Med Phys [JOURNAL]
Sun
200 papers
Proton therapy range uncertainty reduction using vendor-agnostic tissue characterization on a virtual photon-counting CT head scan.
Vrbaski S, Stanic G, Molinelli S … +4 more , Bhattarai M, Abadi E, Ciocca M, Samei E
Med Phys · 2026 May · PMID 42027118 · Full text
BACKGROUND: Photon-counting CT (PCCT) is the latest technology enabling imaging with reduced noise and inherent spectral separation, with the potential to directly calculate a more accurate tissue stopping power from spe...
BACKGROUND: Photon-counting CT (PCCT) is the latest technology enabling imaging with reduced noise and inherent spectral separation, with the potential to directly calculate a more accurate tissue stopping power from spectral data. This potential benefit is difficult to quantify in practice and is currently evaluated mainly in phantoms with simplified geometries that only approximate real patient anatomy. PURPOSE: In this work, we proposed virtual imaging simulators as an alternative approach to experimental validation of beam range uncertainty in complex patient geometry using a computational model of a human head and a CT system. In addition, we validate the accuracy of stopping power ratio (SPR) calculations on a model of a PCCT scanner using a conventional stoichiometric calibration approach and a prototype software TissueXplorer. METHODS: A validated CT simulator (DukeSim) was used to generate PCCT projections of a computational head phantom, which were reconstructed with an open-source toolbox (ASTRA). The dose of 2 Gy was delivered through protons in a single fraction to target two different cases of nasal and brain tumors. The ground-truth treatment plan was made directly on the computational phantom using clinical treatment planning software (RayStation). This plan was then recalculated on the corresponding CT images for which SPR values were estimated using both the conventional method and the prototype software TissueXplorer. The resulting dose distributions were subsequently compared against the ground-truth plan to quantify dose differences arising from SPR estimation. RESULTS: The mean percentage difference in estimating the SPR with TissueXplorer in all head tissues inside the scanned volume was 0.28%. SPRs obtained with this method showed smaller dose distribution differences from the ground truth plan than the conventional stoichiometric calibration method on the computational head phantom. CONCLUSIONS: Virtual imaging offers an alternative approach to validation of the SPR prediction from CT imaging, as well as its effect on the dose distribution and thus downstream clinical outcomes. According to this simulation study, software solutions that utilize spectral information hold promise for more accurate prediction of the SPR than the conventional stoichiometric approach.
Correction to "Lightweight deep training network for lymph nodes segmentation from head and neck CT images".
Med Phys · 2026 May · PMID 42015337 · Publisher ↗
Abstract loading — click title to view on PubMed.
Determining a Gamma Knife's isocenter using a starburst shot.
Corns R, Calistro-Allen K, Yang K … +3 more , Jung J, Cao Y, Belcher W
Med Phys · 2026 Apr · PMID 42007767 · Full text
BACKGROUND: The Gamma Knife is an important treatment unit for stereotactic radiosurgery. The isocenter is well understood from the collective behavior of all its 192 Co60 sources but is not known from the behavior of in...
BACKGROUND: The Gamma Knife is an important treatment unit for stereotactic radiosurgery. The isocenter is well understood from the collective behavior of all its 192 Co60 sources but is not known from the behavior of individual sources. PURPOSE: To introduce and validate the "starburst shot," a new quality assurance (QA) technique to better understand the Gamma Knife's isocenter by analyzing the convergence of all individual collimator beam axes. METHODS: We built cylindrical phantoms to acquire panoramic radiographic films for 4-, 8-, and 16-mm collimator sizes. We developed an algorithm to locate the centers of each entrance and corresponding exit exposure on the panoramic film, map their centers back to the cylinder, and connect the centers with a line segment in 3D. These lines represent the collimators' central axes, and we analyzed the resulting cluster of 192 lines using an optimization algorithm to locate the smallest sphere that simultaneously touches all lines. This sphere's center is an estimate of the isocenter, and the radial distances from the isocenter were scored by the probability of passing within 0.5 mm of the isocenter. RESULTS: This optimization method provided a robust determination of the isocenter. For the 4-mm collimators, was 1.00. For the 8-mm collimators, ranged from 0.81 to 0.93, and for the 16-mm collimators, it ranged from 0.74 to 0.92. The isocenters determined for the 4-, 8-, and 16-mm collimators were found to be within 0.2 mm of each other. In a double-exposure experiment, the starburst shot accurately found a 1-mm shift introduced between exposures. CONCLUSIONS: The starburst shot is a new, effective technique for detailed Gamma Knife isocenter QA, providing a comprehensive analysis of all 192 beams. It successfully verifies the isocenter's position and size with submillimeter accuracy. The starburst shot differs significantly from standard techniques, such as the pinprick tool test, and we propose a tiered system of acceptable values of 0.90 for 4 mm, 0.80 for 8 mm, and 0.70 for 16 mm, as a QA standard for clinical use.
Internal stimulation source near-field small target electrical impedance tomography methodology (ISEIT) in pulmonary interventional surgery.
Zhang W, Wang J, Zhao P … +3 more , He W, Jiang Q, Yang H
Med Phys · 2026 Apr · PMID 42003790 · Publisher ↗
BACKGROUND: Dynamic imaging of the surgical region during lung interventional procedures can provide clinicians with valuable real-time guidance, assisting them in making informed clinical decisions. However, conventiona...
