Medical physics最新文献

Background: Estimation of the radiation dose to the fetus is essential for the assessment of radiation risks and benefits to pregnant patients undergoing radiological examinations. During the past decade, the global twinning rate has soared resulting from embryo assistance and increased delivery age. However, to the best of our knowledge, radiation dosimetry in multiple pregnancies from radiological imaging has never been reported before.

Purpose: The purpose of this study is to develop personalized computational models for twin fetuses based on clinical CT images of real pregnant patients and to estimate personalized radiation doses for twin fetuses from abdominal/pelvic CT examinations.

Methods: Personalized computational phantoms representing pregnant females with twins at the second and third trimesters were constructed based on CT images of two pregnant patients. Monte Carlo calculations were performed using the MCNP transport code and three validated CT scanners to estimate the radiation dose of twin fetuses during abdominal and pelvic CT examinations.

Results: The absorbed fetal organ dose was calculated and compared between twins. For the same patient, the absolute difference in fetal organ dose between twins varies between 0.63% and 39.64% with an average value of 12.85%. The estimated total-body dose differences for twin fetuses were 11.55% and 7.51%, respectively, for pregnant patients at 22 and 30 weeks gestational age.

Conclusion: The variations of body weight and organ mass affect the absorbed dose of twin fetuses. Personalized computational models provide more accurate fetal radiation dosimetry estimates for pregnant patients with twins. This work also contributes to a better understanding of model-induced uncertainties in external radiation dosimetry for the developing fetus.

Background: Carbon-ion beam radiotherapy offers substantial physical and biological advantages due to its distinct Bragg peak (BP) depth dose distribution and higher linear energy transfer (LET) in the peak region that enhances its efficacy in tumor eradication compared to x-ray beams. Porous structures, such as those found in lung and lung-equivalent tissues, unfortunately, introduce significant uncertainties in both dose and LET distributions, which current treatment planning systems (TPS) inadequately address.

Purpose: This study aims to investigate the effects of porous lung-equivalent structures on LET distribution using Monte Carlo (MC) simulations and CR-39 measurements. It seeks to understand how porous structures influence LET spectra and dose-averaged LET (LET_d) in carbon-ion beams.

Methods: A Gammex LN300 phantom and a binary voxel virtual phantom composed of water and air were used to represent lung-equivalent tissues for measurements and MC simulations. LET spectra measured with CR-39 at different depths within the LN300 slabs were compared with MC-calculated LET_d distributions. The impact of porous structures on dose and LET_d distributions was evaluated using various beam configurations, including single-beam and multi-beam setups. Additionally, a convolution method with modulation power (P_mod) was proposed to improve LET_d prediction in porous media.

Results: The study demonstrated that porous structures broaden both the dose and LET_d distributions, especially around the BP region. Multiple beam angles helped mitigate dose degradation but did not resolve discrepancies in the LET_d distributions. Compared with calculation results based on CT images, intensity-modulated particle therapy (IMPT) using a distal LET_d patching method in porous structure increased the median LET_d in the target from 67.2 to 69.6 keV/µm, and the minimum LET_d from 51.5 to 58.0 keV/µm, respectively. Moreover, to improve the prediction of LET_d in porous structures, analytical convolution-based predictions showed good agreement with the MC simulations, with mean LET_d deviations of -1.9% ± 1.6% in the plateau, -3.1% ± 4.9% in the BP, and -1.1% ± 7.7% in the tail region.

Conclusions: Porous lung-equivalent structures significantly affect LET_d distributions in carbon-ion therapy, as confirmed by both CR-39 measurements and MC simulations. IMPT with LET_d optimization may be more impacted by porous structures in terms of median and minimum LET_d values within the target. The Gaussian convolution function shows promise for enhancing LET_d calculation accuracy, but further validation in anatomically complex models is needed to assess its clinical feasibility.

