Physics in medicine and biology最新文献

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.

Objective. Clinical implementation of deformable image registration (DIR) requires voxel-based spatial accuracy metrics such as manually identified landmarks, which are challenging to implement for highly mobile gastrointestinal (GI) organs. To address this, patient-specific digital twins (DTs) modeling temporally varying motion were created to assess the accuracy of DIR methods.Approach. A total of 21 motion phases simulating digestive GI motion as 4D image sequences were generated from static 3D patient scans using published analytical GI motion models through a multi-step semi-automated pipeline. Eleven datasets, including six T2-weighted FSE MRI (T2w MRI), two T1-weighted 4D golden-angle stack-of-stars, and three contrast-enhanced computed tomography scans were analyzed. The motion amplitudes of the DTs were assessed against real patient stomach motion amplitudes extracted from independent 4D MRI datasets using hierarchical motion reconstruction. The patient-specific DTs were then used to assess six different DIR methods using target registration error, Dice similarity coefficient (DSC), and the 95th percentile Hausdorff distance using summary metrics and voxel-level granular visualizations. Finally, for a subset of T2w MRI scans collected from patients treated with magnetic resonance-guided radiation therapy, dose distributions were warped and accumulated to assess dose warping errors (DWEs), including evaluations of DIR performance in both low- and high-dose regions for patient-specific error estimation.Main results. Our proposed pipeline synthesized patient-specific DTs modeling realistic GI motion, achieving mean and maximum motion amplitudes and a mean log Jacobian determinant within 0.8 mm and 0.01, respectively, similar to published real-patient gastric motion data. It also enables the extraction of detailed quantitative DIR performance metrics and supports rigorous validation of dose mapping accuracy prior to clinical implementation.Significance. The developed pipeline enables rigorously testing DIR tools for dynamic, anatomically complex regions facilitating granular spatial and dosimetric accuracies.

Introduction.The clinical interest in particle therapy continues to grow due to its precise dose distribution and increased biological effectiveness. Recent attention has focused on ultra high dose rate (UHDR) irradiation, which has been associated with reduced normal tissue toxicity, an effect that is partially attributed to altered radiolysis chemistry. However, most existing experimental studies only compare UHDR irradiation at a fixed dose rate with conventional dose rates, providing limited insights into the continuous dose rate dependence of radiolysis processes.Methods.This study aims to systematically investigate the dose rate dependence of hydrogen peroxideH2O2formation, a long-lived product of water radiolysis, across a wide range of dose rates and under varying pulse structures and irradiation conditions using protons and carbon-ions.Results.The results demonstrate a non-linear relationship, withH2O2yields increasing up to 20-30 Gy s^-1before declining at higher dose rates. These findings help to reconcile discrepancies between current experimental observations and simulation predictions, emphasizing the critical role of instantaneous dose rate distribution and pulse characteristics. The data provide essential input for advancing radiolysis modeling and for understanding UHDR effects in particle therapy.

Objective.During radiotherapy, the radiation dose delivered to circulating blood can result in radiation-induced lymphopenia, which is correlated with adverse clinical outcomes like lower survival. Increasingly complex models to simulate radiation dose delivery to circulating blood have been developed in response, and their inclusion during radiotherapy treatment planning has been suggested. However, performing full dynamic blood dose simulations which take into account temporal considerations such as blood flow dynamics and treatment delivery time during the iterative treatment planning process is currently infeasible. This work presents a quasi-instantaneous deep learning-based approach to estimate blood dose simulation results to allow for their inclusion during treatment planning.Approach.We used treatment planning computed tomography images and dose-volume histograms of 157 head-and-neck cancer patients to perform dynamic blood dose simulations (HEDOS). Subsequently, a deep neural network composed of fully-connected layers and a Transformer encoder was trained to estimate the blood dose distribution obtained from HEDOS, using the same inputs as HEDOS. We used 126 patients' data for training and internal validation and the remaining 31 patients' data for testing. To evaluate the proposed method, we calculated the Kullback-Leibler (KL) divergence between the prediction results and the ground truth data. Additionally, we compared the minimum dose delivered to 90% of the blood particles receiving the highest dose (D_90%) to estimate the model's clinical efficacy.Main results.The average and standard deviation of KL divergence between the prediction and the ground truth were 0.099 and 0.092, respectively. The D_90%calculated from the predicted distribution showed a mean-absolute-percentage error of 4.60% compared to the ground truth.Significance.A deep learning-based model capable of accurately and quasi-instantaneously predicting the results of dynamic blood dose simulations was developed, paving the way for the inclusion of dynamic blood dose simulations during radiotherapy treatment planning.