BACKGROUND: Dynamic imaging of the surgical region during lung interventional procedures can provide clinicians with valuable real-time guidance, assisting them in making informed clinical decisions. However, conventional medical imaging methods have limitations, such as poor real-time capability (CT, MRI), radiation exposure risk (X-ray), limited imaging resolution (ultrasound), and restricted fields of view (endoscopy), making them unsuitable for effective intraoperative monitoring. PURPOSE: Evaluate the proposed Internal Stimulation Source Near-Field Small Target electrical impedance tomography (EIT) Methodology to verify its effectiveness. METHODS: Unlike conventional EIT methods relying on external stimulation, this approach employs the interventional probe tip as an embedded electrode to safely deliver excitation current directly into tissues, creating a local electric field. The position of the probe tip is at the center of the lesion or close to it. The potentials of the remaining electrodes are alternately measured to reconstruct images of lesions adjacent to the probe tip, thereby establishing an intraoperative imaging model for interventional procedures. The feasibility of this method was evaluated through sensitivity distribution analysis and image reconstruction simulations. An in vitro physical model platform was developed, and a 16+1 electrode internal EIT system was designed for experimental validation. RESULTS: The simulation results show that ISEIT can accurately determine the target position in both single-target and multi-target scenarios. When the size of the lung nodules is less than 15 mm, the imaging quality of the traditional EIT method deteriorates, with the appearance of artifacts and position errors. However, the ISEIT method is basically unaffected by the images and still has stable imaging results and accuracy even for small target sizes. In different lesion area sizes, the imaging quality indicators of the ISEIT method are all superior to those of the traditional EIT method. Its internal source excitation feature enables it to have better robustness and anti-interference ability, verifying its boundary potential distribution characteristics and frequency response characteristics, as well as its ability to distinguish changes in the real part and imaginary part. CONCLUSION: The results demonstrated the effectiveness of the internal stimulation-based EIT method. Compared with conventional EIT, this approach achieves superior imaging performance and image quality, providing surgeons with real-time functional monitoring capabilities during surgical interventions, thereby enhancing intraoperative imaging accuracy and supporting the advancement of intelligent surgical systems.
On the role of demagnetizing tensors in arbitrary orientations of general ellipsoids and implications for MRI safety assessment.
Pakarinen T
Med Phys · 2026 Apr · PMID 42003228 · Full text
BACKGROUND: Demagnetizing tensors describe the shape anisotropic contribution of magnetic susceptibility, that is, how an object's shape and orientation affect its internal magnetization. Ellipsoids possess a unique geom...
BACKGROUND: Demagnetizing tensors describe the shape anisotropic contribution of magnetic susceptibility, that is, how an object's shape and orientation affect its internal magnetization. Ellipsoids possess a unique geometric property by exhibiting homogeneous internal magnetization, enabling a purely geometrical characterization. A general description of the demagnetizing tensor under arbitrary rotations would broaden their applicability across various fields, most notably in magnetic resonance safety evaluation. PURPOSE: Demagnetizing tensor is a well-defined concept in the ellipsoids principal frame, though its transformation under three-dimensional reorientation is often overlooked, a justifiable omission for spheroidal solutions of the Poisson equation. However, this does not hold for general ellipsoids under arbitrary reorientation. This work is motivated by the concerns in magnetic resonance imaging safety and its practical evaluation in clinical environments, aiming to extend and simplify the process. METHODS: This work demonstrates the validity of directly rotating the orthogonal basis solutions, derived from Poisson's equation, and uses the procedure to evaluate a practical approximation, based on orthogonal area-projections. The approach is also applied to generalize force and torque calculations for ellipsoids under three-dimensional reorientation. RESULTS: The method shows an exact match with the numerical solution and agrees with the standard scalar expressions for translation force and torque. Additionally, a unique connection to the MRI magic angle is found as the point of convergence for prolate spheroid aspect ratios under rotation. The area-projection approximation was demonstrated to perform fairly across prolate and oblate spheroids. Similar approximations might extend to irregular shapes, but numerical approaches remain preferable due to the complexity of internal field distributions. CONCLUSION: The presented approach offers an alternative method for force and torque calculations, such as those provided by ASTM, also generalizing the conventional approach. The area-projection approximation via SVD offers a straightforward extension to the original method. Finally, the connection between demagnetizing tensor rotation and the magic angle provides a new perspective on the phenomenon.
An Interpretable Radiomics Model Based on Multi-DLCT Images for Differentiating Benign and Malignant Solid Solitary Pulmonary Nodules.
Lin Z, Huang P, Xiao S … +7 more , Fan B, Tan Y, Luo S, Wang Y, Yu X, Zhang D, Zuo M
Med Phys · 2026 Apr · PMID 42003192 · Publisher ↗
BACKGROUND: Early-stage lung cancer frequently presents as a solitary pulmonary nodule. Compared with subsolid pulmonary nodules, solid pulmonary nodules are associated with greater invasiveness, earlier metastasis, fast...
BACKGROUND: Early-stage lung cancer frequently presents as a solitary pulmonary nodule. Compared with subsolid pulmonary nodules, solid pulmonary nodules are associated with greater invasiveness, earlier metastasis, faster growth, and worse clinical outcomes. Therefore, early diagnosis is essential for improving survival and prognosis. PURPOSE: To develop an interpretable radiomics model using multi-parametric images derived from dual-layer CT (DLCT) for the non-invasive differentiation of benign and malignant solid solitary pulmonary nodules (SSPNs). METHODS: This retrospective study included 236 patients with pathologically confirmed SSPNs who underwent DLCT-enhanced scanning at two centers. Patients from one center were randomly divided into a training cohort (n = 111) and an internal test cohort (n = 48) at a 7:3 ratio, while patients from the other center served as an external test cohort (n = 77). Radiomic features were independently extracted from seven venous-phase image series, including conventional images (CI), iodine density (ID) maps, effective atomic number (Zeff) maps, electron density (ED) maps, virtual monochromatic images (VMI) at 40 and 100 keV, and virtual non-contrast (VNC) images, and were subsequently selected using the Mann-Whitney U test, Spearman correlation analysis, and LASSO. Radiomics models based on individual image series and combined models were constructed using logistic regression, SVM, and XGBoost. Model performance was evaluated using the area under the receiver operating characteristic curve (AUC), accuracy, sensitivity, and specificity. SHapley Additive Explanations (SHAP) were applied to interpret the fusion model. RESULTS: The SVM-based combined model demonstrated stable diagnostic performance across cohorts (AUC = 0.909, 0.852, and 0.793 for the training, internal test, and external test cohorts, respectively), achieving the highest AUC among all models in the external test cohort. In the external test cohort using the SVM algorithm, the combined model showed a higher AUC than the CI model (AUC = 0.793 vs. 0.667; P = 0.125), with positive NRI and IDI. CONCLUSIONS: An interpretable radiomics model based on multi-parametric DLCT images using the SVM algorithm enables accurate and robust noninvasive differentiation of benign and malignant SSPNs.
