Purpose This study aims to identify the optimal virtual monoenergetic image (VMI) energy levels for fast-kilovolt-switching dual-energy CT (FVS-DECT) to replace conventional single-energy CT (SECT) in proton therapy, and validate the accuracy of DECT-derived stopping power ratio (SPR) maps for dose calculation. Methods The Catphan503 phantom was scanned using FVS-DECT to reconstruct VMIs (40–140 keV). Image quality metrics were evaluated to identify the VMI energy level most comparable to 120 kVp SECT. The Cirs062 phantom was used to establish the optimal VMI pair for SPR estimation. Validation was performed using the Gammex467 phantom and a custom phantom containing porcine tissue samples. Reference SPR values for the porcine tissues were derived from their measured mass densities and elemental compositions obtained by chemical analysis. Six single-field intensity-modulated proton therapy (IMPT) plans were generated on the custom phantom to evaluate dose calculation accuracy. Results 70 keV VMIs provided CT numbers and image quality comparable to 120-kVp SECT. The combination of 70 keV and 120 keV VMIs enabled accurate SPR prediction. Mean absolute SPR errors were 0.8% and 0.7% for the Gammex467 and custom phantoms, respectively. Compared with the reference dose calculated on SPRREF maps, IMPT plans recalculated on DECT-derived SPR maps achieved mean gamma passing rates of 93.8% and 98.2% under 2 mm/2% and 3 mm/3% criteria, respectively. Except for the plan through the porcine lung, D1% and D99% changes were within 1%, and the mean ΔR80 was within 0.5 mm. Conclusion 70 keV VMIs can potentially replace SECT for contouring. DECT VMIs at 70 and 120 keV allow accurate SPR prediction and improve dose calculation accuracy compared with SECT-based methods.

Stereotactic arrhythmia radioablation (STAR) is an emerging treatment for refractory or recurrent arrhythmias. Compared with conventional stereotactic body radiotherapy (SBRT), STAR involves greater complexity in target delineation, motion management, and organs at risk (OARs) protection, yet it lacks established consensus clinical guidelines, and workflow-specific risk analyses remain limited. To develop a C-arm linear accelerator (LINAC)-based STAR workflow and perform a failure modes and effects analysis (FMEA) with fault tree analysis (FTA), supplemented by an exploratory segment-based anatomical proximity analysis. A multidisciplinary team constructed a process map for C-arm LINAC-based STAR and identified potential failure modes across the workflow. Risks were scored using occurrence, severity, and detectability to calculate risk priority numbers (RPNs), and FTA was used to analyze causal pathways. In a predefined exploratory imaging subgroup of eight patients with ventricular tachycardia and simulation-acquired coronary computed tomography angiography (CCTA), we divided the left ventricle according to the 17-segment model and measured the minimum distances from each segment to adjacent OARs as an anatomical surrogate of exposure likelihood. Seventy-nine failure modes were identified, of which 17 were classified as high risk. The highest-risk failure modes were concentrated in target delineation and motion-related steps, including inaccurate multimodal image registration (RPN 432), improper motion evaluation/management (RPN 336), and diagnostic error of arrhythmia substrate definition (RPN 320). In the exploratory imaging subgroup (n = 8), segmental spatial patterns were moderately consistent across patients, with most standard deviations of minimum distance below 2.0 cm, although variability remained for several segment-OAR relationships. Segments 4, 5, and 10 were each located within 2 cm of both the stomach and esophagus, indicating relatively higher anatomical proximity-related risk. In this FMEA of C-arm LINAC-based STAR, the principal high-risk workflow steps were concentrated in substrate definition, image registration, and motion management. The segment-based analysis provides an exploratory anatomical risk-mapping framework for OAR awareness. Further multicenter studies are needed to refine workflow risk prioritization and evaluate the clinical relevance of the segment-based anatomical findings. Not applicable.

Respiratory motion tracking is critical for optimizing thoracoabdominal radiotherapy accuracy but remains constrained by the system latency of medical linear accelerators. Neural signals that precede the emergence of respiratory motion have the potential to mitigate this system latency issue in respiratory motion tracking radiotherapy. However, the real-time decoding of respiratory-related neural signals is challenging, creating translational bottlenecks that surpass the technical barriers encountered in conventional imaging-based tracking systems. This prospective review aims to provide an overview of the technical challenges and potential solutions for translating neural signals-based respiratory motion tracking into clinical practice.

