Med Phys [JOURNAL]
Sun
200 papers
Unsupervised deep learning-based 4DCT deformable image registration and carbon-ion 4D dose calculation.
An X, Wang J, Zhang J … +9 more , Liu Y, Abuduxiku M, Chai N, Luo Y, Wu W, Hu Y, Hei D, Li Q, He P
Med Phys · 2026 Mar · PMID 41833529 · Publisher ↗
BACKGROUND: Accurate four-dimensional dose calculation (4DDC) is essential for carbon-ion lung radiotherapy and relies on deformable image registration (DIR). However, conventional DIR methods are computationally intensi...
BACKGROUND: Accurate four-dimensional dose calculation (4DDC) is essential for carbon-ion lung radiotherapy and relies on deformable image registration (DIR). However, conventional DIR methods are computationally intensive, hindering the implementation of online adaptive workflows. PURPOSE: This study investigates the feasibility and efficacy of unsupervised deep learning-based DIR models, TransMatch and VoxelMorph, in accelerating clinical lung four-dimensional computed tomography (4DCT) registration and facilitating accurate carbon-ion 4D dose calculation. METHODS: A total of 150 clinical lung 4DCT datasets were utilized (120 for training, 20 for validation, and 10 for testing), with a conventional B-spline method serving as the baseline. Registration accuracy was evaluated using the Mean Absolute Error (MAE), Dice Similarity Coefficient (DSC), 95th percentile Hausdorff Distance (HD95), and Jacobian determinant (|J|). Carbon-ion 4D dose distributions were accumulated using the generated deformation vector fields (DVFs). Dosimetric impacts on the gross tumor volume (GTV) and organs at risk (OARs) were quantified using Dose-Volume Histogram (DVH) metrics under a gating with 6× rescanning scenario. RESULTS: TransMatch and VoxelMorph achieved superior registration accuracy with lower MAE, mean DSC > 0.97, and HD95 < 2.5 mm. In DVF analysis, TransMatch and VoxelMorph showed negligible folding rates (<0.02%) and significantly higher Jacobian standard deviations than the B-spline method, indicating superior capability in capturing fine local deformations. Dosimetrically, differences in GTV and OARs metrics between the deep learning and B-spline methods were less than 2% of the prescription dose, falling within clinically acceptable tolerances. Crucially, TransMatch and VoxelMorph models achieved sub-second registration times (<1 s), whereas the conventional B-spline method required more than 10 min. CONCLUSIONS: TransMatch and VoxelMorph achieve geometric and dosimetric accuracy comparable to the conventional B-spline method for carbon-ion lung radiotherapy while offering substantially higher computational speed, highlighting their potential for real-time adaptive carbon-ion therapy.
Experimental characterization and mitigation of scatter induced spectral contamination in photon counting CBCT for radiotherapy.
Sabounchi R, Pyakurel U, Shukla A … +2 more , Bayat F, Altunbas C
Med Phys · 2026 Mar · PMID 41833525 · Publisher ↗
BACKGROUND: Photon counting detectors (PCDs) have the potential to enable spectral cone beam computed tomography (CBCT) in radiotherapy, which may improve tissue visualization, tissue characterization, and dose calculati...
BACKGROUND: Photon counting detectors (PCDs) have the potential to enable spectral cone beam computed tomography (CBCT) in radiotherapy, which may improve tissue visualization, tissue characterization, and dose calculation accuracy. However, spectral imaging with PCDs such as material decomposition require highly quantitative raw data that accurately reflects X-ray attenuation through the imaged object. PURPOSE: In CBCT, X-ray scatter is the primary cause of degraded raw-data fidelity. This study experimentally characterizes the effect of scatter on PCD counts and energy spectra and investigates a hardware-driven approach to mitigate scatter contamination in photon counting CBCT (pcCBCT). METHODS: A photon-counting detector with a cadmium telluride (CdTe) sensor and two adjustable energy thresholds was integrated into a benchtop X-ray imaging system mimicking the CBCT geometry of C-arm linacs. Scatter contamination as a function of energy was characterized using phantoms of varying thicknesses and the beam-stop method. Bias in the detected energy spectrum due to scatter, as well as the scatter-to-primary ratio (SPR) as a function of detected energy, were quantified. The effect of increased X-ray fluence due to scatter on pulse pile up was measured under X-ray fluence conditions relevant to image-guided radiotherapy. To mitigate scatter, all experiments were repeated using a 2D antiscatter grid (ASG) prototype, which also served as a reference for quantifying scatter-induced bias in contaminated raw data. The effect of scatter contamination on the energy-specific CT number accuracy was evaluated in pcCBCT scans acquired with two energy bins. RESULTS: Scatter bias increased counts by up to 96% in the 25-50 keV range in the normalized energy spectra. The mean energy of the spectrum shifted to lower energies by up to 8 keV as a result of scatter bias. Average scatter to primary ratio (SPR) was up to a factor of 5 higher at 35 keV than the SPR at 100 keV. SPR reached 4 in the 25-60 keV range. The incremental pulse pile up related count underestimation attributable to scatter-induced fluence was < 3%. The 2D ASG reduced SPR by 90%-94% across all energies and phantom configurations. The efficacy of 2D ASG in SPR reduction varied only 3% across 25-120 keV range. Without scatter mitigation, mean CT-number loss was 453 and 390 HU in the 25-60 keV and 60-120 keV bins, respectively. The 2D ASG reduced CT number loss to 34 and 22 HU in the respective energy bins. CONCLUSIONS: Scatter fraction is substantially higher at lower energies and biases the energy spectrum toward lower energies under CBCT imaging scenarios relevant to radiotherapy. Consequently, CT number degradation is significantly greater at lower energies, which may challenge the quantitative utilization of low-energy data in future pcCBCT systems and spectral imaging applications. The effect of scatter-induced fluence increase on pulse pile up is deemed small. The 2D ASG effectively reduced scatter contamination across the entire energy spectrum and improved the fidelity of raw data for pcCBCT imaging.
