Physics in medicine and biology最新文献

This study validates the effectiveness of a quantitative detector calibration-originally developed for monochromatic synchrotron radiation-when applied with a polychromatic x-ray source, using both a Hamamatsu CMOS flat panel detector and a WidePIX photon-counting detector with Timepix chips. We show that, despite the energy dependence of the detector response, this calibration is a simple yet robust tool for suppressing computed tomography (CT) ring artefacts in both attenuation contrast and phase contrast CT. Furthermore, we demonstrate that this algorithm is effective even when the ring artefacts are present at a level below the shot noise. This makes it especially useful for data that is filtered in a way that suppresses noise, such as low-dose imaging via phase retrieval.

Objective.Due to the emergence of ultra-high dose rate proton therapy techniques, the need for advancedin-vivodosimetry tools able to monitor such beams has become more critical. This study evaluates the potential of silica optical fibers as real-time scintillating detectors for quality assurance of scattered and scanned pulsed proton beams at high and standard dose rates.Approach.The radioluminescence (RL) of short pieces of thin Gd³⁺-, Ce³⁺-, and N-doped silica fibers was studied focusing on their response to dose rate and energy, and their ability to perform depth-dose measurements, and for pencil beam scanning proton beams, spot size evaluation and real-time beam monitoring. The fibers response was evaluated in a continuous scattered proton beam and a synchrocyclotron scanned pulsed proton beam.Main results.Up to 130 Gy·s^-1, fibers response was linear as a function of dose rate. For both proton beams, range was measured with less than 1.1 mm deviation with the three fibers, despite scintillation quenching (SQ) from 63 to 226 MeV. With the same data, spot size was estimated with less than 0.3 mm difference from the reference, taking advantage of the real-time RL acquisitions for scanned beams. Also, characterization of synchrocyclotron specific dose delivery by repainting, was investigated. Fibers were able to measure the relative dose contribution of each painting, according to the system log-files (less than 2% difference). Moreover, the ability of the system to resume properly the irradiation after beam interruption was validated in real-time through fibers acquisitions and were correlated to the system log-files.Significance.RL-based silica fiber sensors appeared to be viable detectors for proton beam quality assurance (proton range, spot size) at standard and high dose rates despite showing SQ and showed potential for real-time dosimetry for scanned proton beams and log-files QA, in addition to high dose rate proton beams.

Objective.To develop a machine learning-based framework for accurately modeling the anode heel effect in Monte Carlo(MC) simulations of x-ray imaging systems, enabling realistic beam intensity profiles with minimal experimental calibration.Approach.Multiple regression models were trained to predict spatial intensity variations along the anode-cathode axis using experimentally acquired weights derived from beam measurements across different tube potentials. These weights captured the asymmetry introduced by the anode heel effect. A systematic fine-tuning protocol was established to minimize the number of required measurements while preserving model accuracy. The models were implemented in the OpenGATE 10 and GGEMS MC toolkits to evaluate their integration feasibility and predictive performance.Main results.Among the tested models, gradient boosting regression (GBR) delivered the highest accuracy, with prediction errors remaining below 5% across all energy levels. The optimized fine-tuning strategy required only six detector positions per energy level, reducing measurement effort by 65%. The maximum error introduced through this fine-tuning process remained below 2%. Dose actor comparisons within MC simulations demonstrated that the GBR-based model closely replicated clinical beam profiles and significantly outperformed conventional symmetric beam models.Significance.This study presents a robust and generalizable method for incorporating the anode heel effect into MC simulations using machine learning. By enabling accurate, energy-dependent beam modeling with limited calibration data, the approach enhances simulation realism for applications in clinical dosimetry, image quality assessment, and radiation protection.

Objective. Magnetic resonance imaging (MRI) guided radiotherapy requires the delineation of gross tumor volumes (GTV) in daily MRI from MRI-linacs. Specialized models have been developed for this task for certain tumors. This study investigated an alternative, using promptable foundation models.Approach. Promptable foundation models were prompted with six different sparse geometric prompt types (points, boxes, 2D masks) to produce GTV segmentation masks, including Segment-anything 2 (SAM2), SAM2 fine-tuned for medical imaging (MedSAM2), and nnInteractive, an nnUnet-based promptable model for medical imaging. A diverse multi-institutional dataset of clinical GTV masks from the abdomen, lung, liver, pancreas, and pelvis sites on MRI scans from MRI-linacs was used to evaluate model outputs using various metrics, including the Dice similarity coefficient (DSC).Main results. The models produced segmentation masks comparable or superior to those from domain-specific models with median DSCs of up to 0.85 (nnInteractive-mask3 prompt). Prompts with more spatial information yielded better results with lower variance, with the effect reduced for nnInteractive and MedSAM2. These produced overall better results (median DSC over all prompt types 0.75 for nnInteractive, 0.70 for MedSAM2, 0.54 for SAM2).Significance. This investigation showed that promptable foundation models can in principle be used for GTV segmentation in MRI across multiple tumor types, although more research is necessary to reduce the variance and improve model performance.

