BACKGROUND:Patient-specific quality assurance (PSQA) plays a pivotal role in intensity-modulated radiotherapy (IMRT) to ensure accurate dose delivery. However, conventional measurement-based PSQA approaches are labor-intensive and provide limited insight into the underlying factors contributing to variations in gamma passing rates (GPRs). Anatomical characteristics of the planning target volume (PTV) and organs at risk (OARs) may contain predictive information relevant to GPR performance, yet their potential has not been fully explored within interpretable machine learning frameworks. PURPOSE:This study aimed to develop an interpretable machine learning (ML) framework for predicting GPRs in IMRT based on anatomical features extracted from the PTV and OARs. METHODS:A retrospective cohort of 243 clinical chest IMRT plans was analyzed. Radiomic and dosimetric features were extracted for each anatomical structure. Two ML regression models-Random Forest (RF) and eXtreme Gradient Boosting (XGBoost)-were developed to predict GPRs for the PTV and OARs under four gamma criteria (3%/3 mm, 3%/2 mm, 2%/3 mm, and 2%/2 mm). The GPR obtained by comparing the dose distribution reconstructed using the independent Monte Carlo (MC) dose calculation software ArcherQA (Wisdom Technology Company Limited, Hefei, China)-based on linear accelerator delivery log files-with the original planned dose distribution was used as the reference standard, and calculated using global gamma analysis with a 10% dose threshold. Model performance was evaluated using the mean absolute error (MAE), root mean square error (RMSE), and Spearman's rank correlation coefficient. Shapley Additive Explanations (SHAP) were applied to interpret feature contributions in the best-performing model. RESULTS:Both models demonstrated robust predictive performance across different anatomical structures and gamma criteria. As the gamma criteria became less stringent, prediction errors decreased accordingly. Prediction accuracy was relatively high for OARs; for example, under the 3%/3 mm criterion, the test-set MAE was 0.06% ± 0.01% for the heart and 0.26% ± 0.04% for the whole lung. In contrast, the prediction error was relatively larger for the PTV, with a test-set MAE of 1.98% ± 0.31% under the same criterion. SHAP analysis revealed that texture-related radiomic features contributed most substantially to model predictions. Moreover, feature importance patterns varied according to organ type and gamma-criterion stringency. CONCLUSIONS:Multi-omics descriptors derived from anatomical structures can reliably predict GPRs in IMRT. The proposed interpretable ML framework not only achieves accurate prediction but also enhances mechanistic understanding through SHAP-based explanations. These findings provide valuable insights into dose verification variability and offer a practical, transparent tool for IMRT patient-specific quality assurance.

PURPOSE:Distant metastasis remains a major cause of treatment failure in head and neck (HN) cancer, highlighting the need for more accurate early risk stratification. This study developed and validated a deep radiomics framework to characterize tumor heterogeneity from pre-treatment CT images and improve prediction of distant metastasis-free survival (DMFS). MATERIALS AND METHODS:This multicenter study included 3,421 HN cancer patients from four cohorts across twelve institutions (RADCURE, HN1, HN-PET-CT and TCGA-HNSC). Radiomics features were extracted from primary tumors and transformed into OmicsMaps, a structured representation that spatially organizes inter-feature relationships to facilitate learning of complex prognostic patterns. A convolutional neural network was trained to derive prognostic signatures, which were integrated with key clinical variables to construct an OmicsMap-clinical fusion model for patient risk stratification. Model performance was assessed using the concordance index (C-index) and time-dependent area under the receiver operating characteristic curve (AUC) in the RADCURE, HN1, and HN-PET-CT cohorts. Radiogenomic analyses using RNA-seq data were conducted in the TCGA-HNSC cohort to investigate biological characteristics associated with the imaging-defined risk groups. RESULTS:The OmicsMap achieved C-index values of 0.742, 0.768, and 0.671 in the RADCURE, HN1, and HN-PET-CT cohorts, outperforming conventional radiomics approach by 5.84-6.37%. Incorporating clinical variables further improved generalizability, yielding C-index of 0.864 (HN1) and 0.730 (HN-PET-CT), with time-dependent AUCs of 0.727-0.895. The fusion model consistently stratified patients into distinct high- and low-risk groups for both DMFS and overall survival across cohorts (P < 0.01). Radiogenomic analyses revealed enrichment of immune-related pathways in the low-risk group, whereas the high-risk group exhibited a more aggressive phenotype enriched for proliferation, hypoxia, and epithelial-mesenchymal transition pathways, along with a fibrosis-prone tumor microenvironment characterized by extracellular matrix remodeling. CONCLUSION:Modeling inter-radiomic feature relationships using the OmicsMap representation substantially improves CT-based prediction of DMFS and characterization of tumor heterogeneity in HN cancer, supporting precision risk stratification in clinical oncology.