A universal framework for IMRT dose prediction.
Wang Q, Chen M, Zhu Y … +6 more , Kazemimoghadam M, Zhang K, Yang Z, Jiang H, Gu X, Lu W
Med Phys · 2026 Mar · PMID 41833521 · Full text
BACKGROUND: Dose prediction has great potential in improving plan quality and efficiency by estimating optimal dose distribution. However, most existing deep learning (DL) based dose prediction models for intensity-modul...
BACKGROUND: Dose prediction has great potential in improving plan quality and efficiency by estimating optimal dose distribution. However, most existing deep learning (DL) based dose prediction models for intensity-modulated radiation therapy (IMRT) have been primarily developed under simplified conditions, such as fixed beam configuration and/or disease site. These constraints limit the generalizability and clinical usability of such models across the diverse scenarios encountered in real-world practice. PURPOSE: We proposed a DL-based universal dose prediction model, named UniDose, designed to accommodate a wide range of disease sites and support diverse clinical scenarios, especially for IMRT treatment plans with arbitrary beam configurations. METHODS: UniDose is built on a customized nnU-Net framework, adapted into an image-to-image mapping network tailored for 3D dose prediction and trained using the Huber loss. The network takes three generalized input channels: a normalized prescription dose map that encodes planning goals for the target, a weighted avoidance mask that consolidates multiple organs at risk (OARs) and body structures into a single channel with clinical relevance-based voxel weights, and a beam trace image that captures beam configuration using a non-modulated, cumulative dose approximation generated via a ray-tracing based algorithm. The model was trained, validated and tested on a heterogeneous dataset of 871 patients encompassing 25 disease sites and a wide spectrum of prescription doses and beam configurations. To assess the deliverability of the predicted dose, we incorporated a reference-based in-house optimization engine into the UniDose framework to generate feasible plans constrained by machine limitations. Model performance was evaluated by comparing predicted doses, optimized doses, and clinical plans using gamma passing rate (GPR) with a 3%/2 mm criteria and 10% lower dose threshold and dose-volume histogram (DVH) metrics. RESULTS: The UniDose predictions achieved an average GPR of 92.36% compared to the optimized doses and demonstrated strong DVH consistency. The average GPR between predicted and clinical doses was 86.13%. DVH comparisons showed that the predictions and the optimized dose achieved improved OAR sparing while maintaining comparable target coverage relative to clinical dose, particularly in prostate, liver, and brain cases. Case studies across six disease sites with variable beam configurations further confirmed that the predicted and optimized doses exhibited similar dose deposition patterns along beam paths, suggesting that the predicted dose is physically feasible and approachable following dose optimization. Additionally, adjusting voxel weights in the avoidance input channel enabled flexible trade-offs between OAR sparing and target coverage, supporting patient-specific treatment planning. CONCLUSIONS: UniDose demonstrates strong potential as a universal DL-based dose prediction framework capable of generalizing across diverse disease sites and beam configurations. By combining a generalized input design, robust network customization, and integration with a reference-guided optimization engine, UniDose generates physically feasible dose predictions and allows efficient user interaction through adjustable input conditions.
Energy dependence of Farmer-type ionization chambers used in kilovoltage x-ray beam reference dosimetry.
Tyler M, Butler D, Duncan M … +1 more , McNamara J
Med Phys · 2026 Mar · PMID 41833510 · Publisher ↗
BACKGROUND: Absorbed dose determination for therapeutic kilovoltage (kV) x rays is performed following international recommendations to ensure consistency in practice and equivalence in dose delivery for patients worldwi...
BACKGROUND: Absorbed dose determination for therapeutic kilovoltage (kV) x rays is performed following international recommendations to ensure consistency in practice and equivalence in dose delivery for patients worldwide. These recommendations include limits on the variation of ionization chamber calibration coefficients with beam quality. To date, data on energy-dependence have only been published for a limited set of ionization chambers and kV beams, and there are limited guidelines regarding which models of cylindrical ionization chambers are suitable for kV reference dosimetry. PURPOSE: To provide cylindrical ionization chamber energy-dependence data with enough beam qualities and chambers to assist medical physicists to assess the compliance of chamber models with protocol recommendations regarding chamber characteristics. METHODS: Air kerma calibration coefficients for 150 cylindrical ionization chamber calibrations performed between 2017 and 2024 in 59 medium-energy x-ray beam qualities at the Australian Primary Standards Laboratory were analyzed. Chamber response was scrutinized and assessed for compliance with international recommendations on absorbed dose determination for kilovoltage x-ray beams. RESULTS: All chambers showed some degree of energy dependence, with none of the chamber models investigated fully satisfying international recommendations. Chamber response is dominated by filtration (half-value layer), with a weaker dependence on generating potential (kVp); differences of up to ±4.3% were observed for a 100 kVp beam spanning 1.30-6.61 mm Al HVL. Distinct differences in response were observed between chamber models, providing rationale for use or exclusion in kV absorbed dose determination. Energy dependence was observed to vary between chambers of the same model type, indicating that calibration coefficients cannot be interchanged between individual chambers of the same model. CONCLUSIONS: Cylindrical ionization chambers frequently used for reference dosimetry of kV x-ray beams exhibit different energy dependence based on model type. Some models frequently employed and recommended for use in absorbed dose determination do not satisfy international protocol recommendations regarding energy dependence. International recommendations and protocols for absorbed dose determination for kilovoltage x-rays based solely on kVp should be reviewed and further developed to explicitly account for the effect of filtration (HVL) on chamber response. The results of this work support the adoption or refinement of the recommended chamber types within reference dosimetry protocols.