Objective.To develop an online quality control (QC) system for accurate assessment of dwell position and dwell time in high-dose-rate (HDR) brachytherapy, and to investigate the potential of neural networks so as to improve the robustness and stability of the proposed system.Approach.An integrated framework was constructed using a Basler high-speed camera (144 fps, 1920 × 1200 pixels), custom illumination, and dedicated software. Experiments were conducted with a GammaMedPlus iX afterloader equipped with a stepping192Irsource, testing various step sizes (0.2 cm, 0.5 cm, 1.0 cm) and dwell times (2.0 s, 3.0 s, 10.0 s). The core online algorithm employed a frame-difference method for source tracking, while offline analysis evaluated the RT-DETRv2 neural network for source localization.Main results.The system achieved high spatial resolution (0.083 mm) and temporal resolution (7.0 ms). Primarily due to guidewire bending, positional deviations remained below 0.1 cm and increased with guidewire length. After position correction, the positional deviation was reduced to about 0.01 cm. Dwell time deviations were within 10.0 ms. RT-DETRv2 demonstrated outstanding localization accuracy (91% of predictions within 0.26 mm) in various conditions. However, its processing latency of 0.35 s per image makes it unsuitable for online monitoring but well-suited for offline or auxiliary verification in this system.Significance.This work presented a technically feasible online QC method for HDR brachytherapy that enabled precise verification of source delivery parameters. Moreover, the successful application of deep learning for source detection established a foundation for future enhancements in reliability and automation of the proposed QC system.

Objective.The noninvasive characterization of soft tissue microstructure remains challenging and has a significant clinical impact on diagnosis and therapy monitoring. During high-intensity focused ultrasound (HIFU) treatments, coagulation necrosis is accompanied by mechanical changes. The objective of this work is to use the anisotropy arising at the cellular level as a new biomarker for treatment evaluation.Approach.We demonstrate that HIFU induces anisotropic alterations in the cellular architecture of liver tissue, which are detectable through the angular dependence of the backscattering coefficient (BSC). Also,in vivoexperiments reveal a distinct anisotropic histological pattern localized in the HIFU-treated region.Main results.We show that the degree of anisotropy deduced from BSC measurements is correlated with the histological observations. Moreover, anisotropy increases with delivered energy, providing a quantitative link between treatment parameters and tissue response.Significance.These findings establish BSC anisotropy as a previously unexplored signature of thermal lesions, offering a promising approach for monitoring and feedback in thermal therapeutic ultrasound applications. This breakthrough could open the door to next-generation imaging tools, accelerating the widespread adoption of this highly effective therapeutic modality.

Objective: To develop an efficient deep learning framework for precise 3D segmentation of complex orbital structures in multi-sequence MRI and robust assessment of thyroid eye disease (TED) activity, thereby addressing limitations in computational complexity, segmentation accuracy, and integration of multi-sequence features to support clinical decision-making. Approach: We propose RQNet, a U-shaped 3D segmentation network that incorporates the novel Refined Query Transformer Block (RQT Block) with Refined Attention Query Multi-Head Self-Attention (RAQ-MSA). This design reduces attention complexity from O(N²) to O(N·M) (M ll N) through pooled refined queries. High-quality segmentations then feed into a radiomics pipeline that extracts features per region of interest-including shape, first-order, and texture descriptors. The MRI features from the three sequences-T1-weighted imaging (T1WI), contrast-enhanced T1-weighted imaging (T1CE), and T2-weighted imaging (T2WI)-are subsequently integrated, with support vector machine (SVM), random forest (RF), and logistic regression (LR) models employed for assessment to distinguish between active and inactive TED phases. Main Results: RQNet achieved Dice Similarity Coefficients of 83.34-87.15% on TED datasets (T1WI, T2WI, T1CE), outperforming state-of-the-art models such as nnFormer, UNETR, SwinUNETR, SegResNet, and nnUNet. The radiomics fusion pipeline yielded area under the curve (AUC) values of 84.65-85.89% for TED activity assessment, surpassing single-sequence baselines and confirming the benefits of multi-sequence MRI feature fusion enhancements. Significance: The proposed RQNet establishes an efficient segmentation network for 3D orbital MRI, providing accurate depictions of TED structures, robust radiomics-based activity assessment, and enhanced TED assessment through multi-sequence MRI feature integration. .