Correction to "Efficient and accurate commissioning and quality assurance of radiosurgery beam via prior-embedded implicit neural representation learning".
Med Phys · 2026 Apr · PMID 42003093 · Full text
Abstract loading — click title to view on PubMed.
Support vector machine-based MRI radiomics predict response to neoadjuvant therapy and progression-free-survival of Colorectal Liver Metastases patients.
Li ZF, Guo J, Sun C … +3 more , Tian S, Ma ZL, Ye ZX
Med Phys · 2026 Apr · PMID 41988745 · Publisher ↗
BACKGROUND: The efficacy of neoadjuvant therapy (NAT) and surgical prognosis stratification in patients with colorectal cancer liver metastasis (CRLM) remain unclear. PURPOSE: This study aims to develop and validate two...
BACKGROUND: The efficacy of neoadjuvant therapy (NAT) and surgical prognosis stratification in patients with colorectal cancer liver metastasis (CRLM) remain unclear. PURPOSE: This study aims to develop and validate two radiomics models using baseline and delta radiomics features to predict the response to NAT and prognosis in CRLM. METHODS: Baseline and Delta MRI radiomics features were extracted from 116 CRLM patients who underwent NAT combined with hepatectomy in the training set. Radiomics and clinical features were selected using Boruta algorithm to construct predictive models. In the external test set, the predictive performance of both models was evaluated using the area under the curve (AUC), decision curve analysis (DCA), calibration curves, and Brier score. Additionally, the delta model was compared with the clinical risk score (CRS). RESULTS: The two models contain 6 and 4 variables, respectively, for predicting treatment response and progression-free survival (PFS). Both the response model and delta model demonstrated comparable predictive performance in the external test set (AUC = 0.806 for both). Furthermore, the delta model demonstrated superior performance to CRS in predicting PFS (AUC: 0.806 vs. 0.623). CONCLUSIONS: Radiomics models based on baseline and delta MRI effectively predicted treatment response and postoperative PFS in CRLM patients, with the delta model showing superior accuracy for PFS prediction compared to CRS.
Masked pre-training and pseudo-vessel enhancement-based semi-supervised cerebral vascular segmentation.
Wang S, Wang H, Tang Y … +3 more , Lin J, Yu J, Wu G
Med Phys · 2026 Apr · PMID 41988737 · Publisher ↗
BACKGROUND: Accurate segmentation of cerebral vessels in computed tomography angiography (CTA) and computed tomography perfusion (CTP) images is essential for diagnosing cerebrovascular diseases. However, due to the scar...
BACKGROUND: Accurate segmentation of cerebral vessels in computed tomography angiography (CTA) and computed tomography perfusion (CTP) images is essential for diagnosing cerebrovascular diseases. However, due to the scarcity of annotated data and the complexity of small and branching vessels, fully supervised deep learning methods face limitations in clinical scenarios. PURPOSE: This study aims to develop a semi-supervised learning framework that can achieve high-accuracy cerebrovascular segmentation using only limited labeled data, while improving robustness to and enhancing small vessel continuity. METHODS: We propose a novel semi-supervised segmentation framework integrating three core components: (1) a dual-scale masked pre-training strategy, which jointly reconstructs masked vessel regions and performs segmentation to enhance vascular continuity; (2) pseudo-vessel augmentation, which simulates vessel-like artifacts to improve discrimination between true vessels and false positives; and (3) multi-level consistency losses, which enforce output consistency across different enhancement and model predictions to enhance learning under low supervision. The model was evaluated on 948 CTA and CTP scans using five annotation ratios (5%, 10%, 20%, 50%, and 100%). RESULTS: With only 5% of labeled data, the proposed method achieved 98.7% of the performance of a fully supervised model. Under full supervision, it reached Dice scores of 0.8379 on CTP and 0.9177 on CTA, outperforming the second-best method by 1.12% and 0.87%, respectively. CONCLUSIONS: The proposed semi-supervised framework enables accurate cerebrovascular segmentation with minimal labeled data. It effectively enhances small vessel continuity, reduces over-segmentation from noise, and demonstrates strong robustness and generalizability, making it a promising solution for real-world clinical applications.
Refining feature representation for accurate fundus lesion segmentation.
Sun H, Gao E, Li Y … +11 more , Xiang D, Lu F, Zhou Y, Zhu W, Shi F, Nie B, Chen X, Zhang L, Cai J, Peng D, Peng T
Med Phys · 2026 Apr · PMID 41988735 · Publisher ↗
BACKGROUND: Accurate segmentation of retinal lesions is essential for early detection of age-related macular degeneration (AMD), particularly its key indicators, choroidal neovascularization (CNV) and choroidal non-perfu...