Propagation-based methods have drawn increasing research attention in interactive medical image segmentation. However, existing propagation-based methods face two significant challenges: 1) Due to the continuous nature of anatomical structures within the organs and tumors throughout the volume, over-propagation is likely to occur as the propagation process reaches the end of structures, leading to a degradation in segmentation performance. 2) During the multi-round refinement process, selecting the worst-segmented slice for refinement tends to hinder the optimization of segmentation results. To overcome these challenges, we propose the Discrepancy Aware Network (DANet), which includes a Discrepancy Learning Module (DLM) and employs a confidence loss to achieve accurate segmentation. Specifically, DLM captures the temporal-contextual discrepancy between previous and current slices, enabling the model to perceive the variations of the target. Furthermore, the confidence loss is responsible for regularizing the over-confident segmentation at the image level by estimating the target foreground. Additionally, we design a straightforward slice selection strategy to optimize the refinement process. Extensive experimental results on five public medical datasets demonstrate significant improvements over state-of-the-art methods (e.g., with +1.07% improvement on the MSD-Spleen dataset).

BACKGROUND:Failure mode and effects analysis (FMEA) is a widely used proactive tool for risk assessment in radiotherapy. With the introduction of 3D-printed template-assisted intracavitary/interstitial brachytherapy (3DP-IC/IS), the increasing procedural complexity and newly introduced steps highlight the need for systematic risk management. However, the traditional FMEA approach has been criticized for its mathematical and logical limitations. PURPOSE:To develop an enhanced FMEA framework using a cloud model and data envelopment analysis (DEA), and to validate its performance through a comparative risk assessment against the traditional FMEA method in 3DP-IC/IS. METHODS:The analysis was performed on a dataset from 80 patients who underwent 3DP-IC/IS. The proposed framework incorporates a cloud model to manage the fuzziness and randomness of linguistic risk assessments and a modified DEA model for multi-criteria decision-making. Both the proposed and traditional FMEA methods were used to rank failure modes (FMs) in the 3DP-IC/IS workflow. Risk priority numbers and model-derived efficiency scores were calculated to identify high-priority FMs, and the ranking results from both methods were compared to evaluate performance. RESULTS:A total of 66 FMs were analyzed. The average rank shift between the traditional and introduced FMEA was 4.39 (max: 17, min: 0). Among the top 20 FMs, 18 were consistent between the two methods. "The introduced FMEA uniquely identified "Insufficient checking of needle labels" and "Excessive number of needles" as high risk." The proposed framework also increased risk-ranking discrimination, reducing the number of FMs with identical scores from 12 groups to a single group. The two highest risk FMs-"Unreasonable needle track design" and "Error in identifying non-coplanar needle number"-were consistently ranked first by both methods. Notably, the proposed framework reclassified four FMs as high risk, which demonstrated improved sensitivity to risks potentially underestimated by traditional FMEA. CONCLUSION:We developed a hybrid cloud model-DEA-FMEA framework that mitigates key limitations of traditional risk assessment. When applied to 3DP-IC/IS, this framework demonstrated both feasibility and an enhanced ability to identify and prioritize critical FMs. Our findings highlight its clinical value for improving quality assurance and patient safety, especially in complex, high risk radiotherapy procedures.