ABSTRACT With the rapid growth of image storage requirements, image deduplication has become a necessary technology to reduce storage costs by removing redundant images. Here, we propose Simage‐Dedup, a similar image deduplication scheme based on semantic segmentation. The scheme employs scalable semantic segmentation for fine‐grained deduplication and uses a semantic‐aware similarity detection method to analyze local semantic block features, enabling quick identification and alignment of similar images. Lossless compression strategies are applied to the aligned image blocks to mitigate the impact of local misalignment on compression efficiency. Finally, the images and their difference information are stored in the cloud to minimize overall storage requirements. To evaluate the performance of Simage‐Dedup, we developed a prototype system. It improves throughput by 1.8 to 3.4 times and compression ratio by 1.3 to 1.6 times, demonstrating its excellent performance.

[This corrects the article DOI: 10.1016/j.adro.2025.101891.].

Background:With the development of technology, stereotactic body radiotherapy (SBRT) has been increasingly applied in the treatment of patients with central lung cancer. Three-dimensional conformal radiotherapy (3DCRT) and intensity-modulated radiotherapy (IMRT) are common SBRT techniques for lung cancer. Although 3DCRT plans have low complexity, the quality of such plans is highly dependent on the experience of the planner. IMRT plans can provide better organ-at-risk (OAR) sparing, but they are associated with a high degree of modulation. Therefore, an SBRT planning strategy that combines the advantages of 3DCRT and IMRT has been developed. The purpose of this study was to investigate the differences in dosimetric parameters and delivery accuracy between CRT-IMRT-combined (Co-CRIM) and IMRT plans for central lung cancer patients treated with SBRT. Methods:Twenty patients were retrospectively included in this study. Four plans were designed for each patient. Co-CRIM-E and IMRT-E plans were designed based on Co-CRIM and IMRT, and delivered with Varian Edge. Co-CRIM-T and IMRT-T plans were designed based on Co-CRIM and IMRT, and delivered with Varian TrueBeam. Dosimetric parameters were compared between Co-CRIM-E and IMRT-E plans, as well as between Co-CRIM-T and IMRT-T plans. Monitor unit (MU), modulation factor (MF) and gamma (γ) passing rate were calculated for each plan. Paired-samples t-test was used to check the differences between datasets and P<0.05 was considered statistically significant. Results:Conformity index (CI), ratio of 50% prescription isodose volume to the planning target volume (PTV) volume (R50) and maximum dose (in % of dose prescribed) at 2 cm from PTV in any direction (D2cm) were significantly higher in Co-CRIM-E plans than those in IMRT-E plans. The CI, D2cm in Co-CRIM-T plans were also significantly higher than those in IMRT-T plans. Mean lung dose (MLD), percentage of the total lung volume receiving more than 10 Gy (V10), 12.5 Gy (V12.5), 13.5 Gy (V13.5) and 20 Gy (V20) were significantly higher in Co-CRIM-E plans than those in IMRT-E plans. MLD, V10, V12.5, V13.5 and V20 of total lung in Co-CRIM-T plans were also significantly higher than those in IMRT-T plans. MU and MF were significantly lower in Co-CRIM-E and Co-CRIM-T plans than those in IMRT-E and IMRT-T plans and γ passing rate was significantly higher than that in IMRT-E and IMRT-T plan. Conclusions:For central lung cancer patients treated with SBRT, the Co-CRIM approach was able to achieve plans that met clinical constraints. Compared with IMRT, Co-CRIM significantly reduced plan complexity and improved delivery accuracy with slightly reduced target conformability.