Background: In fluoroscopy, particularly fluoroscopically guided interventions (FGIs), accurate estimation of peak skin dose (PSD) is crucial for identifying potential radiation-induced skin injuries. Most current methodologies for PSD calculation methods rely on analytical methods, which may introduce uncertainty due to their limited consideration of the complexities of x-ray beam conditions, patient geometry, and positioning. Methods based on full Monte Carlo (MC) simulations can enhance accuracy, but their practical application is limited due to the intensive requirement of computational resources and time.

Purpose: We aimed to develop a novel method that combines MC simulation with a noise reduction technique to calculate PSD, as well as skin dose distributions, more efficiently and accurately. The goal was to overcome the limitations of current methods, providing a more practical solution for clinical and academic use.

Methods: Our method to calculate the PSD and skin dose distributions consists of two steps of rough MC simulation and iterative noise reduction. The performance of the methodology was demonstrated for six fluoroscopy scenarios, with results compared against those from full MC simulation with high particle history, which is considered a gold standard for radiation dosimetry relative to conventional analytical methods.

Results: Our method was demonstrated for various fluoroscopy scenarios, and the result showed that the iterative noise reduction procedure successfully estimates PSD and skin dose distribution for rough MC simulations with a maximum dose statistical error of up to 20%. For successful dose estimations, PSD discrepancies from the values obtained by full MC simulation were within 3%, and voxel-wise dose differences in skin dose distributions were less than 10% of the average skin dose. The computation time of our method was on the order of a few seconds on a personal computer, which is estimated to be at least 10⁴ times faster than full MC simulation when using the same computing resources.

Conclusion: Our method rapidly and accurately calculates PSD and skin dose distribution, making it a useful tool for research and clinical applications. The planned integration of our method into the National Cancer Institute Dosimetry System for Radiography and Fluoroscopy (NCIRF) will enhance accessibility. Additionally, future upgrades of NCIRF will include a comprehensive phantom library and pregnant phantoms that will enable our method to account for patient-specific body shapes, further improving the accuracy and personalization in dose assessments.

Background: Magnetic resonance imaging (MRI) is a highly sensitive modality for diagnosing breast cancer, providing an expanding range of clinical usages that are crucial for the care of women at elevated risk of breast cancer development. Segmentation of the whole breast and fibroglandular tissue (FGT), used to evaluate breast cancer risk, is often manually delineated by radiologists in clinical practice. In this paper, we aim to substitute handcrafted breast density segmentation and categorization. The traditional fuzzy C-means (FCM) enable automatic segmentation but may be susceptible to heterogeneity or sparse FGT distribution in MRI.

Purpose: We develop a new automated technique for the segmentation of whole breast and FGT for the coronal-view MRI.

Methods: We propose a Chan-Vese (CV) aided FCM segmentation approach for estimating the FGT in the whole breast using fat-suppressed (FS) precontrast T1-weighted breast MRI. We present a methodology pipeline comprising region-of-interest (ROI) extraction, nonparametric non-uniform intensity normalization N4 algorithm-based intensity inhomogeneity correction, skin-layer extraction, and then whole breast and FGT segmentation. Our approach involves the FCM algorithm to assign membership degrees to pixels, distinguishing FGT regions from surrounding adipose tissues by assessing their probability of belonging to specific FGT regions, and subsequently, the region-based active contour CV model leverages these membership degrees to direct contour evolution and enhance segmentation boundaries. The proposed method adeptly tackles common challenges in MRI, including blurred edges, low contrast, and intensity inhomogeneity, with efficiency.

Results: We evaluated our approach on the Duke Breast Cancer MRI data (DBCM-data) and achieved good segmentation accuracy in terms of Dice similarity coefficient (DSC), Intersection-over-Union (IoU), and Sensitivity (SEN). Our method demonstrates significant accuracy, achieving a DSC (%) of 93.2 ± 3.3 and 84.1 ± 4.9, IoU (%) of 86.4 ± 3.5 and 73.2 ± 5.1, and SEN 87.3 ± 4.1 and 76.7 ± 4.1 for the segmentations of whole breast and FGT, respectively.

Conclusion: Our results demonstrated that the CV-aided FCM approach significantly outperformed the existing methods and resulted in significantly more accurate whole breast and FGT segmentation in MRI data.