Magnetic resonance imaging (MRI)-guided radiotherapy (MRIgRT) integrates MRI with linear accelerators (MRI-linacs), enabling real-time motion management based on temporally resolved 2D MRI (cine-MRI). Current systems rely on template matching or deformable image registration for radiotherapy target (typically the gross tumor volume) localization, which allows beam gating. Further advances in localization could support more precise and efficient delivery methods. https://trackrad2025.grand-challenge.org/ was organized to provide a common dataset to benchmark algorithms for MRIgRT target tracking in 2D+t cine-MRI. Participants propagated target segmentation masks from an initialization frame across subsequent frames. The dataset comprised sagittal cine-MRI scans of 585 cancer patients undergoing radiotherapy at 0.35 T and 1.5 T MRI-linacs at six different institutions, with expert-annotated targets in 108 sequences. Target sites included the thorax (179 cases), abdomen (266 cases), and pelvis (140 cases). A total of 477 unlabeled and 50 labeled cases were provided for training purposes, 58 cases were kept private for preliminary testing (8) and final evaluation (50). The algorithms submitted by participants were executed on the challenge platform and assessed using metrics in three categories: geometric accuracy, surrogate dose accuracy and execution speed. Rankings were derived via a Rank-Then-Mean scheme. TrackRAD2025 attracted 148 registrations from 28 countries, 100 preliminary submissions and 24 final submissions from 14 teams. The top five methods achieved mean Dice similarity coefficients >0.87 and Euclidean center distances <2.1 mm, comparable to interobserver variability. Leading top five solutions featured foundation models with (4) or without (1) finetuning. Field strength had minimal effect on performance and tracking worked better for the pelvis with reduced motion amplitude compared to the thorax and abdomen cases, which achieved equivalent performance. TrackRAD2025 established a benchmark for MRIgRT tracking on multi-institutional cine-MRI data, highlighting foundation models as promising for clinical translation.

In the TrackRAD2025 challenge, we propose a three-stage semi-supervised fine-tuning method for real-time MRI tumor segmentation based on Segment Anything Model 2 (SAM2). Our approach employs the sam2_hiera_tiny variant as the backbone network, which is applied uniformly across all anatomical locations and field strengths. The core of our methodology is a progressive learning strategy: (1) we first perform supervised full-parameter fine-tuning on 38 labeled cases; (2) the trained model is then used to generate high-quality pseudo-labels for 100 unlabeled cases; (3) finally, we conduct a second round of fine-tuning on the combined set of labeled and pseudo-labeled data to enhance generalization. For data preprocessing, .mha files were converted to JPGs, normalized with min–max scaling, and cropped to an input size of 512× 512 . Standard geometric and color augmentations are also applied. The model is optimized using AdamW with a cosine decay learning rate schedule, initialized at 2.0 × 10^-7 . The loss function was a weighted combination of mask, Dice, class, and IoU losses. The final model submitted was selected based on its superior performance on a comprehensive set of metrics on a local validation set of 12 cases. On the official test set, our method achieved a DSC of 0.8860 (second place) and a relative D98 dose of 0.963 (first place), achieving second place in the challenge. The code is available at https://github.com/togetherhkl/TrackRAD2025 .

Artificial intelligence (AI) has been widely applied in healthcare; however, the potential risks it introduces still require attention. At present, safety risk analysis in healthcare largely relies on the traditional failure mode and effect analysis (FMEA), which has certain limitations in risk evaluation and prioritization. To address these gaps, we developed an enhanced FMEA (eFMEA) framework tailored to such workflows. Expert ratings for three risk criteria were converted into fuzzy values, and a consensus-reaching process (CRP) was employed to measure and improve consensus levels across failure modes(FMs). Expert weights were assigned using the Ordered Weighted Averaging (OWA) operator based on consensus contributions, while criteria weights were derived via the Entropy Weighting Method (EWM). Risk scores were computed using the Gained and Lost Dominance Score (GLDS). Among 86 FMs analyzed, 27 initially showed low consensus; only two remained low after consensus adjustment. Severity received the highest weight, aligning with clinical expectations. Comparative analysis with traditional FMEA and another multi-criteria decision-making model further demonstrated the robustness of the eFMEA method. And eFMEA resolved tied ranks and revealed distinct prioritization differences for several FMs. By integrating fuzzy theory, EWM, CRP, and GLDS, the proposed eFMEA framework overcomes key methodological shortcomings of traditional FMEA and offers a more robust and clinically applicable risk assessment tool.