Objective.Patients with locally advanced non-small cell lung cancer (LA-NSCLC) exhibit heterogeneous prognoses despite receiving standard treatments, highlighting the need for more reliable prognostic biomarkers. This study aims to develop and validate OmicsMap model, a deep radiomics biomarkers derived from computed tomography images for the prediction of progression-free survival (PFS) in LA-NSCLC patients.Approach.We retrospectively analyzed data from 329 LA-NSCLC patients who underwent definitive radiotherapy. The cohort was randomly divided into development (N= 220) and independent testing set (N= 109). The prognostic signature was derived from integrated radiomics features extracted from both the primary tumor and involved lymph nodes, and inter-patient radiomics feature interactions. To achieve this, high-dimensional radiomics data from all patients were transformed into structured two-dimensional representations, termed OmicsMap, wherein radiomics feature interactions were encoded within the pixelated configuration. Deep radiomics features from the OmicsMaps were then extracted using a convolutional neural network for prognostic prediction. Model performance was evaluated by time-dependent area under the receiver operating characteristic curves area under the curve (AUC). Kaplan-Meier curves were plotted and hazard ratios (HR) were calculated via Cox proportional hazards model.Main results.The OmicsMap model achieved time-dependent AUCs of 0.76, 0.78 and 0.76 at 1, 2 and 3 years in the independent testing set, significantly outperforming the clinical model (AUC: 0.57, 0.57, 0.64;p< 0.05). The proposed model improved predictive discrimination with 7.69% increase in C-index over conventional radiomics approaches. It effectively stratified patients into high-risk and low-risk subgroups for both PFS (p< 0.001, HR = 0.380) and overall survival (p= 0.0021, HR = 0.525) in the testing set.Significance.The proposed OmicsMap model provides a novel paradigm for enhancing prognostic prediction in patients with LA-NSCLC. By improving risk stratification, the framework may help inform clinical decision-making and support future efforts toward more individualized management strategies.

Post-deduplication in traditional cloud environments primarily focuses on single-node, where delta compression is performed on the same deduplication node located on server side. However, with data explosion, the multi-node post-deduplication, also called global deduplication, has become a hot issue in research communities, which aims to simultaneously execute delta compression on data distributed across all nodes. Simply setting up single-node deduplication systems on multi-node environments would significantly affect storage utilization and incur secondary overhead from file migration. Nevertheless, existing global deduplication solutions suffer from lower data compression ratios and high computational overhead due to their resemblance detection's inherent limitations and overly coarse granularities. Similar blocks typically have high correlations between sub-blocks; inspired by this observation, we propose IBNR (Intra-Block Neighborhood Relationship-Based Resemblance Detection for High-Performance Multi-Node Post-Deduplication), which introduces a novel resemblance detection based on relationships between sub-blocks and determines the ownership of blocks in entry stage to achieve efficient global deduplication. Furthermore, the by-products of IBNR have shown powerful scalability by replacing internal resemblance detection scheme with existing solutions on practical workloads. Experimental results indicate that IBNR outperforms state-of-the-art solutions, achieving an average 1.99x data reduction ratio and varying degrees of improvement across other key metrics.

As the prevalence of cloud storage increases, many individuals and companies prefer outsourcing their data for backup and management. However, this has led to a significant increase in redundancy, decreasing storage utilization and wasting network bandwidth. While conventional resemblance detection methods remove redundancy among similar data by comparing the features extracted from each chunk's content. However, we observed that small changes between similar data chunks may cause false dissimilarity detection by conventional resemblance detection techniques. This is because features derived solely from the chunk content are highly susceptible to various modification patterns. Fortunately, we have discovered that two chunks are likely to be similar if their surrounding chunks are also similar, a concept we refer to as "chunk- context". Therefore, we propose a novel chunk-context aware resemblance detection method, called CARD, which includes a network-based chunk-context aware model and an N-sub-chunk shingles-based initial feature extraction strategy. By leveraging the Neural network, it can discover the complex patterns between the chunk- context and chunk content itself. A high-level understanding of the contextual information with chunk content can be synthesized into the representation of a chunk. The primary difference compared with others is that our design can significantly improves the accuracy or efficiency of resemblance detection by considering the chunk- context with chunk content itself. Furthermore, we implemented a CARD prototype and conducted extensive experiments using real workload, demonstrating that CARD can detect up to 75.03% more redundant data and accelerate the resemblance detection operations by 5.6 x to 86.7 x faster than state-of-the-art work.