Background: Kidney ultrasound (US) image segmentation is one of the key steps in computer-aided diagnosis and treatment planning of kidney diseases. Recently, deep learning (DL) technology has demonstrated promising prospects in automatic kidney US segmentation. However, due to the poor quality, particularly the weak boundaries in kidney US imaging, obtaining accurate annotations for DL-based segmentation methods remain a challenging and time-consuming task. This issue can hinder the application of data-hungry deep learning methods.

Purpose: In this paper, we explore a novel cross-modal data augmentation method aimed at enhancing the performance of DL-based segmentation networks on the limited labeled kidney US dataset.

Methods: In particular, we adopt a novel method based on contrastive unpaired translation network (CUT) to obtain simulated labeled kidney US images at a low cost from labeled abdomen computed tomography (CT) data and unlabeled kidney US images. To effectively improve the segmentation network performance, we propose an instance-weighting training strategy that simultaneously captures useful information from both the simulated and real labeled kidney US images. We trained our generative networks on a dataset comprising 4418 labeled CT slices and 4594 unlabeled US images. For segmentation network, we used a dataset consisting of 4594 simulated and 100 real kidney US images for training, 20 images for validation, and 169 real images for testing. We compared the performance of our method to several state-of-the-art approaches using the Wilcoxon signed-rank test, and applied the Bonferroni method for multiple comparison correction.

Results: The experimental results show that we can synthesize accurate labeled kidney US images with a Fréchet inception distance of 52.52. Moreover, the proposed method achieves a segmentation accuracy of 0.9360 ± 0.0398 for U-Net on normal kidney US images, and 0.7719 ± 0.2449 on the abnormal dataset, as measured by the dice similarity coefficient. When compared to other training strategies, the proposed method demonstrated statistically significant superiority, with all p-values being less than 0.01.

Conclusions: The proposed method can effectively improve the accuracy and generalization ability of kidney US image segmentation models with limited annotated training data.

Background: Ultra-high dose rate radiotherapy (UHDR) delivers therapeutic doses at rates >40 Gy/s in a fraction of a second, aiming to enhance the therapeutic ratio through the FLASH effect. The substantial increase in UHDR beam current poses serious challenges for conventional active dosimeters. Integrating current transformers (ICT) offer a nondestructive solution for accurate monitoring, enabling the type of fast transient readout that will be crucial for UHDR treatment verification.

Purpose: The aim of this study is to build and characterize a clinically deployable ICT system and to develop an accurate calibration methodology for absolute charge determination that is traceable to primary electrical standards.

Methods: The ICT was constructed from a Super MuMetal^® toroid, and its secondary winding was made from 50 Ω coaxial cable. A 3D-printed case with an internal conductive coating shields the toroid assembly from interference. The ICT readout involves a custom differential amplifier and a commercial flash analog-to-digital converter. The system was calibrated using a bespoke sub-µs current pulser in the range of (2 to 16) mA, which itself is traceable to electrical standards via an in-house built electrometer with a calibrated feedback capacitor. The performance of the ICT was evaluated as a milliampere-scale beam monitor against concurrent absorbed dose graphite calorimetry irradiation measurements acquired on a specially tuned medical accelerator for UHDR delivery.

Results: The ICT responses for nominal test pulses, generated by a function generator and a current pulser, exhibited accurate reproduction of rise and fall times within the 1 ns sampling frequency. A systematic droop effect of 0.6(1)%/µs was observed but is accounted for through the calibration chain. The calibration of the current pulser exhibited a repeatability typically better than 0.05%, with a slowly-varying leakage that can be subtracted using a linear regression of the leakage current. The ICT charge calibration demonstrated a repeatability in the range of (0.5 to < 0.05)% for charge per pulse values in the range of (0.5 to 50) nC, respectively. The ICT response showed a strong linear relationship (adj-R² = 0.99997) to charge per pulse. The in-beam comparison with a graphite calorimeter demonstrated the effectiveness of the ICT as an online beam monitor, independent of pulse repetition frequency in the range of (25 to 200) Hz, reducing the mean excess (i.e., independent of accelerator output) calorimeter variation to 0.3% (0.1% standard error on the mean).

