Physics in medicine and biology最新文献

Objective.This study aims to develop a digital twin (DT) framework to achieve adaptive proton prostate stereotactic body radiation therapy (SBRT) with fast treatment plan selection and patient-specific clinical target volume (CTV) setup uncertainty. Prostate SBRT has emerged as a leading option for external beam radiotherapy due to its effectiveness and reduced treatment duration. However, interfractional anatomy variations can impact treatment outcomes. This study seeks to address these uncertainties using DT concept to improve treatment quality.Approach. A retrospective study on two-fraction prostate proton SBRT was conducted, involving a cohort of 10 randomly selected patient cases from an institutional database (n= 43). DT-based treatment plans were developed using patient-specific CTV setup uncertainty, determined through machine learning predictions. Plans were optimized using pre-treatment CT and corrected cone-beam CT (cCBCT). The cCBCT was corrected for CT numbers and artifacts, and plan evaluation was performed using cCBCT to account for actual patient anatomy. The ProKnow scoring system was adapted to determine the optimal treatment plans.Main Results.Average CTV D98 values for original clinical and DT-based plans across 10 patients were 99.0% and 98.8%, with hot spots measuring 106.0% and 105.1%. Regarding bladder, clinical plans yielded average bladder neck V100 values of 29.6% and bladder V20.8 Gy values of 12.0cc, whereas DT-based plans showed better sparing of bladder neck with values of 14.0% and 9.5cc. Clinical and DT-based plans resulted in comparable rectum dose statistics due to SpaceOAR. Compared to clinical plans, the proposed DT-based plans improved dosimetry quality, improving plan scores ranging from 2.0 to 15.5.Significance.Our study presented a pioneering approach that leverages DT technology to enhance adaptive proton SBRT, potentially revolutionizing prostate radiotherapy to offer personalized treatment solutions using fast adaptive treatment plan selections and patient-specific setup uncertainty. This research contributes to the ongoing efforts to achieve personalized prostate radiotherapy.

Objective: The purpose of this study was to develop a robust deep learning approach trained with a small in-vivo MRI dataset for multi-label segmentation of all eight carpal bones for therapy planning and wrist dynamic analysis.

Approach: A small dataset of 15 3.0-T MRI scans from five health subjects was employed within this study. The MRI data was variable with respect to the Field Of View (FOV), wide range of image intensity, and joint pose. A two-stage segmentation pipeline using modified 3D U-Net was proposed. In the first stage, a novel architecture, introduced as Expansion Transfer Learning (ETL), cascades the use of a focused Region Of Interest (ROI) cropped around ground truth for pretraining and a subsequent transfer by an expansion to the original FOV for a primary prediction. The bounding box around the ROI generated was utilized in the second stage for high-accuracy, labeled segmentations of eight carpal bones. Different metrics including Dice Similarity Coefficient (DSC), Average Surface Distance (ASD) and Hausdorff Distance (HD) were used to evaluate performance between proposed and four state-of-the-art approaches.

Main results: With an average DSC of 87.8 %, an ASD of 0.46 mm, an average HD of 2.42mm in all datasets (96.1 %, 0.16 mm, 0.38mm in 12 datasets after exclusion criteria, respectively), the proposed approach showed an overall strongest performance than comparisons.

Significance: To our best knowledge, this is the first CNN-based multi-label segmentation approach for MRI human carpal bones. The ETL introduced in this work improved the ability to localize a small ROI in a large FOV. Overall, the interplay of a two-stage approach and ETL culminated in convincingly accurate segmentation scores despite a very small amount of image data.