Background:To improve patients' quality of life, it is necessary to preserve the high-functioning regions of the lung. This study aimed to propose a planning strategy that prioritizes sparing ventilation function for stereotactic body radiation therapy (SBRT) in patients with central lung cancer. Methods:Twenty patients with central lung cancer were retrospectively enrolled. Ventilation maps were obtained from 4-dimensional computed tomography by deformable image registration and quantitative analysis. The top 60% and 30-60% of the total lung were defined as high- and medium-ventilation regions, respectively. Clinical plans and functional lung avoidance (FLA) plans were designed for each patient, and FLA plans were optimized to spare high-ventilation (H-V) regions. Dosimetric parameters were compared in terms of tumor coverage, plan heterogeneity, and dose of organs-at-risk (OARs). A patient-specific dose-function response model of Owen was applied to characterize personalized functional damage. Results:FLA plans led to significant reductions in the mean dose of H-V lung than in clinical plans (361.33 vs. 398.51 cGy, P<0.001), but at the expenses of degraded gradient index (4.95 vs. 4.58, P<0.001) and conformity index (0.82 vs. 0.86, P<0.05). Generalized estimated equations demonstrated the decreased risk of grade 2+ radiation pneumonitis (RP) by 1.51% (P<0.01), and grade 3+ by 0.39% (3.81% vs. 3.42%, P<0.01) for FLA plans. We also observed the average ventilation preservation for FLA as 0.17% (P=0.002) higher than clinical plans in one year after SBRT. Conclusions:FLA plans can preserve the post-radiotherapy lung function, limit the pulmonary toxicity, and further improve life quality for SBRT patients with central lung cancer.

This study aims to construct and train the WingsNet model, which leverages the parameters recorded in log files to rapidly and accurately predict the patient-specific three-dimensional (3D) dose distribution for IMRT quality assurance (QA). We conducted a retrospective analysis of data from 286 lung cancer patients treated with a prescription of 60 Gy in 30 fractions, with 242 cases used for model training and 44 for testing. Log files containing information such as multi-leaf collimator (MLC) positions, monitor units (MU), and gantry angles were collected from Varian treatment accelerators. Pylinac software was employed to extract mechanical parameters from the log files, generating 2D fluence maps, which were then converted into 3D volumes using a ray-tracing algorithm. CT images, RT structures, and 3D volumes were resampled to a uniform dimension of 128*128*128 to serve as input for the WingsNet model, with the 3D dose distribution calculated by the treatment planning system (TPS) serving as output. The model training utilized L1 loss and mean squared error (MSE) as evaluation metrics. The results of this study demonstrate that the WingsNet model can effectively predict the 3D dose distribution of IMRT plans based on the parameters recorded in the log files. Evaluation through metrics such as mean absolute error (MAE), root mean square error (RMSE), and dose-volume histogram (DVH) indices reveals that the model performs well in most areas, with some errors observed in the planning target volume (PTV) region and at high dose levels, yet it retains potential for clinical use. Visually, the isodose line distributions are consistent. The Dice coefficients between the predicted and reference dose distributions at varying isodose line levels indicate a decreasing trend as the dose level increases. The WingsNet model developed in this study successfully predicts the patient-specific 3D dose distribution for QA by parsing the parameters recorded in log files. This model shows promise for use in 3D dose distribution verification for IMRT, providing an efficient and reliable tool for the verification of 3D dose distributions in patient-specific QA.

The primary aim of this paper is to investigate the asymptotic distribution of the zeros of certain classes of hypergeometric F-q+1(q) polynomials. We employ classical analytical techniques, including Watson'slemma and the method of steepest descent, to understandthe asymptotic behavior of these polynomials:F-q+1(q)(-n,kn+alpha,& mldr;,kn+alpha+(q-1)/(q);kn+beta,& mldr;,kn+beta+(q-1)/(q);z)(n ->infinity), where n is a nonnegative integer, q is a positive integerand the constant parameters alpha and beta are constrained by alpha<beta. By applying the general results established in this paper, we generate numerical evidence and graphical illustrations using Mathematica to show the clustering of zeros on certain curves.