Out-of-beam artifact suppression in charged nuclear fragment based carbon-ion radiotherapy monitoring.
Kirchgässner R, Martišíková M, Schlegel P … +5 more , Ochoa-Parra P, Harrabi S, Jäkel O, Debus J, Kelleter L
Med Phys · 2026 Mar · PMID 41807765 · Full text
BACKGROUND: Carbon-ion radiotherapy offers highly precise targeting of tumors while sparing healthy tissue compared to X-ray therapy. However, this precision comes at the cost of an increased sensitivity of the treatment...
BACKGROUND: Carbon-ion radiotherapy offers highly precise targeting of tumors while sparing healthy tissue compared to X-ray therapy. However, this precision comes at the cost of an increased sensitivity of the treatment to range uncertainties, which can arise from anatomical changes of the patient. Our group develops an in-vivo treatment monitoring method by tracking of charged nuclear fragments using hybrid silicon pixel detectors. PURPOSE: Anatomical changes outside of the region accessed by carbon-ion beams are clinically not relevant, as they do not affect the dose distribution. However, they can potentially influence the fragment data, producing artifacts, which might be interpreted as signals produced by clinically relevant anatomical changes. This misinterpretation would cause unnecessary clinical action, like performing a CT scan. This work proposes methods for the identification and suppression of clinically irrelevant artifacts with the aim of avoiding unnecessary clinical action. METHODS: A clinically relevant and an irrelevant anatomical change are emulated by introducing coin-sized air cavities at different positions in a homogeneous cylindrical plastic head phantom. Charged nuclear fragments are detected by a Timepix3-based mini-tracker during irradiations of this phantom with a clinically realistic treatment plan. All measurements are performed for two different positions of the mini-tracker. The reconstructed fragmentation vertex distributions are analyzed and compared to those of reference measurements. RESULTS: A significant signal from the clinically irrelevant air cavity was observed. This artifact was found to differ from the signal of the clinically relevant cavity. Most importantly, the location of the artifact changes with the mini-tracker position, whereas the relevant signal remains unchanged. This facilitates identification of the artifact as well as its suppression by combining the data from several mini-trackers at different positions around the patient. CONCLUSIONS: Clinically irrelevant changes were shown to potentially impede carbon-ion treatment monitoring by tracking of charged nuclear fragments. However, positioning several mini-trackers around the patient, which monitor the treatment from different perspectives, was found to be the key to the identification and suppression of artifacts from anatomical changes outside of the region accessed by carbon-ion beams. This is implemented in the detection system of an ongoing clinical trial.
Cross-technique transfer learning to predict the dose distribution for radiotherapy planning based on a limited sample size.
Wang X, Wang K, Ding J … +3 more , Chen X, Men K, Dai J
Med Phys · 2026 Mar · PMID 41807749 · Publisher ↗
BACKGROUND: Accurate dose prediction is challenged by the lack of available training samples and the rapid evolution of radiotherapy techniques. PURPOSE: A cross-technique transfer learning strategy was developed to pred...
BACKGROUND: Accurate dose prediction is challenged by the lack of available training samples and the rapid evolution of radiotherapy techniques. PURPOSE: A cross-technique transfer learning strategy was developed to predict the dose distribution for radiotherapy planning using limited training samples. METHODS: Data were collected from 154 patients with nasopharyngeal carcinoma: 60 treated with intensity-modulated radiotherapy (IMRT) and 94 treated with volumetric modulated arc therapy (VMAT). The Res-U Net was selected as the base deep learning network. Cross-technique models were pretrained on the IMRT dataset and subsequently fine-tuned on VMAT data using limited samples (five and seven cases). Independent models were trained from scratch using the same limited samples, while a standard model trained on the full VMAT training set served as the reference. Model performance was evaluated on a test set using metrics including the dose-volume histogram (DVH), voxel-based mean absolute error (MAE), and the Dice similarity coefficient (DSC) of the isodose volume. RESULTS: The cross-technique models exhibited clinically acceptable performance with only five training samples and were comparable to the standard model (MAE deviation: 0.15%, p > 0.01 after Bonferroni correction; DSC deviation: 0.11%-0.72%). Performance improved further with seven training samples (MAE deviation: 0.05%, p > 0.01; DSC deviation: 0.02%-0.40%). However, the independent models trained with five or seven samples showed significantly inferior performance (five samples: MAE deviation: 1.14%, p < 0.01, DSC deviation: 0.98%-2.48%; seven samples: MAE deviation: 0.50%, p < 0.01, DSC deviation: 0.48%-1.05%). CONCLUSION: The cross-technique models accurately and reliably predicted the dose distribution for a new radiotherapy technique using a limited sample size.
Plaque-adaptive flexible film dosimetry system for Ru-106 plaque brachytherapy commissioning and quality assurance.