Orthotopic tumor models in pre-clinical translational research are becoming increasingly popular, raising the demands on accurate tumor localization prior to irradiation. This task remains challenging both in x-ray and proton computed tomography (xCT and pCT, respectively), due to the limited contrast of tumor tissue compared to the surrounding tissue. We investigate the feasibility of gadolinium oxide nanoparticles as a multimodal contrast enhancement agent for both imaging modalities. We performed proton radiographies at the experimental room of the Trento Proton Therapy Center using a MiniPIX-Timepix detector and dispersions of gadolinium oxide nanoparticles in sunflower oil with mass fractions up to 8wt%. To determine the minimum nanoparticle concentration required for the detectability of small structures, pCT images of a cylindrical water phantom with cavities of varying gadolinium oxide concentration were simulated using a dedicated FLUKA Monte Carlo framework. These findings are complemented by simulating pCT at dose levels from 80 mGy to 320 mGy of artificially modified murine xCT data, mimicking different levels of gadolinium oxide accumulation inside a fictitious tumor volume. To compare the results obtained for proton imaging to x-ray imaging, cone-beam CT images of a cylindrical PMMA phantom with cavities of dispersions of oil and gadolinium oxide nanoparticles with mass fractions up to 8wt% were acquired at a commercial pre-clinical irradiation setup. For proton radiography, considerable contrast enhancement was found for a mass fraction of 4wt%. Slightly lower values were found for the simulated pCT images at imaging doses below 200 mGy. In contrast, full detectability of small gadolinium oxide loaded structures in xCT at comparable imaging dose is already achieved for 0.5wt%. Achieving such concentrations required for pCT imaging inside a tumor volume inin-vivoexperiments may be challenging, yet it might be feasible using different targeting and/or injection strategies.

Objective.Time resolution is crucial in positron emission tomography (PET) to enhance the signal-to-noise ratio and image quality. Moreover, high sensitivity requires long scintillators, which can cause distortions in the reconstructed images due to parallax effects. This study evaluates the performance of a time-of-flight (TOF)-PET module that makes use of a single-side readout of a4×43.1×3.1×15mm³LYSO:Ce matrix with an array of4×4silicon photomultipliers (SiPMs) and a light guide to extract high-resolution TOF and depth of interaction (DOI) information.Approach.This study assesses the performance of the detector prototype using the commercially available TOFPET2 ASIC and SiPMs from various producers. DOI and TOF performance are compared to results using custom-made NINO 32-chip based electronics.Main results.Using a Broadcom NUV-MT array, the detector module read out by the TOFPET2 ASIC demonstrates a DOI resolution of 2.6 ± 0.2 mm full width at half maximum (FWHM) and a coincidence time resolution (CTR) of 216 ± 6 ps FWHM. When read out using the NINO 32-chip based electronics, the same module achieves a DOI resolution of 2.5 ± 0.2 mm and a CTR of 170 ± 5 ps.Significance.The prototype module, read out by commercial electronics and using state-of-the-art SiPMs, achieves a DOI performance comparable to that obtained with custom-made electronics and a CTR of around 200 ps. This approach is scalable to thousands of channels, with only a deterioration in timing resolution compared to the custom-made electronics, which achieve a CTR of 140 ps using a standard non-DOI module.

Objective.A new nuclear Overhauser enhancement (NOE)-mediated saturation transfer MRI signal at -1.6 ppm, potentially from choline phospholipids and termed NOE(-1.6), has been reported in biological tissues at high magnetic fields. This signal shows promise for detecting brain tumors and strokes. However, its proximity to the water peak and low signal-to-noise ratio makes accurate quantification challenging, especially at low fields, due to the difficulty in separating it from direct water saturation and other confounding signals. This study proposes using a machine learning (ML) method to address this challenge.Approach.The ML model was trained on a partially synthetic chemical exchange saturation transfer dataset with a curriculum learning denoising approach. The accuracy of our method in quantifying NOE(-1.6) was validated using tissue-mimicking data from Bloch simulations providing ground truth, with subsequent application to an animal tumor model at 4.7 T. The predictions from the proposed ML method were compared with outcomes from traditional Lorentzian fit and ML models trained on other data types, including measured and fully simulated data.Main results.Our tissue-mimicking validation suggests that our method offers superior accuracy compared to all other methods. The results from animal experiments show that our method, despite variations in training data size or simulation models, produces predictions within a narrower range than the ML method trained on other data types.Significance.The ML method proposed in this work significantly enhances the accuracy and robustness of quantifying NOE(-1.6), thereby expanding the potential for applications of this novel molecular imaging mechanism in low-field environments.