BACKGROUND: Accurate segmentation of retinal lesions is essential for early detection of age-related macular degeneration (AMD), particularly its key indicators, choroidal neovascularization (CNV) and choroidal non-perfusion (CNP). However, the highly variable shapes and appearances of lesions pose substantial challenges for automated segmentation. PURPOSE: To tackle these challenges, we propose a deep-learning model, the Retinal Feature Enhancement Network (RFENet), designed to improve the accuracy and robustness of retinal lesion segmentation under challenging imaging conditions. METHODS: RFENet builds upon the UNeXt backbone and introduces two modules tailored for retinal lesion segmentation: an Adaptive Feature Refinement Unit (AFRU), which selectively emphasizes informative features, and an Optimized Channel-Wise Convolution Unit (OCCU), which captures fine structural details in irregular lesions. Two expert-annotated datasets were used: 1070 OCT B-scans from 83 CNV patients and 184 FFA images from 107 CNP patients. Images were split at the patient level into training (70%), validation (10%), and test (20%) sets. Performance was compared against state-of-the-art segmentation models, including UNet++, Swin-UNet, SelfReg-UNet, DAEFormer, and Mamba-UNet. Evaluation metrics included Dice coefficient and Intersection over Union (IoU). Statistical significance was assessed using paired two-tailed t-tests ( ), and effect sizes were reported with Cohen's . RESULTS: On the CNV dataset, RFENet achieved a Dice of 82.50% and an IoU of 71.66%, with clear improvements over competing models. On the CNP dataset, RFENet achieved a Dice of 74.43% and an IoU of 60.80%, slightly surpassing the best benchmark. Most pairwise comparisons were statistically significant, including CNV compared with Swin-UNet and SelfReg-UNet (p 0.001) and CNP compared with DAEFormer (p 0.035). Statistical analyses, including effect size evaluation, further confirmed the robustness of these improvements, with the most notable gains observed in the CNV dataset. CONCLUSIONS: RFENet demonstrated consistent improvements over strong benchmarks on both CNV and CNP tasks, with statistical evidence supporting the reliability of these gains. These results indicate that RFENet provides a reliable technical advance for automated segmentation of CNV and CNP lesions associated with AMD. To facilitate reproducibility and further research, the source code is publicly available at https://github.com/HaoSun223/RFENet.
Enhanced anchor-free pulmonary nodule detection utilizing distribution prediction and task-aligned learning strategies.
Ma L, Ji Y, Lu J … +5 more , Feng X, Li Z, Wei T, Liu L, Zhao S
Med Phys · 2026 Apr · PMID 41988734 · Publisher ↗
BACKGROUND: Efficient lung cancer screening relies on the automated and precise detection of pulmonary nodules in computed tomography (CT) images. Conventional anchor-based methodologies often rely on pre-defined anchor...
BACKGROUND: Efficient lung cancer screening relies on the automated and precise detection of pulmonary nodules in computed tomography (CT) images. Conventional anchor-based methodologies often rely on pre-defined anchor boxes, resulting in complex design and constrained adaptability to nodules of varying sizes. In contrast, anchor-free models, characterized by their simpler design, have garnered increasing interest. However, existing anchor-free methods may suffer from spatial misalignment across tasks and limited adaptability to diverse nodule shapes. PURPOSE: This study proposes a novel anchor-free model, named DistAlignNet, which integrates distribution prediction and task-aligned learning to address challenges such as spatial misalignment and the diversity of nodule shapes. METHODS: Our model utilizes a one-stage object detection architecture, enabling automatic learning of nodule shapes without explicit anchor box definitions. The model is optimized for both classification and localization tasks. In the localization task, we incorporate distribution prediction to improve noise robustness and adapt to the distinctive shape patterns of each nodule. In addition, we introduce a task-aligned learning strategy to mitigate inconsistencies between different tasks and dynamically adjust the label assignment process. RESULTS: Experimental evaluations were performed on the publicly available LUNA16 dataset, where our model achieved an average Competition performance metric (CPM) score of 0.924. Compared with the SCPM-Net model that reported the strongest performance among the existing approaches, the statistical analysis yielded and , with a Cohen's of 1.582, reflecting a statistically significant improvement. These results show that our model achieves superior performance compared with state-of-the-art anchor-based and anchor-free methods for pulmonary nodule detection. CONCLUSIONS: Quantitative results and qualitative assessments demonstrate the effectiveness of our model, indicating its potential for clinical deployment in pulmonary nodule detection.
Determinants and expected differences in the calculation of CT effective dose conversion factors for various phantom sizes and scan parameters.
Chu R, Mihailovic J, Lee C … +3 more , Yu L, Zhang D, Yang K
Med Phys · 2026 Apr · PMID 41988721 · Publisher ↗
BACKGROUND: Estimating effective dose (E) for CT is often done with dose-length product (DLP) and effective dose conversion factors, a.k.a. k-factors. Previous studies have primarily focused on k-factor values derived fr...