The one-shot multiple object tracking (MOT) framework simultaneously generates detected targets and re-identification (ReID) embeddings, and employs them to associate previous tracks. However, the continuity between the learning and extraction within the ReID task is unconsciously neglected because of treating them as two isolated stages. This leads to unreliable ReID embeddings and poses a challenge to target matching especially in complex and crowded scenarios. To this end, we propose a guided embedding enhancement multiple object tracker, named G2EMOT. It consists of three innovative designs aiming at improving the embedding representation capability for individual targets. First, based on the original feature map, a global instance-specific context decoupling (GICD) module is devised to facilitate the respective feature learning for detection and ReID branches. Then, a heatmap-guided embedding enhancement (HGEE) module is introduced to connect the processes of embedding learning and extraction, ensuring that the detected target coordinates are accurately aligned with the discriminative ReID embeddings. Finally, to associate and stabilize matched targets, we present a novel two-gate guided embedding update (TG-GEU) strategy that dynamically updates ReID embeddings. With the proposed three components, G2EMOT achieves outstanding performance on popular MOT benchmarks while outperforming the existing one-shot tracking methods by a large margin. In particular, it realizes IDF1 of 76.1% on MOT17 and IDF1 of 74.2% on MOT20. The source codes are released at https://github.com/ydhcg-BoBo/G2EMOT.

Landmark detection is a common task that benefits downstream computer vision tasks. Current landmark detection algorithms often train a sophisticated image pose encoder by reconstructing the source image to identify landmarks. Although a well-trained encoder can effectively capture landmark information through image reconstruction, it overlooks the semantic relationships between landmarks. This contradicts the goal of achieving semantic representations in landmark detection tasks. To address these challenges, we introduce a novel Siamese comparative transformer-based network that strengthens the semantic connections among detected landmarks. Specifically, the connection between landmarks with the same semantics has been enhanced by employing a Siamese contrastive regularizer. In addition, we integrate a lightweight direction-guided Transformer into the image pose encoder to perceive global feature relationships, thereby improving the representation and encoding of landmarks. Experiments on the CelebA, AFLW, and Cat Heads benchmarks demonstrate that our proposed method achieves competitive performance compared to existing unsupervised methods and even supervised methods.

Multiple object tracking (MOT) in Unmanned Aerial Vehicle (UAV) videos is important for diverse applications in computer vision. Current MOT trackers rely on accurate object detection results and precise matching of target reidentification (ReID). These methods focus on optimizing target spatial attributes while overlooking temporal cues in modelling object relationships, especially for challenging tracking conditions such as object deformation and blurring, etc. To address the above-mentioned issues, we propose a novel Spatio-Temporal Cohesion Multiple Object Tracking framework (STCMOT), which utilizes historical embedding features to model the representation of ReID and detection features in a sequential order. Concretely, a temporal embedding boosting module is introduced to enhance the discriminability of individual embedding based on adjacent frame cooperation. While the trajectory embedding is then propagated by a temporal detection refinement module to mine salient target locations in the temporal field. Extensive experiments on the VisDrone2019 and UAVDT datasets demonstrate our STCMOT sets a new state-of-the-art performance in MOTA and IDF1 metrics. The source codes are released at https://github.com/ydhcg-BoBo/STCMOT.

Four-dimensional cone-beam computed tomography (4D CBCT) provides respiration-resolved images and can be used for image-guided radiation therapy. However, the ability to reveal respiratory motion comes at the cost of image artifacts. As raw projection data are sorted into multiple respiratory phases, there is a limited number of cone-beam projections available for image reconstruction. Consequently, the 4D CBCT images are covered by severe streak artifacts. Although several deep learning-based methods have been proposed to address this issue, most algorithms employ ordinary network models, neglecting the intrinsic structural prior within 4D CBCT images. In this paper, we first explore the origin and appearance of streak artifacts in 4D CBCT images.Specifically, we find that streak artifacts exhibit a periodic rotational motion along with the patient's respiration. This unique motion pattern inspires us to distinguish the artifacts from the desired anatomical structures in the spatiotemporal domain. Thereafter, we propose a spatiotemporal neural network named RSTAR-Net with separable and circular convolutions for Rotational Streak Artifact Reduction. The specially designed model effectively encodes dynamic image features, facilitating the recovery of 4D CBCT images. Moreover, RSTAR-Net is also lightweight and computationally efficient. Extensive experiments substantiate the effectiveness of our proposed method, and RSTAR-Net shows superior performance to comparison methods.