Inverse treatment planning is pivotal in tumor treatment planning. It enables the multi-objective optimization of radiation dose delivery, ensuring precise tumor targeting while sparing surrounding healthy tissues. This process often requires frequent parameter adjustments to achieve the desired balance between objectives, making it both labor-intensive and time-consuming. Deep reinforcement learning (DRL) provides an automated, model-based planning solution, aimed at reducing reliance on human expertise and enhancing the efficiency of objective parameter optimization. However, most current approaches apply DRL to inverse planning without fully leveraging the knowledge embedded in the continuous state-action space, defined by the coupling between nonstationary planning states and continuous decision variables. This may result in insufficient exploration and exploitation, leading to inefficient optimization. This work introduces an innovative action-guided DRL (AgDRL) approach for automatic inverse planning. Our goal is to enhance exploration and exploitation by leveraging insightful guidance from reward-guided actions. The implementation of AgDRL incorporates both exploitation and exploration in the action-state space. For exploitation, high-reward actions are employed as guidance to achieve the optimal action adjustment. For exploration, low-reward actions are recommended as training resets to explore a broader range of the latent state space. Quantitative and qualitative experiments are conducted in various settings to evaluate the proposed method. The results are assessed using DRL-related metrics (e.g. reward gains) and clinical-related measurements (e.g. dose-volume histograms, DVHs). Experimental results on a real-world rectal cancer dataset empirically demonstrate that the proposed AgDRL-based approach significantly improves optimization efficiency through a high-reward strategy while enhancing exploration diversity via a low-reward strategy, consistently outperforming the MatRad treatment planning optimization platform.

Objectives.Stereotactic arrhythmia radioablation (STAR) offers a non-invasive treatment option for refractory tachycardia; however, precise dose delivery remains challenging due to the complexity of cardiorespiratory motion. This study evaluated the geometric and dosimetric performance of multiple deformable image registration (DIR) algorithms using ECG-gated four-dimensional CT (ECG-4DCT) in both virtual phantom and clinical datasets.Approach. ECG-4DCT data from the extended cardiac-torso phantom and 20 patients were analyzed across 10 cardiac phases using six DIR algorithms, with a seventh algorithm, TransMorph, additionally evaluated on the clinical datasets. Registration accuracy was assessed using the Dice similarity coefficient (DSC), Hausdorff distance (HD95), and average surface distance, while dosimetric accuracy was evaluated using dose-volume histogram metrics andγanalysis. 4D dynamic dose (4DDD) variability was quantified using the coefficient of variation (CV) and maximum pairwise absolute dose difference (MPADD).Main Results. Registration accuracy was lowest between the end-systolic and end-diastolic phases in both phantom and clinical datasets. In the phantom study, MIM achieved the highestγpassing rate (89.6% at 1%/1 mm) and the slightest deviation from the reference, with differences of -0.02 Gy inD95and -0.2% inV25. In the clinical datasets, patients without metallic implants exhibited reduced geometric accuracy and increased 4DDD variability (mean CVs of 0.01 forV25and 0.02 forD95; MPADDs up to 11.3% and 1.15 Gy). TransMorph achieved the highest geometric accuracy, with mean DSC values of 0.86 in patients without implants and 0.89 in those with implants; however, this improved geometry was accompanied by steeper deformation gradients and pronounced localized dose discrepancies.Significance. Current DIR algorithms remain limited in capturing complex cardiac motion for STAR. Geometric accuracy alone is insufficient to ensure physiologically plausible deformation, underscoring the need for cardiac-specific, physiology-constrained DIR frameworks to enable robust and clinically reliable 4DDD evaluation.

Class incremental medical image segmentation (CIMIS) aims to preserve knowledge of previously learned classes while learning new ones without relying on old-class labels. However, existing methods 1) either adopt one-size-fits-all strategies that treat all spatial regions and feature channels equally, which may hinder the preservation of accurate old knowledge, 2) or focus solely on aligning local prototypes with global ones for old classes while overlooking their local representations in new data, leading to knowledge degradation. To mitigate the above issues, we propose Prototype-Guided Calibration Distillation (PGCD) and Dual-Aligned Prototype Distillation (DAPD) for CIMIS in this paper. Specifically, PGCD exploits prototype-to-feature similarity to calibrate class-specific distillation intensity in different spatial regions, effectively reinforcing reliable old knowledge and suppressing misleading information from old classes. Complementarily, DAPD aligns the local prototypes of old classes extracted from the current model with both global prototypes and local prototypes, further enhancing segmentation performance on old categories. Comprehensive evaluations on two widely used multi-organ segmentation benchmarks demonstrate that our method outperforms state-of-the-art methods, highlighting its robustness and generalization capabilities.