Cho JD, Kim H, Kim DW … +3 more , Nam H, Kim JS, Kim JI
Med Phys · 2026 Mar · PMID 41807326 · Full text
BACKGROUND: Ruthenium-106 (Ru-106) plaque brachytherapy is an established treatment modality for choroidal melanoma that delivers localized beta radiation while sparing surrounding ocular structures. It is widely used in...
BACKGROUND: Ruthenium-106 (Ru-106) plaque brachytherapy is an established treatment modality for choroidal melanoma that delivers localized beta radiation while sparing surrounding ocular structures. It is widely used in Europe and parts of Asia, whereas COMS Iodine-125 and Palladium-103 plaques remain the standard in the United States. However, its concave geometry presents challenges for accurate dose verification, as conventional flat dosimeters are ill-suited for use during commissioning and quality assurance (QA). PURPOSE: This study aimed to assess the feasibility and dosimetric accuracy of a plaque-adaptive flexible film dosimeter (PFD) for planar and central axis (CAX) dose measurements in Ru-106 plaque brachytherapy. METHODS: The PFD consisted of an active layer of lithium salt of pentacosa-10,12-diynoic acid (LiPCDA) enclosed between upper and lower silicone sheets. The flexible film was shaped to conform to the curvature of the Ru-106 plaques and mounted within a dual-layer film holder. A 3D-printed QA tool was used for reproducible positioning. Dose measurements were performed at depths of 1 and 3 mm for CCA, COB, and CIB plaque types using a dual-film configuration. Planar dose distributions were assessed at 33 manufacturer-certified reference points at a 1 mm depth. CAX dose profiles were independently acquired using a microDiamond detector in a water phantom. All measurements were compared with manufacturer-certified reference data. Calibration curves were established based on optical density and dose uncertainties were analyzed. RESULTS: The dose ratio between 1 and 3 mm depths measured with the PFD differed by -0.2% to 0.4% from the reference data across all plaque types. Planar dose measurements at 33 points yielded a mean difference of 3.0% ± 2.0% from reference values. In the high-dose central region, the mean differences for CCA, COB, and CIB were 2.1% ± 0.8%, 2.1% ± 0.8%, and 2.2% ± 0.7%, respectively. CAX dose profiles measured with a microDiamond detector differed by 0.89% ± 0.58%. Total dose uncertainty ranged from 4.43% at 0.3 Gy to 4.19% at 2.4 Gy. CONCLUSIONS: The PFD enabled accurate and high-resolution dose measurements on the curved surfaces of Ru-106 plaques. Both planar and CAX measurements demonstrated strong agreement with manufacturer-certified reference data across all plaque types. This system offers a practical and reproducible solution for commissioning and QA in Ru-106 plaque brachytherapy, supporting its clinical integration for choroidal melanoma treatment.
Robust cascade bidirectional triple capsule network with OOA for deep neural network-based improved brain tumor identification.
Thangavel K, Murugavelu M, Nelson SC … +1 more , Sreekanth GR
Med Phys · 2026 Mar · PMID 41807104 · Publisher ↗
BACKGROUND: Abnormal growth of brain cells may produce serious neurological symptoms (migraines, seizures, and cognitive impairments), which are called brain tumors. Early and precise diagnosis is of utmost importance in...
BACKGROUND: Abnormal growth of brain cells may produce serious neurological symptoms (migraines, seizures, and cognitive impairments), which are called brain tumors. Early and precise diagnosis is of utmost importance in enhancing prognosis and treatment options such as surgery, radiation, or chemotherapy. Although the current systems have improved in Deep Learning (DL), they continue to record poor accuracy and high false positive rates, which warrant more trustworthy solutions. PURPOSE: In this paper, the authors propose a Robust Triple Extraction with Cascade Bidirectional Capsule Network and Osprey Optimization Algorithm (RT-CBCN-OOA) to achieve a higher quality, accuracy, and dependability of brain tumor detection and classification. METHODS: The pre-processing of the BraTS and Figshare MRI images is performed with the help of the Modified Square Root Sage-Husa Adaptive Kalman Filter (MSRS-HAKF) in order to eliminate noise and enhance image clarity. Dual-Domain Attention CNN based on EfficientNet-B3 CNN (EN-B3 CNN-2DA) is used to extract features and segment and classify these features using the Geometric Algebra Transformer-based Robust Cascade Bidirectional Triple Capsule Network with Triple Attention (GAT-RCBTCN-TA). The Osprey Optimization Algorithm (OOA) optimizes the performance of model weights. RESULTS AND CONCLUSION: The proposed RT-CBCN-OOA has a recall and accuracy of 99.9 and 99.8, respectively, which is better than the current models. It provides a powerful, precise, and efficient brain tumor detection, which proves to have a great future in clinical use in medical imaging and diagnosis.
AAPM task group 317 report: A joint AAPM and ESTRO report on brachytherapy catheter, needle, and applicator tracking technology.
Beaulieu L, Bert C, Borot de Battisti M … +11 more , Cormack RA, Cunha JAM, Damato AL, Deufel CL, Hautvast GLTF, Hsu IC, Kolkman-Deurloo IK, Moerland MA, Niatsetski Y, Weersink RA, Tanderup K
Med Phys · 2025 Oct · PMID 41805152 · Publisher ↗
In recent years, various tracking technologies that work independently of imaging systems have been proposed to automate, simplify, and enhance various tasks in the brachytherapy treatment workflow. These tasks, critical...