Purpose: This study investigated the role of Planning Target Volume (PTV) in determining the suitability of manual and automatic Stereotactic Body Radiation Therapy (SBRT) planning for lung cancer patients. Methods: We retrospectively created manual and automatic lung SBRT plans for ninety-eight patients using the Pinnacle 16.2 Treatment Planning System (TPS). The superior plan, whether manual or automatic, was selected for each patient through a combined index. Receiver Operating Characteristic (ROC) analysis was used to assess the predictive potential of the PTV volume in determining the superior plan and to establish a cutoff value. Patients were then stratified into two groups based on this value, and dosimetric variances were evaluated. Results: The ROC analysis highlighted the PTV volume's proficiency in predicting the superior choice between manual and automatic lung SBRT plans (Area Under Curve (AUC): 0.918, p = 0.005). The delineated cutoff value was set at 22.675 cc. For PTV volumes below this threshold, automatic plans surpassed manual plans in the Conformity Index (CI), Gradient Index (GI), and lung doses. Conversely, for PTV volumes exceeding 22.675 cc, manual plans exhibited improved results in the Heterogeneity Index (HI), GI, and dosimetric metric of heart and lung. Conclusion: The PTV volume is a significant determinant in guiding the optimal selection between manual and automatic lung SBRT plans with the Pinnacle TPS. Automatic plans are recommended for patients with PTV volumes below 22.675 cc, and manual plans for those above this threshold.

This study aimed to improve the image quality and CT Hounsfield unit accuracy of daily cone-beam computed tomography (CBCT) using registration generative adversarial networks (RegGAN) and apply synthetic CT (sCT) images to dose calculations in radiotherapy. The CBCT/planning CT images of 150 esophageal cancer patients undergoing radiotherapy were used for training (120 patients) and testing (30 patients). An unsupervised deep-learning method, the 2.5D RegGAN model with an adaptively trained registration network, was proposed, through which sCT images were generated. The quality of deep-learning-generated sCT images was quantitatively compared to the reference deformed CT (dCT) image using mean absolute error (MAE), root mean square error (RMSE) of Hounsfield units (HU), and peak signal-to-noise ratio (PSNR). The dose calculation accuracy was further evaluated for esophageal cancer radiotherapy plans, and the same plans were calculated on dCT, CBCT, and sCT images. The quality of sCT images produced by RegGAN was significantly improved compared to the original CBCT images. ReGAN achieved image quality in the testing patients with MAE sCT vs. CBCT: 43.7 ± 4.8 vs. 80.1 ± 9.1; RMSE sCT vs. CBCT: 67.2 ± 12.4 vs. 124.2 ± 21.8; and PSNR sCT vs. CBCT: 27.9 ± 5.6 vs. 21.3 ± 4.2. The sCT images generated by the RegGAN model showed superior accuracy on dose calculation, with higher gamma passing rates (93.3 ± 4.4, 90.4 ± 5.2, and 84.3 ± 6.6) compared to original CBCT images (89.6 ± 5.7, 85.7 ± 6.9, and 72.5 ± 12.5) under the criteria of 3 mm/3

In the dim-small target detection field, background suppression is a key technique for stably extracting the target. In order to effectively suppress the background to enhance the target, this paper presents a novel background modeling algorithm, which constructs base functions for each pixel based on the local region background and models the background of each pixel, named single pixel background modeling (SPB). In SPB, the low-rank blocks of the local backgrounds are first obtained to construct the background base functions of the center pixel. Then, the background of the center pixel is optimally estimated by the background bases. Experiments demonstrate that in the case of extremely low signal-to-noise ratio (SNR < 1.5 dB) and complex motion state of targets, SPB can stably and effectively separate the target from the strongly undulant sky background. The difference image obtained via SPB background modeling has the characters: the non-target residual could be white noise, and the target is significantly enhanced. Compared with the other typical five algorithms, SPB remarkably outperforms other algorithms to detect the target of a low signal-to-noise ratio.