Objective.To develop and validate UNetrDose, a Transformer-based deep learning model designed for fast and accurate photon beamlet dose prediction. The study aims to achieve Monte Carlo (MC)-level dosimetric accuracy using only CT-derived electron density images and beamlet coordinates as input, enabling the efficient reconstruction of complete 3D dose distributions for intensity-modulated radiation therapy (IMRT) plans.Approach.For each beamlet, a fixed-size 3D image patch was extracted along its propagation path, centered on the beamlet trajectory. The geometric information, defined as the beamlet's relative position within the beam field, was used alongside the image patch as model input. The ground-truth dose distributions were generated using MC simulations. The proposed UNetrDose model combined convolutional layers for local feature extraction with Transformer modules to capture long-range dependencies. A total of 90 fixed-beam IMRT plans (51 esophageal and 39 rectal cases) were used for model training, validation, and testing. Model performance was comprehensively assessed, evaluating spatial accuracy with 3D gamma pass rates and clinical acceptability through dose-volume histogram comparisons and other dosimetric parameters.Main results.UNetrDose demonstrated high fundamental accuracy at the individual beamlet level, where pass rates for the stringent(1 mm, 1%) criterion exceeded 96% for both esophageal and rectal cases. This translated to excellent clinical performance on full IMRT plans, where the model achieved mean(2 mm, 2%) pass rates ranging fromfor esophageal cases tofor rectal cases. The model was also highly efficient, with an average inference time of approximately 28 ms per beamlet.Significance.UNetrDose offers a promising alternative to traditional dose calculation engines by providing a balance between high dosimetric accuracy and fast computation. Its ability to predict dose distributions using only electron density images and beamlet positions simplifies the workflow, making it highly applicable for time-sensitive clinical scenarios.

Purpose To accurately predicting and classifying the magnitude of tumor movement in lung cancer patients, a machine learning model based on clinical information and CT-based radiomics feature with artificial neural network (ANN) was developed. Methods and materials CT images of lung tumor of 110 patients were enrolled, comprising 165 tumor sites. Manually extracted 18 clinical features, while 1218 radiomics features were calculated based on CT images. Superior-inferior (SI) direction tumor amplitude values were computed from in-house 4DCT. The three models for predicting tumor motion and classifying active or no motion management strategy based on ANN are as follows: i) C-model which based on clinical and image features; ii) R-model which based on radiomic features; iii) hybrid model that combines two previous models. were utilized for assessing the predictive accuracy of three models. For assessing classification performance, the area under the curve (AUC), specificity and sensitivity were calculated. Results In the test dataset, the C-model showed an MAE of 2.95 mm and an R2 value of 0.51, while the R-model demonstrated an MAE of 1.34 mm and an R2 value of 0.88. The hybrid model exhibited the best performance with an MAE of 1.23 mm and an R2 value of 0.91. Sensitivity and AUC were 0.79 and 0.95, 1.00 and 0.99, 1.00 and 1.00 for the C-model, R-model and hybrid model. Conclusion Utilizing machine learning techniques based on clinical features combined with CT image information enables accurate prediction of lung cancer tumor motion range as well as effective classification of motion management strategies.

Purpose This study aims to develop novel dose-transfer-based calibration methods (central field [CF] and daisy chain [DC]) to overcome the limitations of conventional wide-field (WF) calibration for IC PROFILER (TM) ionization chamber arrays in MR-linac environments, specifically resolving dose profile anomalies and enhancing measurement consistency with water tank reference data. Methods The central field (CF) and daisy chain (DC) calibration methods were evaluated with water tank measurements as a reference and compared to the wide-field (WF) method from Sun Nuclear Corporation (SNC). Profiles were measured at depths of 0.9, 2.9, and 4.9 cm for field sizes ranging from 10 x 10 cm(2) to 40 x 22 cm(2). Accuracy was evaluated using flattening filter-free (FFF) beam characterization and gamma passing rate (GPR) with a 1 %/1 mm criterion. Statistical analyses included the Kruskal-Wallis and Mann-Whitney U tests. Effect size of GPR calculated using Cohen's d. Results The CF and DC methods showed superior agreement with water tank profiles compared to the SNC method. In larger fields (40 x 22 cm(2) and 30 x 22 cm(2)), the CF method achieved mean GPRs above 95.9 % and median values over 99.0 %, outperforming the SNC method (mean and median GPR: 89.2 %; all p-values <0.05 and d-values >0.9). The DC method also performed well, achieving higher GPRs in the 40 x 22 cm(2) field and consistent accuracy across smaller fields. Additionally, beam characterization metrics-including peak position, symmetry, penumbra, unflatness, and slope-exhibited smaller deviations from water tank measurements when using CF and DC methods, as reflected in more concentrated difference distributions. Conclusion The SNC method is insufficient for MR-linac due to magnetic field-induced variations in detector response, leading to significant discrepancies in dose profiles. In contrast, the CF and DC methods demonstrate superior accuracy, improving the reliability of QA workflows. By minimizing the risk of beam miscalibration, these methods ensure the safety and effectiveness of MR-linac-based radiotherapy.