BACKGROUND: Estimating effective dose (E) for CT is often done with dose-length product (DLP) and effective dose conversion factors, a.k.a. k-factors. Previous studies have primarily focused on k-factor values derived from certain fixed-sized phantoms but have not extensively explored all the contributing factors and associated variations in E. PURPOSE: In this study, we investigated the contributing factors and variations of E using k-factors as surrogates in adult and pediatric CT exams and provided empirical formula to estimate a comprehensive new set of k-factors. METHODS: Effective dose was estimated using NCICT (National Cancer Institute, USA), including organ and effective dose conversion coefficient library estimated by Monte-Carlo radiation transport techniques. The k-factors were derived as a surrogate of DLP-normalized E, via E/DLP (unit: mSv/mGy·cm). We examined the variations of effective dose estimates for different phantom and scan parameters represented by the calculations of k-factors. This included variations in patient sex, weight, height, kVp, and scan length across typical adult and pediatric CT examinations of anatomical regions such as the head, neck, chest, abdomen, and pelvis. Multivariable regression analyses with/without standardization were conducted to evaluate the significance of the dependency of k-factors on the studied parameters. Tube current modulation (TCM) parameters, modeled as a function of patient attenuation using variable modulation strengths, were also incorporated to reflect realistic clinical CT dose variations. RESULTS: The variations of DLP-normalized E were substantially affected by multiple factors, exceeding 100% for adults (weight for adult pelvis) and more than 200% for pediatrics (height for pediatric head). The analysis of standardized regression coefficients revealed that patient weight is the most critical factor, except for head scans in adults, where scan length and patient height are the main determinants. In pediatric CT, both patient weight and height had critical impact. Similarly, for pediatric head CT scans, patient height and scan length are the dominant factors, as observed in adults. Consequently, the k-factors have a wide range, from 0.0099 to 0.0357 mSv/mGy·cm for adult chest and from 0.0085 to 0.0363 mSv/mGy·cm for adult abdomen. Empirical formulas to estimate k-factors were obtained based on multi-variable regression with adjusted R square ranging from 0.71 to 0.93. Based on NCICT's TCM model, the use of TCM resulted in an additional average variation of 110% in the k-factor compared to non-TCM results. CONCLUSIONS: Our study demonstrated the large variations of DLP-normalized effective dose. Standardized coefficients revealed that patient weight, height and/or scan length were the most crucial determinants. Depending on the specific TCM model used, even greater variations may be observed. This work also provided a practical framework for estimating CT effective dose using the traditional k-factor method with updated new conversion factors.
Feasibility study of a machine learning inspired approach for VMAT optimization.
Wu X, Yang D, Sheng Y … +2 more , Wu QJ, Wu Q
Med Phys · 2026 Apr · PMID 41987551 · Publisher ↗
BACKGROUND: Despite the widespread clinical adoption of volumetric modulated arc therapy (VMAT), advances in its fundamental optimization methodology have remained relatively limited, particularly with respect to open an...
BACKGROUND: Despite the widespread clinical adoption of volumetric modulated arc therapy (VMAT), advances in its fundamental optimization methodology have remained relatively limited, particularly with respect to open and researcher-accessible optimization frameworks. PURPOSE: This study introduces a novel machine learning (ML) inspired approach for VMAT optimization, reformulating the problem as a multilayer neural network solvable with modern ML toolkits. METHODS AND MATERIALS: In this framework, multileaf collimator (MLC) leaf positions and control-point weights are optimized. They are represented as trainable parameters embedded within parameterized activation functions and the final weighting layer, respectively. The dose-deposition matrix provides a fixed linear mapping. Optimization was performed using PyTorch's built-in L-BFGS optimizer with GPU acceleration. Machine-specific constraints, including maximum dose rate, gantry speed, MLC motion limits, and trajectory smoothness, were incorporated as regularization terms. The framework was evaluated using prostate cases with two arcs and head-and-neck (HN) cases with two and four arcs, with results compared against corresponding benchmark IMRT plans. RESULTS: All VMAT optimizations converged successfully, with stable reduction of total objective values and reasonable trends in machine-related regularization terms. The optimized plans were successfully imported into Eclipse TPS and delivered on a TrueBeam linac without interlocks, confirming deliverability. For prostate cases, two-arc VMAT plans achieved planning target volume (PTV) coverage and organ-at-risk (OAR) sparing comparable to benchmark IMRT plans with similar DVH characteristics. For HN cases, four-arc VMAT plans provided plan quality comparable to benchmark IMRT, and consistently improved target dose conformity and OAR sparing compared with two-arc plans, particularly in regions adjacent to complex target geometries. All observations and comparisons are consistent with established clinical experience on VMAT optimization. CONCLUSION: The proposed ML based VMAT optimization framework bridges modern machine learning optimization with treatment plan optimization and demonstrates strong potential as a flexible and extensible platform for future algorithmic development and research-driven innovations.
Experimental data-driven design of 3D range modulators for proton conformal FLASH applications.
Wang A, Setianegara J, Lin Y … +6 more , Xiao P, Gerard N, Nys J, Chen RC, Xie Q, Gao H
Med Phys · 2026 Apr · PMID 41984546 · Publisher ↗
BACKGROUND: Proton ultra-high-dose-rate (FLASH) radiotherapy has shown great potential in proton therapy owing to its superior sparing of organs at risk. Current FLASH-capable cyclotrons are restricted to single-energy d...
BACKGROUND: Proton ultra-high-dose-rate (FLASH) radiotherapy has shown great potential in proton therapy owing to its superior sparing of organs at risk. Current FLASH-capable cyclotrons are restricted to single-energy deliveries and hence, preclinical FLASH experiments have been performed with transmission beams. Recently, 3D range modulators (3D RMs) have been increasingly investigated to confer better dose conformality versus transmission beams. However, existing RM designs rely heavily on time-consuming and resource-intensive simulation-based iterations, posing substantial barriers to clinical applicability. PURPOSE: This work developed and experimentally validated a data-driven design method of 3D RMs for proton conformal FLASH. METHODS: Three 3D RMs corresponding to different target geometries were designed from an extensive base data library consisting of experimental proton spot profiles and IDDs through varying 3D RM material thicknesses from a FLASH-capable proton synchrocyclotron's (IBA ProteusONE) beam model. Each RM was designed to create conformal dose distributions of varying geometries from a single-energy beam by creating spread-out-Bragg-peaks per spot. Expedient dose calculations were performed with a MATLAB-based simplified dose engine consisting of pencil beam algorithms. All RMs were 3D-printed with resin. All plans were experimentally delivered. 1D absolute dose measurements were performed with a plane-parallel ion chamber (PPC05), 2D profile measurements were performed with radiochromic films (EBT-XD) and an end-to-end test performed with a 2D ionization chamber array (MatriXX ONE) with absolute dose calibration. Compared with traditional simulation-based approaches, which often require several days to develop a reliable machine-specific model, the proposed data-driven framework enables model establishment using approximately one day of experimental measurements. RESULTS: Good dose conformities were achieved for all targets with the FLASH dose-rates achieved through all 3D RMs. Experimentally measured doses had gamma passing rates above 95% at 2%/2 mm showing good dose calculation agreements with our expedient data-driven approach. CONCLUSIONS: This study experimentally validated an expedient data-driven design method of 3D RMs, demonstrating the feasibility of a non-simulation-based approach for 3D RM design and providing a practical foundation for clinical translation of proton conformal FLASH.