In recent years, various tracking technologies that work independently of imaging systems have been proposed to automate, simplify, and enhance various tasks in the brachytherapy treatment workflow. These tasks, critical to the overall accuracy of the therapeutic dose delivery, include applicator, catheter and needle insertion guidance, and reconstruction as well as transfer tube connection in afterloading technique. Task Group 317 was established as a joint American Association of Physicists in Medicine (AAPM) and European Society for Radiotherapy and Oncology (ESTRO) committee to review: the current state-of-the-art scientific literature as it pertains to tracking technology in the field of brachytherapy; the benefits and issues related to the use of the technology for automated reconstruction of brachytherapy implants, quality control (QC) tasks such as channel path and tip reconstruction, and real-time guidance tasks; their limitations, in particular in the clinical environment and, finally, to develop recommendations related to commissioning, quality assurance (QA) and clinical use. The Task Group has looked in detail at key tracking technologies in advanced brachytherapy applications: infrared, electromagnetic, fiber optic shape sensing (fiber Bragg grating), and active radiofrequency coil tracking. For each, the performance and accuracy in well-controlled conditions as well as in clinically relevant environments are provided. Guidelines for clinical implementations, including target accuracy and performance needed for critical tasks, are summarized. Risk-based analysis is discussed in the context of an electromagnetic-based tracking system used as part of a clinical trial. The report concludes with the essential elements of an effective quality management program dedicated to the advanced features enabled by the above-described technology.
Correction to "Reliability of log file-based patient-specific IMRT QA in a multi-vendor LINAC enterprise depends on inter-vendor compatibility".
Med Phys · 2026 Mar · PMID 41804049 · Publisher ↗
Abstract loading — click title to view on PubMed.
A server-based solution for misregistration correction in PET/CT using limited-coverage and data-driven gated CT.
Pan T, Luo D
Med Phys · 2026 Mar · PMID 41795690 · Publisher ↗
BACKGROUND: Misregistration between CT and PET images can compromise lesion localization and the accuracy of tracer uptake quantification. Although repeating a limited-coverage (LC) CT scan may resolve the issue, most PE...
BACKGROUND: Misregistration between CT and PET images can compromise lesion localization and the accuracy of tracer uptake quantification. Although repeating a limited-coverage (LC) CT scan may resolve the issue, most PET/CT systems require extending the LC CT to cover one to two PET bed positions, substantially increasing the patient's CT dose and adding the complexity of matching the LC CT to the corresponding PET positions. Consequently, many clinics instead repeat an LC PET/CT scan, which is operationally simpler but adds several minutes of unnecessary PET acquisition and still results in a significant increase in CT dose. A more efficient solution is needed-one that minimizes CT doses without necessitating an additional PET scan. PURPOSE: We aimed to develop a misregistration correction server (MCS) to solve the problems of excessive CT dose and unnecessary PET scans associated with current limited coverage (LC) CT and LC PET/CT procedures. METHODS: A new MCS was developed to embed the CT of PET/CT with either an LC CT or data-driven gated (DDG) CT to enable PET attenuation correction. Both LC CT and DDG CT can be positioned across PET bed positions and does not need a repeat PET. The MCS currently supports misregistration correction of 9 scanners within our hospital network: 8 GE DMI and 1 Siemens Quadra. It can simultaneously process data from all 9 scanners and return the WB CT embedded with LC CT or DDG CT to each requesting scanner for misregistration correction. Over 2,033 patient studies have been corrected. The dose implications of the MCS will be assessed. RESULTS: The processing time from data transfer to output was approximately 1 min for LC CT and 3 min for DDG CT. The MCS workflow has minimal operational impact and eliminates the need for repeat PET acquisitions. The LC/DDG CT scan ranges were 17.5 ± 4.7 cm on the GE DMI and 22.1 ± 5.6 cm on the Siemens Quadra, with corresponding doses of 7.52 ± 4.18 mGy and 5.63 ± 4.46 mGy. Registration improved in 2,004 of 2,013 DMI studies and 13 of 20 Quadra studies. The dose length product (DLP) values for LC/DDG CT (DMI: 133.9 ± 87.8 mGy-cm; Quadra: 108.3 ± 90.6 mGy-cm) were substantially lower than the WB CT values (DMI: 569.0 ± 305.0 mGy-cm; Quadra: 641.8 ± 321.5 mGy-cm), representing only 23.5% and 16.8% of the corresponding WB DLP. CONCLUSIONS: A new MCS has been developed to correct misregistration in PET/CT scanners from GE and Siemens. It can generate DDG CT from the cine CT data to correct for PET or DDG PET, or enable repeat CT of any coverage, avoiding the need for repeat PET and thereby reducing both CT dose and PET scan time. The additional doses in DLP accounted for 23.5% of the WB DLP for the GE DMI scanner and 16.8% for the Siemens Vision Quadra scanner.
Multileaf Faraday cup for beam energy verification in radiation therapy with ultra-high dose-rate electron beams and ion beams.
Makowski C, Deutsch M, Schmitzer CS … +1 more , Schüller A
Med Phys · 2026 Mar · PMID 41795689 · Full text
BACKGROUND: In radiotherapy and particle therapy, the stability of the beam energy must be checked routinely during quality assurance. This is of particular importance when varying the dose rate via the beam intensity of...