Radiation therapy is regarded as the mainstay treatment for cancer in clinic. Kilovoltage cone-beam CT (CBCT) images have been acquired for most treatment sites as the clinical routine for image-guided radiation therapy (IGRT). However, repeated CBCT scanning brings extra irradiation dose to the patients and decreases clinical efficiency. Sparse CBCT scanning is a possible solution to the problems mentioned above but at the cost of inferior image quality. To decrease the extra dose while maintaining the CBCT quality, deep learning (DL) methods are widely adopted. In this study, planning CT was used as prior information, and the corresponding strictly structure-preserved CBCT was simulated based on the attenuation information from the planning CT. We developed a hyper-resolution ultra-sparse-view CBCT reconstruction model, known as the planning CT-based strictly-structure-preserved neural network (PSSP-NET), using a generative adversarial network (GAN). This model utilized clinical CBCT projections with extremely low sampling rates for the rapid reconstruction of high-quality CBCT images, and its clinical performance was evaluated in head-and-neck cancer patients. Our experiments demonstrated enhanced performance and improved reconstruction speed.

The impact of myosteatosis on clinical outcomes following liver transplantation (LT) remains unclear. Articles evaluating the relationship between myosteatosis and the clinical outcomes of LT recipients were comprehensively retrieved from the Embase, PubMed, and Cochrane Library Central databases up to 1 October 2024. Thirteen articles involving 3351 cases were included. Myosteatosis was related to increased mortality risk in patients undergoing LT (HR, 1.764; 95

The integration of magnetic resonance imaging with linear accelerators (Linacs) enhances adaptive radiotherapy by providing real-time imaging for improved treatment precision. However, the long-term performance of MR-Linac systems, particularly in clinical settings, remains insufficiently studied. Traditional quality assurance (QA) methods, relying on binary pass/fail criteria, may overlook critical system variations. This study applies statistical process control (SPC) techniques to evaluate the long-term performance of a 1.5T MR-Linac, focusing on optimization in beam quality, MR-to-MV alignment, MR imaging, and geometric distortion. A dual-phase SPC framework was applied to 1 year of daily and weekly QA data from an Elekta Unity MR-Linac. Phase I established performance benchmarks, while Phase II monitored deviations online. Evaluated parameters included beam output, symmetry, MR-to-MV alignment, signal-to-noise ratio (SNR), spatial linearity, slice profile, and geometric distortion across spherical volumes (DSVs). Stability and variability were quantified using control charts and process performance indices (Ppk). Beam quality was stable overall (Ppk ≥ 1.33), though output dose and transverse symmetry showed increased variability in Phase II, with dose Ppk declining from 3.13 to 1.33. MR-to-MV alignment was consistent, but Phi rotational and Z translational offsets showed variability after system upgrades. Imaging metrics, including SNR and spatial linearity, achieved A + performance (Ppk ≥ 1.67) in Phase II, while vertical spatial resolution was lower (Ppk 1.04–1.10). Geometric distortion was well-controlled, though larger DSVs (≥ 500 mm) showed increased AP-axis distortion (2.44 mm) compared to RL (1.37 mm) and FH (0.93 mm). SPC techniques dynamically identified stable parameters and areas for improvement. Key recommendations include enhanced alignment protocols for beam quality and MR-to-MV offsets, as well as targeted strategies to address geometric distortion in larger volumes and along the AP axis.