Enable preclinical, high-precision image-guided radiotherapy research with a stand-alone small animal IGRT system.
Deng W, Han D, Byrne KE … +7 more , Bhandary B, Jiang K, Poirier YP, Mossahebi S, Xu J, Shukla H, Sawant A
Med Phys · 2026 Apr · PMID 41984545 · Full text
BACKGROUND: Many preclinical studies involving novel particle radiotherapy techniques have been conducted without precise image guidance due to the lack of an image-guided small animal radiotherapy research platform for...
BACKGROUND: Many preclinical studies involving novel particle radiotherapy techniques have been conducted without precise image guidance due to the lack of an image-guided small animal radiotherapy research platform for them, such as ultra-high dose rate radiation (UHDR), mini-beam, and grid therapy. This limitation restricts the complexity of research questions that can be effectively addressed. PURPOSE: We developed a workflow to utilize 3D imaging on a Small Animal Radiation Research Platform (SARRP) for precise targeting of internal regions in small animals with a separated treatment beam. METHODS: Our approach mimics the workflow of patient treatment to decouple the imaging process from the radiation system. The SARRP cone-beam CT system was used to image and center the target. Alignment markers were then added to the top surface of the animal platform using the SARRP's well-calibrated laser system, which was consequently transferred and aligned to a collimator in a clinical treatment room. Validation of the method was performed using BB targets within both a mouse phantom and a euthanized rat. The distance between the center of the radiation field and the BB was measured to determine the setup error for the entire procedure. The use of a collimator reduces the impact of setup uncertainty in the treatment room. RESULTS: Our approach achieved a setup accuracy of approximately 0.5 mm, ensuring the target was consistently positioned at the center of the collimated radiation field center after transferring from the SARRP system. This is comparable to the isocentricity of the SARRP itself (∼±0.5 mm). CONCLUSIONS: By decoupling the imaging and radiation systems while maintaining sub-millimeter setup accuracy, this workflow introduces flexibility for conducting high-precision, image-guided preclinical studies across different radiation modalities and techniques, for example, protons, electrons, neutrons, UHDR (FLASH), Small-Field-RT, etc. The workflow uses simple equipment and can be easily implemented by other centers.
A preliminary study of a hybrid transformer with enhanced local feature fusion for nuchal translucency ultrasound image segmentation.
Wang S, Chen B, Yang H … +2 more , Liu L, Wang Z
Med Phys · 2026 Apr · PMID 41972362 · Publisher ↗
BACKGROUND: Nuchal translucency (NT) is an important ultrasound indicator in early pregnancy. Increased NT is strongly associated with chromosomal abnormalities, structural defects, and other adverse outcomes. However, t...
BACKGROUND: Nuchal translucency (NT) is an important ultrasound indicator in early pregnancy. Increased NT is strongly associated with chromosomal abnormalities, structural defects, and other adverse outcomes. However, the low signal-to-noise ratio (SNR), numerous artifacts, and poor contrast present in ultrasound images pose considerable challenges to existing segmentation methods. PURPOSE: To address these challenges, this paper proposes a novel Transformer model, TransNTSeg, for NT segmentation. This study serves as a preliminary proof-of-concept to evaluate the feasibility of the proposed architecture. METHODS: A clinical dataset of 123 B-mode ultrasound images is collected. To ensure rigorous evaluation, the dataset is partitioned into 98 images for training/validation (using 5-fold cross-validation) and 25 images for independent testing. The architecture incorporates two main innovations: first, an enhanced Dense Dilated Mixed Feed-Forward Network (DDMix-FFN) module within the Transformer that more robustly integrates global dependencies and local context to handle image noise and boundary ambiguity; and second, a DDMix-FFN Transformer context bridge that fuses multi-scale encoder features for precise localization of the small NT region. Performance is compared against eight state-of-the-art benchmarks (e.g., U-Net, MT-UNet) using Intersection over Union (IoU) and Dice Similarity Coefficient (DSC). Statistical significance is assessed using the paired t-test with Bonferroni correction to control for multiple comparisons (adjusted p < 0.0015625), and practical significance is quantified using Cohen's d effect size. RESULTS: TransNTSeg achieves the highest performance on the test dataset, with a mean IoU of 0.763 ± 0.05 and a mean DSC of 0.852 ± 0.03. Rigorous statistical analysis (Bonferroni corrected p < 0.0015625) confirms considerable improvements. Based on the evaluation of effect sizes, TransNTSeg is identified as the best-performing method, demonstrating a substantial effect size (d = 0.745) against the closest competitor (MT-UNet). CONCLUSION: While demonstrating competitive performance within this controlled setting, these promising outcomes highlight the potential of TransNTSeg for accurate NT segmentation. This warrants further validation on multicenter, multi-vendor datasets to establish its generalizability for broad clinical use.
An exploratory investigation of dual-domain denoising network for field-free-line magnetic particle imaging.