BACKGROUND: In radiotherapy and particle therapy, the stability of the beam energy must be checked routinely during quality assurance. This is of particular importance when varying the dose rate via the beam intensity of the accelerator in order to study the FLASH effect when comparing irradiations with conventional and ultra-high dose rates. PURPOSE: The standard method of energy verification based on the measurement of percent depth-dose curves in a water phantom is time-consuming and in the case of ultra-high dose-rate beams, may also present a non-negligible radiation protection problem. METHODS: A compact, portable multileaf Faraday cup (MLFC) with 128 channels was developed for energy determination based on the measurement of depth-charge curves. Its design is optimized for clinical electron beams in a range between 3 and 25 MeV. The read-out unit of the MLFC displays the beam energy in real time. The device was tested in strongly pulsed electron beams such as those present in irradiations with ultra-high doses per pulse as well as in conventional clinical electron beams. The same detector was used for proof of concept in a single measurement campaign in pulsed beams of therapeutical protons (conventional dose rates) from a synchrotron source using a suitable range shifter. The MLFC was calibrated with monoenergetic electron beams as well as against depth-dose curves. Simulations were carried out for comparison. RESULTS: The MLFC works well under ultra-high pulse dose rate conditions as well as in conventional electron beams generated by a medical accelerator. Changes in the beam energy of below 20 keV (0.1% of 20 MeV) can be clearly identified by means of the MLFC. For proton beams, well-defined peaks in the depth-charge curves are observed for each single synchrotron spill with a total charge of only about 1 nC. The energy values resulting from MC simulations of the measured MLFC data agree with the actual proton energies within 2 . CONCLUSIONS: The MLFC can be used for quick validation of the energy stability when carrying out experiments comparing electron beams of conventional and ultra-high dose rates. The same device can be used in the pulsed ion beams from a synchrotron.
Improved segmentation of cardiac structures in echocardiograms for diastolic function evaluation.
Lv C, Liu X, Liao K … +3 more , Chen M, Zhan C, Guo X
Med Phys · 2026 Mar · PMID 41795688 · Publisher ↗
BACKGROUND: Noninvasive assessment of diastolic dysfunction relies on multiple echocardiographic indicators, including measurements from both standard B-mode images and Doppler, obtained at various cardiac locations such...
BACKGROUND: Noninvasive assessment of diastolic dysfunction relies on multiple echocardiographic indicators, including measurements from both standard B-mode images and Doppler, obtained at various cardiac locations such as the mitral annulus, tricuspid annulus, left ventricle, and left atrium. The diagnostic process is complex and subject to interobserver variability, making accurate and rapid evaluation challenging. Automated semantic segmentation of key cardiac structures, such as the left atrium, left ventricle, and mitral valve annulus, offers a potential solution by capturing temporal changes throughout the cardiac cycle. PURPOSE: This study aims to improve the accuracy of segmenting the left atrium, left ventricle, and mitral valve annulus in echocardiographic images and to leverage the resulting temporal segmentation features for more reliable identification of diastolic dysfunction. METHODS: This study presents Diff-TransUNet, a novel segmentation model incorporating a noise-robust Differential Transformer module. Evaluations on private (1137 training images, 135 validation images, and 88 test images), CAMUS (1400 training images, 200 validation images, and 200 test images), and EchoNet-Dynamic (5000 training images, 2546 validation images, and 2528 test images) datasets demonstrate improved performance over state-of-the-art methods, assessed by Dice coefficient (Dice), Intersection-over-Union (IoU), and 95th percentile Hausdorff Distance (HD) metrics. Statistical analysis was performed to compare Diff-TransUNet with baseline methods across evaluation metrics. To control for errors arising from multiple comparisons, p-values were adjusted using the Benjamini-Hochberg false discovery rate (FDR) correction. Statistical significance was assessed at a 95% confidence level. In addition to p-values, Cohen's d effect size was computed to quantify the practical significance of performance differences. RESULTS: The proposed Diff-TransUNet achieved a Dice of 87.49%, IoU of 79.07%, and HD of 1.48 on the private dataset. Compared with state-of-the-art models, Dice improved by 1.35%-4.30% (p < 0.05, Cohen's d = 0.32-0.90), IoU by 1.97%-5.67% (p < 0.05, Cohen's d = 0.37-1.03), and HD by 0.16-0.83 (p < 0.05, Cohen's d = 0.21-0.90). On the CAMUS dataset, the model achieved a Dice of 88.74%, IoU of 80.58%, and HD of 2.83, showing improvements of 1.07%-4.96% (p < 0.05, Cohen's d = 0.18-0.63) in Dice, 1.55%-6.89% (p < 0.05, Cohen's d = 0.19-0.71) in IoU, and 0.41-2.85 (p < 0.05, Cohen's d = 0.12-0.46) in HD compared to advanced models. On the EchoNet-Dynamic dataset, the model obtained a Dice of 92.25%, IoU of 85.87%, and HD of 1.65, outperforming other methods by 0.42%-2.00% (p < 0.05, Cohen's d = 0.10-0.40) in Dice, 0.69%-3.21% (p < 0.05, Cohen's d = 0.10-0.43) in IoU, and 0.21-1.12 (p < 0.05, Cohen's d = 0.09-0.34) in HD. Furthermore, by extracting volumetric segmentation features, the proposed method achieved an accuracy of 88.95% (95 % CI 87.15% to 90.08%) in identifying diastolic dysfunction. CONCLUSIONS: The proposed Diff-TransUNet model achieves significant improvements in ultrasound segmentation. Features extracted from the left ventricle, left atrium, and mitral annulus segmented by Diff-TransUNet can be effectively used for the identification of diastolic dysfunction.
A Monte Carlo study of proton LET calculations: Comparison between the codes PENHAN, FLUKA, and TOPAS.
Puerta D, Anguiano M, González W
Med Phys · 2026 Mar · PMID 41795670 · Full text
BACKGROUND: Linear energy transfer (LET) is frequently used to characterize radiation quality in proton therapy and is an important quantity for radiobiological modeling. However, LET calculations can, in principle, vary...