BackgroundNon-alcoholic steatohepatitis (NASH)-associated hepatocellular carcinoma (HCC) has been emerging a predominant reason for liver transplantation (LT). The complexity of comorbidities in this population increases the possibility of poor transplant outcomes. The purpose of this study was to evaluate the differences in survival after transplantation among patients with NASH HCC and those with non-NASH HCC.MethodWe conducted systematic searches of the PubMed, Embase, Web of Science, and Cochrane Library databases. To analyze the data, both fixed and random-effects models were employed to aggregate hazard ratios (HRs) along with 95% confidence intervals (CIs) for recurrence-free survival (RFS) and overall survival (OS) outcomes. This study is registered with PROSPERO as CRD42024578441.ResultsA total of seven studies were included in this study. This study revealed that there was no significant difference in OS between liver transplant recipients with NASH HCC and those with non-NASH HCC. The RFS of NASH HCC patients were significantly longer. The HRs were 0.70 (95% CI: 0.51-0.97, P = 0.03) for RFS and 0.88 (95% CI: 0.72-1.07, P = 0.21) for OS, respectively.ConclusionThis study indicates that patients with NASH HCC who undergo LT have comparable OS as those with non-NASH HCC, while NASH HCC was associated with increased RFS. However, further research in randomized trials is necessary to verify these results and address potential selection biases.

Background:The long acquisition time of three-dimensional (3D) magnetic resonance imaging (MRI) makes it vulnerable to motion-induced artifacts such as blurring and ghosting, which may compromise target delineation and dose accuracy. Although motion management strategies such as four-dimensional MRI and cine MRI have been proposed, the specific influence of respiratory parameters-particularly amplitude and frequency-on the geometric and dosimetric precision of magnetic resonance-guided adaptive radiotherapy (RT) remains inadequately quantified. This study thus aimed to systematically evaluate how respiratory-induced linear translational motion affects delineation accuracy and dose distribution in MRI-based adaptive RT. Methods:An MR-compatible motion phantom was employed to replicate patient-specific respiratory-induced translational motion, with amplitude and frequency variations extracted from real patient waveforms being incorporated. Eight distinct respiratory patterns were generated, and MR images were acquired with standard spin-spin relaxation time-weighted (T2W) three-dimensional (3D) Cartesian sequences for each pattern. The internal target volume (ITV) delineated by clinicians based on 3D MR images was compared with the reference ITV (ITVref) generated with the digital phantom. Key delineation metrics [e.g., Dice similarity coefficient (DSC), Hausdorff distance (HD), and mean surface distance (MSD)] and dosimetric parameters (e.g., dose received by 95% of the volume) were evaluated, and statistical analyses were performed to assess the correlations between respiratory motion characteristics and the observed variations. Results:Respiratory amplitude significantly affected delineation accuracy and dosimetric consistency. The DSC decreased linearly with increasing amplitude, from 0.96 at 2.50 mm to 0.83 at 12.50 mm, while the HD and MSD increased proportionally (2.62 to 6.32 mm and 0.08 to 0.64 mm, respectively). Dosimetric analysis showed a notable reduction in ITVref dose coverage at higher amplitudes, with the dose received by 95% of the volume decreasing by 481.55 cGy at 12.50 mm relative to the prescribed total dose of 4,500 cGy. In contrast, respiratory frequency had minimal impact, with changes remaining within clinically acceptable ranges. Conclusions:This study investigated the impact of respiratory-induced linear translational motion on ITV delineation and dosimetric accuracy in MR-guided RT. It was found that large respiratory amplitudes significantly compromised geometric and dosimetric precision, whereas frequency had minimal influence. The results emphasize the limitations of 3D Cartesian MRI due to motion-averaged artifacts and support the development of advanced imaging techniques for improving ITV delineation accuracy.

Cone-beam computed tomography (CBCT) is a critical imaging modality in various medical fields, yet its repeated use poses radiation risks to patients. Low-dose CBCT image reconstruction aims to mitigate these risks while preserving image quality, which is crucial for clinical diagnosis and treatment. This review paper provides an in-depth analysis of the latest research progress in low-dose CBCT image reconstruction. We explore analytical reconstruction algorithms, iterative reconstruction algorithms, and deep learning approaches, each with distinct characteristics and applications. The paper comprehensively reviews the methods used for dose reduction in CBCT, the evolution of reconstruction algorithms, and their performance evaluations. We also identify challenges and limitations in current techniques, discussing potential future directions for low-dose CBCT reconstruction. Through a systematic literature search and analysis, this review offers a valuable reference for researchers and clinicians alike, aiming to advance the field of CBCT and enhance patient care through reduced radiation exposure and improved imaging outcomes.