Wu X, Xiong F, An Y … +2 more , Tian J, Jiang J
Med Phys · 2026 Apr · PMID 41968776 · Publisher ↗
BACKGROUND: Magnetic particle imaging (MPI) is a functional imaging modality that enables highly sensitive tracking of magnetic nanoparticles. Field-free-line (FFL) MPI provides higher signal-to-noise ratio (SNR) than fi...
BACKGROUND: Magnetic particle imaging (MPI) is a functional imaging modality that enables highly sensitive tracking of magnetic nanoparticles. Field-free-line (FFL) MPI provides higher signal-to-noise ratio (SNR) than field-free-point MPI, however, noise in the sinogram domain and its propagation during reconstruction can introduce artifacts that degrade image quality. PURPOSE: This study aims to develop a dual-domain denoising framework to improve SNR in both the sinogram and image domains for FFL-MPI tomographic reconstruction. METHODS: We propose DudoMTD, a dual-domain cascaded network consisting of two components: (1) a Sinogram Denoising Transformer (SDT) integrates convolutional layers with a Vision Transformer to capture both local and long-range angular dependencies in the sinogram domain; and (2) an edge guiding autoencoder (EGA) operates in the image domain using convolutional filtering and an adaptive Canny operator to preserve structural boundaries. Simulated sinograms were generated using a standard FFL-MPI forward model based on the Langevin magnetization equation and system matrix formulation. The dataset consisted of 13 039 simulated images, with 80% used for training and validation and 20% for testing. In addition, FFL-MPI phantom imaging data were used to evaluate the model under realistic measurement noise. Performance was compared with three benchmarks methods-DuDoNet, RED-CNN, and DnCNN-using peak signal-to-noise ratio (PSNR), structural similarity (SSIM), and root mean square error (RMSE). Statistical significance was assessed using the Wilcoxon signed-rank test, with Benjamini-Hochberg correction for multiple comparisons. Effect sizes (Cohen's d) were reported to quantify the magnitude of improvements. RESULTS: Across four noise conditions (fixed SNRs of 15, 20, 30 dB; 30 dB with additional Poisson noise), DudoMTD showed statistically significant improvements (p < 0.05) over all benchmark methods, with medium-to-large effect sizes. Specifically, the PSNR gains corresponded to Cohen's d values ranging from 0.378 to 1.311, while SSIM improvements yielded d values of 0.323-1.305. Performance gains were pronounced in the 20 dB, 30 dB, and 30 dB + Poisson noise scenarios, where DudoMTD exceeded competing methods by 3%-10% in SSIM and demonstrated consistently superior structural preservation. In phantom experiments acquired on an in-house FFL-MPI system, DudoMTD achieved the highest average SNR (13.809) and effectively suppressed measured noise. CONCLUSION: DudoMTD mitigates dual-domain noise in FFL-MPI and improves tomographic image quality across diverse noise conditions. These improvements may facilitate downstream quantitative MPI applications, particularly in low-dose imaging scenarios.
Diffusion boost: Leveraging diffusion model for groupwise registration in myocardial T1 mapping.
Yue C, Wang Q, Guo Y … +2 more , Tao Q, Wang Y
Med Phys · 2026 Apr · PMID 41968741 · Publisher ↗
BACKGROUND: As a quantitative magnetic resonance imaging (MRI) technique, myocardial T1 mapping plays a crucial role in the diagnosis and treatment of cardiovascular diseases. In practice, involuntary cardiac and respira...
BACKGROUND: As a quantitative magnetic resonance imaging (MRI) technique, myocardial T1 mapping plays a crucial role in the diagnosis and treatment of cardiovascular diseases. In practice, involuntary cardiac and respiratory motion often results in reduced accuracy and precision in T1 estimation. Therefore, image registration remains crucial for accurate and precise myocardial T1 mapping. Compared with pairwise registration that warps each baseline image to a predefined template, groupwise registration aligns all images from one sequence simultaneously without the need for a template. However, a persistent challenge is the difficulty of extracting the structural representation of T1 mapping data that contains vastly varying contrast, which severely undermines the performance of image registration. PURPOSE: The purpose of this study is to incorporate the learning capabilities of the diffusion model to tackle the main challenge encountered in the registration of myocardial T1 mapping. Our goal is to align all images within an image series simultaneously in a groupwise manner. METHODS: In this article, we propose a novel template-free groupwise registration framework that can align one T1-weighted image series through a single forward propagation. Notably, we introduce the diffusion process to effectively boost the structural information extraction under the drastic contrast changes for reliable image registration. Furthermore, we design a Hybrid Attention Feature Fusion (HAFF) module to promote the multi-scale feature fusion from diffusion to registration. To evaluate the registration performance of the proposed model, experiments are conducted on a publicly available myocardial T1 mapping dataset comprising 210 consecutive patients, using an independent test set for comparison experiments and ablation studies. RESULTS: Experimental results demonstrated the great superiority of our proposed method in the registration of myocardial T1 mapping. Quantitatively, the proposed method resulted in a Dice score of 0.839, groupwise Dice score of 0.601, Hausdorff distance of 10.389 mm, and T1 mapping error of 11.372 ms, surpassing the current state-of-the-art approaches. CONCLUSIONS: Our proposed framework realizes robust groupwise registration for myocardial T1 mapping by leveraging the state-of-the-art diffusion model, demonstrating its strong feature extraction capacity for image registration, beyond image generation.
Multi-parametric MRI synthesis for glioblastoma from quantitative MR fingerprinting: Quantitative synthetic neural network (QS-Net).
Ni Y, Liu C, Li W … +11 more , Lin L, Ouyang R, Cao P, Lee HV, Helali AE, Wong YL, Wang X, Wang P, Ren G, Cai J, Li T
Med Phys · 2026 Apr · PMID 41968695 · Full text
BACKGROUND: Diagnosis and treatment of glioblastoma (GBM) rely on multiparametric MRI (mpMRI), but mpMRI is time-consuming and costly. Deep learning-based synthesis methods have been proposed to streamline acquisition; h...