BACKGROUND: Linear energy transfer (LET) is frequently used to characterize radiation quality in proton therapy and is an important quantity for radiobiological modeling. However, LET calculations can, in principle, vary between Monte Carlo (MC) codes, affecting consistency in treatment planning and relative biological effectiveness (RBE) estimation. PURPOSE: This study aims to investigate the variability of LET results when using different MC codes. METHODS: An intercomparison of LET results was performed using penhan, topas MC, and CERN fluka, together with recently published experimental data. Track-averaged ( ) and dose-averaged ( ) LET were evaluated under different irradiation configurations, including pencil and broad beams, with and without nuclear reactions. RESULTS: The results obtained indicate that discrepancies in LET calculations are primarily linked to the choice of stopping powers and the handling of nuclear interactions. Calculations considering only primary protons exhibited great agreement with recent experimental LET measurements, validating our simulations. CONCLUSIONS: showed higher robustness and consistency across codes, suggesting it as potentially a more reliable metric for LET-based treatment planning and RBE modeling. Our findings emphasize the importance of consistent LET calculations, result reporting, and code benchmarking in proton therapy calculations.
Longitudinal analysis of hippocampal subfield atrophy and network centrality associated with cognitive decline in Alzheimer's disease progression.
Lim S, Lee Y
Med Phys · 2026 Mar · PMID 41776730 · Full text
BACKGROUND: Alzheimer's disease (AD) is a progressive neurodegenerative disorder characterized by structural degeneration of the hippocampus. Previous studies have demonstrated that specific hippocampal subfields, such a...
BACKGROUND: Alzheimer's disease (AD) is a progressive neurodegenerative disorder characterized by structural degeneration of the hippocampus. Previous studies have demonstrated that specific hippocampal subfields, such as the cornu ammonis (CA)1 and the subiculum, are susceptible to AD-related atrophy. However, most previous studies have focused on cross-sectional volumetric changes without investigating the interrelationships between subfields or their network-level functions throughout the disease progression. PURPOSE: To examine the longitudinal volumetric changes in the hippocampal subfields over 2 years in individuals who progressed from mild cognitive impairment (MCI) to AD. In addition, we aimed to investigate the associations between cognitive decline, inter-region structural correlation, and network-based centrality profiles of subfields based on atrophy covariance and changes in subfield volume. METHODS: T1-weighted magnetic resonance images of 258 participants who progressed from MCI to AD were obtained from the Alzheimer's Disease Neuroimaging Initiative. Hippocampal subfield volumes were extracted at baseline and during follow-up using FreeSurfer's longitudinal pipeline (v7.4.1). The subfield volume changes were examined using the paired-t tests. Cognitive decline was measured as the percentage change in the Mini-Mental State Examination (MMSE) scores. The partial Pearson's correlations between subfield volume changes and MMSE change were calculated after adjusting for baseline age, sex, education level, and apolipoprotein ε4 status. A structural covariance network was constructed using inter-subfield partial correlations. Four graph-theoretical centrality measurements (degree, betweenness, closeness, and eigenvector) were computed from the network to identify structurally central subfields. RESULTS: Most hippocampal subfields demonstrated important volume decreases over 2 years, with the left fimbria, subiculum head, and dentate gyrus head showing the most atrophy. The left hippocampus showed significantly greater volume decreases than the right hippocampus. Volume changes in the left presubiculum body, the CA3 head, and the dentate gyrus head were strongly correlated with MMSE decline. Notably, structured covariance patterns between anatomically and functionally relevant subfields within the CA1-CA3-CA4-dentate gyrus axis and subiculum complex were found by inter-regional analysis. Network-based analysis identified the left CA1 head and left dentate gyrus head as central hubs across all four-centrality metrics. Other subfields, including the left subiculum head and left molecular layer head, also showed high centrality in several respects, indicating their possible coordinating functions in hippocampal degeneration. CONCLUSIONS: This study provides a comprehensive longitudinal analysis of hippocampal subfield atrophy, inter-regional co-atrophy patterns, and network centrality during MCI-to-AD progression. Our findings demonstrate that several subfields, including the left CA1 and dentate gyrus, are structurally and functionally central in the hippocampal atrophy network. The integration of volumetric, correlation-based, and graph theory-based approaches offers new insights into the coordinated degeneration of the hippocampus in AD, emphasizing the importance of subfield-level network dynamics in understanding the disease progression.
Upright and supine particle therapy of lung cancer: A 4D dosimetric comparison.
Martire MC, Volz L, Durante M … +2 more , Pankuch M, Graeff C
Med Phys · 2026 Mar · PMID 41776380 · Full text
BACKGROUND: A renewed interest in upright particle therapy is currently driven by the availability of upright positioning and imaging systems. The upright positioning system could enhance fixed beamlines for effective ca...