Objective.Targeted radionuclide therapy using Auger electron (AE) emitting radionuclides is promising for the treatment of small tumour lesions and metastases. However, selecting the optimal AE emitting radionuclide and its ideal targeting cell compartment is necessary to reach their full potential. The aim of this study was to compare the absorbed doses in single cells and from neighbouring cells from emerging AE-emitting radionuclides targeting different cellular compartments.Approach.We computedS-values in concentric water spheres of unit density, using simulations with Geant4-DNA (v. 11.1.1) and MIRDcell (v. 3.13), for the radionuclides Pd-103 (including its Rh-103 m daughter), Pd-109, Tb-161, Er-165, and commonly studied radionuclides like Lu-177, I-123, I-125, I-131, In-111, and Y-90, focusing on the cell nucleus, cytoplasm, and the cell membrane as target regions; alongside cross-dose effects representing neighbouring cells.Main Results.A thorough comparison showed our Geant4-DNAS-values align within 10% of other Monte Carlo methods for well-studied geometries (e.g. nucleus, cytoplasm, and the entire cell as targets). When nucleus-bound, Tb-161 showed a 5-fold higher and Pd-103 a 3-fold higher nuclear dose compared to Lu-177. The dose to the cell membrane increased 9-fold for Lu-177, 17-fold for Tb-161, and 30-fold for Pd-103 when the radionuclides are bound to the cell surface compared to their cytoplasmic counterparts. Nuclear dose differences up to 30% from literature values were observed for surface-bound sources and were particularly dependent on the computational method employed. No significant cross-irradiation contributions were seen for the pure AE-emitter Er-165; whereas the nuclear dose doubled due to cross-dose from Tb-161 and Pd-103 located on the surfaces of neighbouring cells in the micrometastasis-mimicking model, mainly from their conversion electron emissions.Significance.This study contributes newS-values for novel AE-emitters and illustrates the value of cellular dosimetry methods to investigate the optimal cellular target for AE-emitting radionuclides and their potential for treating micrometastasis.

Objective.Access to blood flow data in cerebral and retinal vascular beds is crucial for diagnosing cerebrovascular diseases. This study addresses two technological gaps: (1) simultaneous recording of vascular responses in the brain and eye by integrating transcranial Doppler ultrasonography (TCD) with Doppler optical coherence tomography (DOCT); (2) automation of DOCT image processing, including vessel segmentation, to enable quantification of flow in small retinal vessels as a proxy for cerebral microcirculation. The research evaluates vascular reactivity to CO₂ changes, measuring pulsatility index (PI), resistive index (RI), and systolic/diastolic ratio (S/D).Approach.Ten healthy volunteers were recruited. Blood flow was simultaneously recorded in the middle cerebral artery (MCA) using TCD and in retinal microarteries and microveins (diameters 50-130µm) using DOCT (95 Hz sampling rate). A novel software framework was developed for automated DOCT analysis: B-scan preprocessing, alignment, vessel segmentation, filtering, and extraction of parameters (area, lumen diameter, axial velocity component). Data were analyzed across cardiac phases (S/D) during normal breathing, apnea, and hyperventilation. Correlations were assessed statistically, comparing velocity changes and vascular indices.Main results.In the MCA, significant correlations were observed between flow velocity changes and breathing tests (increase during apnea, decrease during hyperventilation;p< 0.01). Retinal vessels showed no statistically significant correlations (p> 0.05), despite DOCT enabling precise measurement of PI, RI, and S/D comparable to TCD. Retinal vessels exhibited lower CO₂ reactivity, confirming fundamental differences in ocular and cerebral microcirculation regulation.Significance.TCD-DOCT integration enables noninvasive, synchronized vascular reactivity assessment, paving the way for studies on brain-eye microcirculation interactions. Results in healthy subjects highlight retinal vessels' uniqueness as a model but indicate limitations in directly mirroring cerebral changes, guiding future hybrid protocols. The automated framework eliminates subjective errors, facilitating scalable clinical analyses in pathologies like ischemic stroke or cerebral small vessel disease.

Objective.Rapid and accurate reconstruction of high-quality three-dimensional magnetic resonance (MR) images from undersampledk-space data with variable sampling patterns remains a challenge due to limited available information and the need to preserve rich anatomical details. Deformable image registration provides a promising solution by warping a fully-sampled reference image to align with undersampled data acquired on-board (from an image-guided treatment delivery platform like MR-LINACs). In this study, we proposed a novel registration framework-cohort-informed meta-learning (cMeta)-to enhance the accuracy and efficiency of implicit neural representations (INR) for limitedk-space data-driven patient-specific deformable image registration.Approach.cMeta-INR incorporated token-aware modulation and population-level deformation priors to guide an INR template-based meta-learning process. By encoding contextual information and leveraging cohort-informed priors, the resulting meta-learning framework enabled the INR to rapidly adapt to new registration cases with undersampledk-space data. Specifically, for the meta learning, a modulation module with token-awareness was introduced to modulate the INR template, and a pre-trained population-based registration network (KS-RegNet) was employed to generate coarse, reference deformation vector fields and latent embeddings for computing the deformation discrepancy loss and embedding similarity loss. During test-time adaptation, the INR, initialized from the meta-learned template, was efficiently fine-tuned to newk-space data with minimal iterations.Main results.Experiments were conducted on 14 abdominal and 11 cardiac 4D magnetic resonance imagings (MRIs) with 5-13 spokes. cMeta-INR outperformed state-of-the-art methods, achieving the best average (± s.d.) Dice similarity coefficients (0.778 ± 0.056 for abdominal and 0.894 ± 0.076 for cardiac data), and center-of-mass errors (3.04 ± 1.48 mm and 1.32 ± 1.02 mm, respectively), while enabling rapid test-time adaptation of only ∼35 s on an NVIDIA H100 GPU.Significance.The proposed cohort-informed meta-learning framework effectively enhanced the adaptation capabilities of INRs to individual patients under highly undersampledk-space scenarios, demonstrating strong potential for fast and accurate patient-specific deformable registration.