Conclusions: This work demonstrates the feasibility of accurately calibrating the ICT in terms of absolute charge and applying it as a clinically deployable monitoring system of mA-scale electron beams delivered by a medical accelerator.

Background: Four-dimensional dynamic contrast-enhanced breast CT (4D DCE-bCT) offers promising high-resolution spatial and temporal imaging capabilities for the characterization and monitoring of breast tumors. However, the optimal combination of parameters for iodine quantification in image space remains to be determined.

Purpose: This study aims to optimize a dedicated bCT system to perform long dynamic contrast-enhanced scans with high spatio-temporal resolution while maintaining a reasonable radiation dose.

Methods: Our protocol includes the acquisition of a high-quality prior image that is reconstructed with a polychromatic iterative algorithm (IMPACT). The acquisition of the post-contrast sequence is continuous but sparse and these images are reconstructed using prior image constrained compressed sensing (PICCS). A four-step optimization process is performed using images of a physical phantom. First, the optimal tube current is selected by taking the noise level into account. Second, the optimal number of angles is selected based on the absence of streak artifacts. Third, the number of iterations in IMPACT is specified at the lowest value that achieves the highest spatial resolution. Finally, the number of iterations in PICCS is determined based on the quantitative accuracy of a range of iodine concentrations.

Results: When a high-quality prior image is available, the imaging of post-contrast images can be performed using just 40 projection angles with a tube current of 32 mA. The noise level in the post-contrast images is inherited from the prior image and no streak artifacts are visible. Mean difference between the linear attenuation coefficients of samples containing iodine reconstructed with IMPACT using all 360 projections and PICCS using 40 projections is 0.0004 ${\mathrm{mm}}^{-1}$ at most. The spatial resolution of images reconstructed with PICCS is lower than the one of IMPACT images and is concentration dependent. The cut-off frequency at 10% modulation transfer function drops from 1.55 ${\mathrm{mm}}^{-1}$ in the prior image to 0.9 ${\mathrm{mm}}^{-1}$ when the target with the largest concentration is evaluated. The total mean glandular dose of the protocol does not exceed 22.5 mGy.

Conclusions: This study found the optimal acquisition and reconstruction parameters for a low-dose dynamic contrast-enhanced bCT protocol. The numerical accuracy of the proposed protocol was ensured by performing a physical phantom study.

Background: Most attention-based networks fall short in effectively integrating spatial and channel-wise information across different scales, which results in suboptimal performance for segmenting coronary vessels in x-ray digital subtraction angiography (DSA) images. This limitation becomes particularly evident when attempting to identify tiny sub-branches.

Purpose: To address this limitation, a multi-scale dual attention embedded network (named MDA-Net) is proposed to consolidate contextual spatial and channel information across contiguous levels and scales.

Methods: MDA-Net employs five cascaded double-convolution blocks within its encoder to adeptly extract multi-scale features. It incorporates skip connections that facilitate the retention of low-level feature details throughout the decoding phase, thereby enhancing the reconstruction of detailed image information. Furthermore, MDA modules, which take in features from neighboring scales and hierarchical levels, are tasked with discerning subtle distinctions between foreground elements, such as coronary vessels of diverse morphologies and dimensions, and the complex background, which includes structures like catheters or other tissues with analogous intensities. To sharpen the segmentation accuracy, the network utilizes a composite loss function that integrates intersection over union (IoU) loss with binary cross-entropy loss, ensuring the precision of the segmentation outcomes and maintaining an equilibrium between positive and negative classifications.

Results: Experimental results demonstrate that MDA-Net not only performs more robustly and effectively on DSA images under various image conditions, but also achieves significant advantages over state-of-the-art methods, achieving the optimal scores in terms of IoU, Dice, accuracy, and Hausdorff distance 95%.

Conclusions: MDA-Net has high robustness for coronary vessels segmentation, providing an active strategy for early diagnosis of cardiovascular diseases. The code is publicly available at https://github.com/30410B/MDA-Net.git.