Objective.To develop a pixel-row-specific retrospective gating technique for dynamic micro-computed tomography (micro-CT) imaging that removes motion artifacts introduced by rolling shutter acquisition on scientific complementary metal-oxide-semiconductor detectors.Approach.The proposed by-row gating technique-which accounts for the differences in acquisition timing of individual rows of pixels in each projection frame-is compared to the conventional method of by-frame gating wherein all pixels of a frame are assumed to have been acquired simultaneously. The errors associated with each gating method are demonstrated analytically, numerically (by simulated scanning of a dynamic phantom), and experimentally (by scanning an inflatable object and anex vivolung sample).Main results.By-row gating minimizes phase errors in the projection image domain prior to reconstruction, resulting in images with less out-of-phase motion, reduced motion artifact, and higher spatiotemporal resolution. Phase errors produced from by-row gating are determined only by the constraints of detector integration time and reconstructed phase bin width. Conversely, when gating by frame, spatially varying phase errors in the projection image domain are propagated into the reconstructed images, resulting in out-of-phase motion and motion artifacts that increase in severity with distance from the center of rotation. The magnitude and spatial pattern of this error depend on the ratio of object phase rate to frame acquisition rate, beam collimation and rolling shutter orientation.Significance.We identify two key parameters that determine the severity of micro-CT imaging artifacts caused by rolling shutter acquisition, and we provide an algorithm for by-row gating that eliminates artifacts and can be implemented on any rolling shutter system. By-row gating outperforms conventional by-frame gating, enabling accurate reconstruction of high-speed phenomena without compromising the high spatiotemporal resolution necessary to study microstructural dynamics.

Objective.To validate a Monte Carlo (MC)-based offline positron emission tomography (PET) range verification framework and quantitatively evaluate range uncertainties associated with material assignment strategies, specifically focusing on the clinical practice of overriding silicone implants as adipose tissue in breast proton therapy.Approach.A FLUKA-based MC model was developed. Validation followed a stepwise program using (1) a PMMA phantom, (2) a homogeneous silicone block, and (3) a composite prosthesis-PMMA setup to verify intrinsic precision and material-specific accuracy. Additionally, a cohort of eight breast cancer patients (three with implants) was included for clinical demonstration. Range deviations were quantified by measuring the depth difference at the 50% distal activity falloff (ΔR50) between simulated and measured PET images. For silicone scenarios, material assignment impact was analyzed by overriding the implant as silicone, adipose tissue, or applying no override.Main results.The MC model demonstrated intrinsic millimeter-level accuracy. In homogeneous silicone phantom experiments, defining the material as silicone yielded the highest agreement with measurements (ΔR50: -0.05-0.88 mm). Reassigning silicone as adipose tissue introduced a systematic deviation but remained within acceptable limits (-1.24--0.83 mm), whereas applying no override resulted in the largest errors (-3.25--2.48 mm). Composite phantom results followed a similar trend. In clinical cases, the method successfully verified range delivery, although positioning uncertainties contributed to larger inter-field variability.Significance.This study validates an MC-based PET verification framework with 2 mm intrinsic accuracy, establishing it as a reliable tool for silicone-implanted breast cancer patients. The results quantitatively confirm that overriding silicone as adipose tissue keeps range uncertainty within 2 mm despite distinct elemental compositions, whereas failing to override the material leads to errors exceeding 3 mm.

Background.For medical imaging, usually, it is crucial to have high spatial resolution. Studies have demonstrated that the novel dual-layer flat panel detectors (FPDs) can acquire extra spatial information to enable super-resolution cone beam CT(CBCT) imaging compared to the conventional single-layer FPDs.Objective.The aim of this study is to investigate the feasibility of realizing near-isotropic super-resolution CBCT imaging with a dual-layer flat panel detector.Methods.To retrieve the near-isotropic super-resolution imaging information, a general mathematical signal model that includes the relative shift, namely, Δualong the horizontalu-axis, Δvalong the verticalv-axis, between the two detector pixel arrays and the gap Δdbetween the two detector layers, is established. Afterwards, an recurrent neural network-based deep neural network, named as two-dimensional 2D-suRi-Net, is employed to efficiently retrieve the projections having near-isotropic super-resolution imaging information. Numerical simulations were performed to investigate the impact of the relative shift (Δu, Δv) and the gap Δdunder different scenarios. The real performance of this proposed super-resolution CBCT imaging approach is validated by a pig leg specimen and an intersecting cylinder phantom.Results.It is found that introducing half pixel shift, i.e. Δu= Δv=0.5δdel(δdeldenotes the pixel dimension), between the two detector layers is necessary for super-resolution CBCT imaging, particularly when the detector gap Δdis less than 3 mm. Results demonstrate that the proposed 2D-suRi-Net can effectively retrieve higher spatial resolution information from the acquired low-energy and high-energy projections having lower spatial resolution. Quantitatively, the spatial resolution difference between the reconstructed super-resolution CBCT images on the axial and coronal plane is less than 10%, demonstrating the near-isotropic super-resolution imaging capability of the 2D-suRi-Net method.Conclusion.In summary, this study demonstrates the feasibility of near-isotropic super-resolution CBCT imaging for dual-layer FPD based CBCT imaging systems.