Objective.Positron emission tomography (PET) has become an important clinical modality, but it is limited to imaging the annihilation radiation from positron-electron collisions. Recently, PET imaging with⁸⁹Zr, which has a half-life of 3 d, has attracted much attention in immuno-PET to visualize immune cells and cancer cells by targeting specific antibodies on the cell surface. However,⁸⁹Zr emits a single gamma ray at 909 keV four times more frequently than positrons, causing image quality (IQ) degradation in conventional PET. To overcome this drawback, use of such single gamma rays for imaging was previously proposed as whole gamma imaging (WGI). In WGI, a single gamma ray is detected by Compton imaging; by inserting a scatter detector ring inside the PET ring, WGI can realize both PET imaging and Compton imaging in one modality. A prototype for WGI was developed and Compton imaging of a mouse after intravenous administration of⁸⁹Zr oxalate was demonstrated. However, the Compton imaging of the single gamma ray still presented a challenge due to its low IQ compared to PET.Approach.In this study, the scatter detector insert of the earlier WGI prototype was redesigned with the aim of improving Compton imaging performance. The new prototype produced WGI images by additive averaging of PET and Compton images after optimizing the ratio of each iteration in the image reconstruction. WGI IQ was then evaluated using the NEMA NU4 IQ phantom, and a tumor-burdened mouse was imaged with WGI up to 12 d after⁸⁹Zr labeled antibody injection.Main results.Consequently, the Compton imaging performance was improved by lowering the angular resolution measure from 6.7 degrees to 6.4 degrees and the sensitivity from 0.11% to 0.18% compared to the previous prototype WGI. The phantom images with WGI showed a 15% reduction in noise and a 3% increase in contrast recovery under low-statistical conditions compared to images reconstructed by PET data alone.Significance. In-vivomouse imaging with the new prototype WGI was successfully performed. This successful imaging leads to the expectation that future whole-body WGI imaging will enable more sensitive and better quantitative⁸⁹Zr antigen-antibody reaction imaging to be obtained.

Objective. Position-sensitive silicon photomultipliers (PS-SiPMs) are promising photodetectors for ultra-high spatial resolution small-animal positron emission tomography (PET) scanners. This paper evaluated the performance of the latest generation of linearly-graded SiPMs (LG-SiPMs), a type of PS-SiPM, for ultra-high spatial resolution PET applications using LYSO arrays from two vendors.Approach. Two dual-ended readout detectors were developed by coupling LG-SiPMs to both ends of the two LYSO arrays. Each LG-SiPM has an active area of 9.8 × 9.8 mm². Both LYSO arrays consist of 20 × 20 arrays of 0.44 × 0.44 × 20 mm³polished LYSOs with a pitch of 0.5 mm. The performance of the two detectors was compared in terms of flood histogram, energy resolution, timing resolution, and depth-of-interaction (DOI) resolutions.Main results. Flood histograms showed clear identification of all LYSO elements except for some edge crystals due to the larger size of the LYSO arrays compared to the active area of the LG-SiPMs and the misalignment between LG-SiPMs and LYSO arrays in the assembled detectors. At a bias voltage of 37.0 V, the detectors utilizing the Tianle LYSO array and EBO LYSO array provided energy resolutions of 17.5 ± 2.2 and 18.6 ± 2.0%, timing resolutions of 0.75 ± 0.03 and 0.78 ± 0.03 ns, and DOI resolutions of 2.16 ± 0.15 and 2.31 ± 0.12 mm, respectively.Significance. The results presented in this paper demonstrate that the new generation LG-SiPMs are promising photodetectors for ultra-high spatial resolution small-animal PET scanner applications.

Objective. Clinical dosimetry in the presence of a 1.5 T magnetic field is challenging, let alone in case small fields are involved. The scope of this study is to determine a set of relevant correction factors for a variety of MR-compatible detectors with emphasis on small fields. Two dosimetry formalisms adopted from the literature are considered.Approach. Six small-cavity ionization chambers (from three manufacturers), four active solid-state detectors and a thermoluminescence dosimeter microcube were modeled in the EGSnrc Monte Carlo code. Phase space files for field sizes down to 1 × 1 cm²of the Unity 1.5 T/7 MV MR-linac (Elekta, UK) were used as source models. Simulations were performed to calculate thekQB,QfB,f(also known askB,Q),kQmsrB,fmsrandkQclin,QmsrB,fclin,fmsrrelevant to two different dosimetry formalisms. Two detector orientations with respect to the magnetic field were considered. Moreover, the effect of the ionization chamber's stem length (a construction parameter) on the correction factor was investigated. Simulations were also carried out to determine whether correction factors obtained in water can be applied in dosimetry procedures involving water-equivalent solid phantoms.Main results. Under thekQB,QfB,f-based formalism, the required corrections to ionization chamber responses did not exceed 1.5% even for the smallest field size considered. A much wider range ofkQB,QfB,fvalues was obtained for the active solid-state detectors included in the simulations. This is the first study to reportkQclin,QmsrB,fclin,fmsrvalues for ionization chambers. The impact of the stem on correction factors is not significant for lengths ⩾0.75 cm. Correction factors determined in water are also valid in dosimetry protocols employing solid phantoms.Significance. This work substantially expands the range of available detectors that can be used in small field dosimetry, enabling more options for commissioning, beam modeling and quality assurance procedures in 1.5 T MR-Linacs. However, more studies are needed to establish a complete and reliable dataset.

