 3D nn-UNet of cervical cancer Ph-sCBCT with contours. This included patients: 63 training, 16 validation 20 each testing Ph-sCBCTs clinical CBCTs. Auto-segmentations bladder, rectum target volume (CTV) evaluated Dice Similarity Coefficient (DSC) 95th percentile Hausdorff Distance (HD95). Initial evaluation occurred before generalizability The performed well generalized CBCTs, yielding median DSC's 0.96 0.94 the 0.88 0.81 rectum, 0.89 0.82 CTV CBCT, respectively. Median HD95's 5 mm 7 CBCT. demonstrates successful training images, necessarily delineating manually.

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

---

Evaluating synthetic computed tomography images for adaptive radiotherapy decision making in head and neck cancer

Physics and Imaging in Radiation Oncology, Journal Year: 2023, Volume and Issue: 27, P. 100478 - 100478

Published: July 1, 2023

Language: Английский

---

Current and future developments of synthetic computed tomography generation for radiotherapy

Physics and Imaging in Radiation Oncology, Journal Year: 2023, Volume and Issue: 28, P. 100521 - 100521

Published: Oct. 1, 2023

Language: Английский

---

Generating synthetic images from cone beam computed tomography using self-attention residual UNet for head and neck radiotherapy

Physics and Imaging in Radiation Oncology, Journal Year: 2023, Volume and Issue: 28, P. 100512 - 100512

Published: Oct. 1, 2023

Background and purposeAccurate CT numbers in Cone Beam (CBCT) are crucial for precise dose calculations adaptive radiotherapy (ART). This study aimed to generate synthetic (sCT) from CBCT using deep learning (DL) models head neck (HN) radiotherapy.Materials methodsA novel DL model, the 'self-attention-residual-UNet' (ResUNet), was developed accurate sCT generation. ResUNet incorporates a self-attention mechanism its long skip connections enhance information transfer between encoder decoder. Data 93 HN patients, each with planning (pCT) first-day images were used. Model performance evaluated two approaches (non-adversarial adversarial training) model types (2D axial only vs. 2.5D axial, sagittal, coronal). compared traditional UNet through image quality assessment (Mean Absolute Error (MAE), Peak-Signal-to-Noise Ratio (PSNR), Structural Similarity Index (SSIM)) calculation accuracy evaluation (DVH deviation gamma (1 %/1mm)).ResultsImage similarity results 2.5D-ResUNet 2.5D-UNet were: MAE: 46±7 HU 51±9 HU, PSNR: 66.6±2.0 dB 65.8±1.8 dB, SSIM: 0.81±0.04 0.79±0.05. There no significant differences models. Both demonstrated DVH below 0.5 % gamma-pass-rate %/1mm) exceeding 97 %.ConclusionsResUNet enhanced number of outperformed generation CBCT. method holds promise generating ART.

Language: Английский

---