Objective.To quantify differences in quantum noise fraction (QNF) between direct and indirect mammography detectors across spatial frequency bands, assessing the ability of QNF to reveal performance degradation at high frequencies.Approach.QNF of various mammography detectors were analyzed by selectively restricting the analysis to medium and high spatial frequency bands. The study included detectors from different manufacturers, both direct and indirect types. Noise components were determined from the sub-bands of a wavelet packet decomposition applied to uniform images acquired at different entrance kerma levels. The noise components were obtained from the terms of a second-degree polynomial fitted to the total noise variance of the sub-bands as a function of entrance kerma to the detector. The results obtained were compared with detective quantum efficiency calculations and contrast-detail curves of the detectors studied.Main results.The QNF results obtained in the different frequency bands studied differ significantly from those calculated in the image domain. In all cases, the QNF decreases with increasing frequency across the entire range of kerma values. Moreover, this decrease is much more pronounced for indirect detectors than for direct ones. Despite showing similar QNF values in the spatial domain, the differences between direct and indirect detectors become substantial in the mid-frequency bands and increase even further in the high-frequency bands. Among the image quality metrics evaluated, the largest differences between detectors were observed in the frequency-dependent evolution of the QNF.Significance.Decreasing values of the QNF reflect the loss in signal-to-noise ratio transfer from input to output in digital detectors, which is directly associated with a loss in detectability. The greater reduction in quantum noise observed in the mid- and high-frequency components of indirect detectors may indicate a decline in their performance for tasks involving the detectability of small objects. QNF exhibited the greatest sensitivity to scintillator-induced signal blurring, making it the metric that best distinguishes between direct and indirect detector designs.

Accurate segmentation of abdominal three-dimensional (3D) vascular structures from computed tomography (CT) scans is crucial for clinical applications yet remains challenging due to dependency on large annotated datasets through effective pretraining and poor generalization via cross-domain feature alignment. In this paper, we proposes a novel transformer-based framework based on masked autoencoder (MAE) and UNEt TRansformers (UNETR), dubbed as Adaptive MAE-UNETR, that integrates self-supervised pretraining and adversarial domain adaptation to achieve robust artery/vein segmentation with enhanced generalization. Specifically, first, we develop a MAE pretraining paradigm with 3D CT scans as input, to learn hierarchical feature representations through self-reconstruction tasks from source domain data. This targeted design ensures high fidelity in feature extraction and improves segmentation accuracy and stability. Second, the pretrained encoder is transferred to a UNETR segmentation network augmented with domain adaptation technique, which adversarially aligns feature distributions between source and target domains via a domain discriminator. Third, we establish an segmentation framework that simultaneously optimizes segmentation accuracy and domain invariance. Comprehensive evaluations on three public datasets demonstrate state-of-the-art performance of our proposed method. On AMOS22 dataset, our model achieves DSC scores of 0.924 (aorta) and 0.892 (inferior vena cava, IVC). Cross-domain tests yield 0.917/0.888 on BTCV dataset and 0.931/0.906 on FLARE23 dataset, showing consistent superiority over conventional methods across diverse datasets.

Objective.Clustered pinhole (CP) collimation currently supports sub-millimeter resolution imaging up to ∼1 MeV, enabling SPECT of alpha and beta emitters with gamma emissions, simultaneous multi-isotope PET and PET/SPECT, and positron range-free PET. Nonetheless, increasing sensitivity in the original CP designs by enlarging pinhole diameters is limited, as the resulting pinhole opening cones would overlap.Approach. To address this limitation, the use of Super-Cluster (SC) collimation was evaluated in a simulation study. Two SC designs were assessed: a standard configuration (SC-ST) offering a resolution-sensitivity trade-off similar to CP, and a high-sensitivity variant (SC-HS) with larger pinhole diameters to enhance sensitivity. Their performance was compared to CP collimation for¹⁸F at concentrations of 1.0, 0.1, 0.05 MBq ml^-1and ⁸⁹Zr at 2.0, 0.2, 0.1 MBq ml^-1, evaluating sensitivity, image resolution, recovery coefficients, and uniformity.Main results.CP and SC-ST showed comparable sensitivity and image resolution. Both resolved¹⁸F rods of 0.9, 1.4, and 1.8 mm at 1.0, 0.1, and 0.05 MBq ml^-1, respectively. For ⁸⁹Zr, rods down to 1.0 mm and 1.6 mm were resolved at 2.0 and 0.2 MBq ml^-1, but none at 0.1 MBq ml^-1. Compared to CP and SC-ST, SC-HS increased sensitivity threefold for¹⁸F and twofold for ⁸⁹Zr. At the highest activity, SC-HS showed slightly reduced resolution for¹⁸F (1.0 mm) and similar for ⁸⁹Zr (1.0 mm). However, it clearly outperformed both other collimators at lower activities, resolving¹⁸F rods of 1.2 and 1.4 mm at 0.1 and 0.05 MBq ml^-1, respectively, and ⁸⁹Zr rods of 1.4 and 1.6 mm at 0.2 and 0.1 MBq ml^-1. Additionally, SC-HS showed superior contrast recovery. Image uniformity remained consistent across all collimators, confirming effective angular sampling.Significance.The new SC geometry enables high-sensitivity collimation for high gamma energies, improving image quality at low activities. These results demonstrate SC collimation's strong potential for sensitivity-critical applications.