Objective. To study the effect of dose-rate in the time evolution of chemical yields produced in pure water versus a cellular-like environment for FLASH radiotherapy research.Approach.A version of TOPAS-nBio with Tau-Leaping algorithm was used to simulate the homogenous chemistry stage of water radiolysis using three chemical models: (1) liquid water model that considered scavenging ofe_aq^-, H^•by dissolved oxygen; (2) Michaels & Hunt model that considered scavenging of^•OH,e_aq^‒, and H^•by biomolecules existing in cellular environment; (3) Wardman model that considered model 2) and the non-enzymatic antioxidant glutathione (GSH). H₂O₂concentrations at conventional and FLASH dose-rates were compared with published measurements. Model 3) was used to estimate DNA single-strand break (SSB) yields and compared with published data. SSBs were estimated from simulated yields of DNA hydrogen abstraction and attenuation factors to account for the scavenging capacity of the medium. The simulation setup consisted of monoenergetic protons (100 MeV) delivered in pulses at conventional (0.2857Gy s^-1) and FLASH (500Gy s^-1) dose rates. Dose varied from 5-20 Gy, and oxygen concentration from 10µM-1 mM.Main Results.At the steady state, for model (1), H₂O₂concentration differed by 81.5%± 4.0% between FLASH and conventional dose-rates. For models (2) and (3) the differences were within 8.0%± 4.8%, and calculated SSB yields agreed with published data within 3.8%± 1.2%. A maximum oxygen concentration difference of 60% and 50% for models (2) and (3) between conventional and FLASH dose-rates was found between 2 × 10⁶and 9 × 10¹³ps for 20 Gy of absorbed dose.Significance.The findings highlight the importance of developing more advanced cellular models to account for both the chemical and biological factors that comprise the FLASH effect. It was found that differences between pure water and cellular environment models were significant and extrapolating results between the two should be avoided. Observed differences call for further experimental investigation.

Objective: The study aims to systematically characterize the effect of CT parameter variations on images and lung radiomic and deep features, and to evaluate the ability of different image harmonization methods to mitigate the observed variations.

Approach: A retrospective in-house sinogram dataset of 100 low-dose chest CT scans was reconstructed by varying radiation dose (100%, 25%, 10%) and reconstruction kernels (smooth, medium, sharp). A set of image processing, convolutional neural network (CNNs), and generative adversarial network-based (GANs) methods were trained to harmonize all image conditions to a reference condition (100% dose, medium kernel). Harmonized scans were evaluated for image similarity using peak signal-to-noise ratio (PSNR), structural similarity index measure (SSIM), and learned perceptual image patch similarity (LPIPS), and for the reproducibility of radiomic and deep features using concordance correlation coefficient (CCC).

Main results: CNNs consistently yielded higher image similarity metrics amongst others; for Sharp/10%, which exhibited the poorest visual similarity, PSNR increased from a mean ± CI of 17.763 ± 0.492 to 31.925 ± 0.571, SSIM from 0.219 ± 0.009 to 0.754 ± 0.017, and LPIPS decreased from 0.490 ± 0.005 to 0.275 ± 0.016. Texture-based radiomic features exhibited a greater degree of variability across conditions, i.e. a CCC of 0.500 ± 0.332, compared to intensity-based features (0.972 ± 0.045). GANs achieved the highest CCC (0.969 ± 0.009 for radiomic and 0.841 ± 0.070 for deep features) amongst others. Convolutional neural networks are suitable if downstream applications necessitate visual interpretation of images, whereas generative adversarial networks are better alternatives for generating reproducible quantitative image features needed for machine learning applications.

Significance: Understanding the efficacy of harmonization in addressing multi-parameter variability is crucial for optimizing diagnostic accuracy and a critical step toward building generalizable models suitable for clinical use.