BACKGROUND: Diagnosis and treatment of glioblastoma (GBM) rely on multiparametric MRI (mpMRI), but mpMRI is time-consuming and costly. Deep learning-based synthesis methods have been proposed to streamline acquisition; however, their generalizability is limited by variability in qualitative input contrasts across sites and scanners. PURPOSE: To overcome this limitation, we developed and evaluated a generalizable deep learning model that synthesizes mpMRI contrasts directly from quantitative magnetic resonance fingerprinting (MRF) maps in GBM patients. The proposed Quantitative Synthesis Network (QS-Net) employs a deeply supervised residual U-Net generator within an adversarial framework, combined with a two-stage training strategy to separate anatomical and pathological learning. METHODS: We collected MRF-derived T1 and T2 maps, along with conventional mpMRI sequences (T1w, T2w, T1-FLAIR, T2-FLAIR, and SWI), from 32 healthy volunteers and retrospectively from 18 GBM patient scans. The proposed QS-Net was initially trained on healthy volunteer data (20 scans for training, 12 for testing) to learn general anatomical features. Subsequently, it was fine-tuned using 9 GBM patient scans to adapt to pathological characteristics, with the remaining 9 patient scans reserved for independent testing. We compared the performance of QS-Net against three state of the art deep learning models: Res-Unet, conditional GAN, and Swin-Transformer, using both quantitative metrics (MAE, SSIM, and PSNR) and qualitative assessments. Additionally, we assessed the generalizability of the models by evaluating their external validation performance when trained with either conventional MRI or quantitative MRF inputs. RESULTS: QS-Net outperformed the comparison models in synthesizing T1w, T2w, SWI, and T2-FLAIR images for GBM patients, achieving the best results across all quantitative metrics: MAE (1.18 ± 0.52, 1.01 ± 0.36, 1.05 ± 0.37, 1.45 ± 0.76), SSIM (0.934 ± 0.037, 0.939 ± 0.039, 0.934 ± 0.034, 0.926 ± 0.053), and PSNR (29.69 ± 3.21, 29.35 ± 2.29, 29.64 ± 2.58, 27.56 ± 3.39), respectively. Qualitative analysis demonstrated that QS-Net generated synthetic images with superior resemblance to ground truth, accurately delineating tumor boundaries and preserving intra-tumoral texture. Furthermore, the generalizability test revealed that models trained on standardized quantitative MRF input maps consistently outperformed models trained on vendor-specific qualitative MRI inputs across all architectures and metrics (p < 0.005). CONCLUSIONS: We developed QS-Net, a deep learning model for high fidelity mpMRI synthesis from quantitative MRF maps, and demonstrated that this quantitative-input paradigm enables superior cross-vendor generalization over conventional qualitative MRI-based approaches.
Megavoltage CT enhancement for cervical cancer tomotherapy using a generative adversarial network with deformable convolution and self-attention.
Jian W, Zhu L, Zhang B … +6 more , He Q, Yang G, Jin H, Li F, Meng H, Wang X
Med Phys · 2026 Apr · PMID 41968645 · Publisher ↗
BACKGROUND: Megavoltage computed tomography (MVCT) is an essential imaging modality for verifying patient positioning in helical tomotherapy. However, its clinical application in daily anatomical monitoring and adaptive...
BACKGROUND: Megavoltage computed tomography (MVCT) is an essential imaging modality for verifying patient positioning in helical tomotherapy. However, its clinical application in daily anatomical monitoring and adaptive radiotherapy is hindered by inherent image artifacts and poor soft-tissue contrast. This issue is particularly pronounced in pelvic radiotherapy, where the intra- and interfraction anatomical variations necessitate high-quality image guidance to ensure precise dose delivery. PURPOSE: We developed a deep-learning-based framework for MVCT enhancement to improve anatomical visualization and facilitate accurate adaptive treatment planning for cervical cancer. METHODS: This study analyzed a retrospective cohort of 170 patients with cervical cancer who underwent helical tomotherapy. The proposed deep-learning-based algorithm employed a generative adversarial network (GAN) that integrated deformable convolution and a self-attention mechanism (SADC-EGAN) to improve MVCT image quality. Comparative analyses were conducted against representative baseline methods, including U-Net, Attention U-Net, U-Net++, Swin-UNet, CycleGAN, and Pix2Pix. Model performance was assessed using quantitative metrics, including mean absolute error (MAE), peak signal-to-noise ratio (PSNR), structural similarity index measure (SSIM), and Fréchet inception distance (FID). RESULTS: The synthetic computed tomography (sCT) images generated by the proposed SADC-EGAN method demonstrated superior Hounsfield unit (HU) accuracy and structural similarity compared to the original MVCT. Specifically, the MAE between the sCT and kilovoltage computed tomography (kVCT) was reduced to 36.32 ± 6.69 HU, compared with 56.72 ± 9.09 HU for MVCT. In terms of image quality, the sCT images exhibited notable enhancements over MVCT images, with higher PSNR (32.54 ± 2.31 vs. 29.40 ± 1.56 dB), improved SSIM (0.93 ± 0.01 vs. 0.89 ± 0.02), and substantially lower FID (66.68 ± 22.31 vs. 153.52 ± 28.77). CONCLUSIONS: The proposed SADC-EGAN framework, integrating deformable convolutions and self-attention, effectively generated high-quality kVCT-like images from MVCT, improving both HU accuracy and image quality. This approach has clinical potential to enable online adaptive helical tomotherapy for cervical cancer.
About
- Frequency
- Sun
- Papers found
- 200
- RSS feed
- Subscribe