BACKGROUND: A renewed interest in upright particle therapy is currently driven by the availability of upright positioning and imaging systems. The upright positioning system could enhance fixed beamlines for effective carbon ion treatments in central body regions, with a substantial cost and space advantage. In addition, few studies have suggested advantages in patient breathing and lung volume in an upright posture. Comparative dosimetric analyses are needed to determine the clinical viability of upright patient positioning for carbon ion therapy of thoracic cancers but are challenged by various sources of bias. PURPOSE: To provide a comprehensive analysis of all parameters influencing the comparison between upright and supine carbon therapy of thoracic patients through 4D dosimetric studies. METHODS: Paired upright and supine 4DCTs were available for six patients treated at the Northwestern Medicine Proton Centre (NMPC), under the Proton Collaborative Group (PCG) registry. Deformable image registration (DIR) between upright and supine CTs was performed on a region of interest (ROI) including the rib cage for target propagation, to avoid failure in DIR caused by thorax anatomical differences. DIR quality was evaluated on lung structures through Dice similarity coefficient (DSC) and average Hausdorff distance (AHD) metrics. Paired 3D plans were optimized on the originally contoured and propagated target volumes, to investigate the effect of segmentation differences. The impact of beam geometry choice was investigated by optimizing plans with a variety of treatment angles. Single-fraction and accumulated 4D doses were calculated with the research treatment planning system TRiP4D to analyze the impact of differences in breathing-induced tumor motion in the two postures. Plan quality between upright and supine plans were assessed through D, HI, and V for the internal target volume (ITV) and V16Gy(lung) and V20Gy(heart) for lung and heart, respectively. RESULTS: Restraining DIR on the ribcage ROI enabled successful DIR. Within the ribcage ROI an average AHD of 1.5 mm and DSC of 0.95 was achieved on the propagated lung structure. Position specific angle selection showed vertical posterior/anterior beams might not be optimal for upright treatments. Comparable 3D treatment quality was achieved for five patients, while an increase of 5 pp occurred in V20Gy(heart) and V16Gy(lung) of patient P6 in upright. The 4D study showed the different positions have clinically relevant impact, increasing D of 3 pp for one patient with halved motion amplitude in upright posture. In addition, robustness was similar between postures, even with a more conservative 5%/5 mm uncertainty setting for upright. When assuming only a fixed beam line is available, as is the case for most carbon ion centers, a comparable plan quality with 360° beam angle flexibility in upright position was observed. CONCLUSIONS: The presented work comprehensively evaluates the influence of various parameters on the comparison of upright and supine therapy of thoracic patients. A solid understanding of these parameters is paramount to reduce bias in future larger patient cohort studies on the viability of upright positioning. The final dosimetric comparison between postures highly depends on patient characteristic and the investigated parameter. More data are needed to provide a resilient comparison between postures.
Hybrid receptive field U-Net for lesion segmentation in breast ultrasound images.
Xue Z, Feng Y, Liu N
Med Phys · 2026 Mar · PMID 41776377 · Publisher ↗
BACKGROUND: Numerous neural networks based on the U-Net architecture have been developed for the segmentation of breast ultrasound images. However, the accuracy of such segmentation tasks is often compromised by the comp...
BACKGROUND: Numerous neural networks based on the U-Net architecture have been developed for the segmentation of breast ultrasound images. However, the accuracy of such segmentation tasks is often compromised by the complex and variable shapes of tumors, the indistinct boundaries of lesion areas, and background noise. PURPOSE: In this study, we develop a hybrid receptive field U-Net (HRF U-Net) to improve lesion segmentation in breast ultrasound images. METHODS: We design an innovative hybrid receptive field module (HRFM) to replace conventional convolutional layers, integrating deformable convolutions that expand the receptive field and enhance the model's ability to capture shape and boundary features. We systematically analyze the differential feature extraction capabilities of deformable convolutions across network layers, combining them with standard convolutions, dilated convolutions, and max pooling. This configuration enables HRF U-Net to achieve a broader and more adaptive receptive field, enhancing its effectiveness in handling the complexities of breast lesion segmentation. Additionally, we introduce a large-kernel attention module (LKAM) within the skip connections, which expands the receptive field and supports adaptive feature selection, capturing long-range dependencies within the convolutional attention mechanism. This novel approach enables more precise feature extraction, effectively mitigates boundary noise during training, and substantially improves the model's segmentation performance. We used three publicly available datasets to conduct extensive experiments, including ablation studies, comparative analyses, robustness evaluations, and external validation. Datasets A and B were divided into training and validation sets for four-fold cross-validation, while Dataset C was used as the test set for external validation. We selected five widely used image segmentation metrics for validation, namely pixel accuracy, precision, recall, Jaccard index, and Dice coefficient. In addition, the statistical significance was evaluated using the paired Student's t-test with Holm-Bonferroni correction . RESULTS: Experimental results on three public datasets demonstrate that the proposed HRF U-Net substantially improves the efficacy of breast ultrasound image segmentation, outperforming several state-of-the-art works. Specifically, in the comparative experiments, HRF U-Net achieved scores of 96.52, 85.03, 83.19, 73.15, and 81.34 for the five metrics on Dataset A. On Dataset B, the scores reached 98.61, 90.22, 83.99, 74.50, and 83.64 . The effect size Cohen's d values for HRF U-Net compared to several typical and advanced networks all exceed 0.32, demonstrating a significant improvement in segmentation performance. In external validation, the metric scores reached 97.02, 84.35, 93.94, 77.74, and 86.86 . CONCLUSIONS: The HRF U-Net proposed in this paper improves the segmentation accuracy of breast ultrasound images.
Motion-compensated diffusion encoding gradients for segmented thick-slab 3D magnetic resonance diffusion-weighted imaging of the brain.
Johansson J, Lagerstrand K, Hebelka H … +1 more , Maier SE
Med Phys · 2026 Mar · PMID 41776355 · Full text
BACKGROUND: Routine clinical magnetic resonance diffusion-weighted imaging (DWI) is generally performed with 2D echo planar sequences. A single thick-slab 3D approach could offer higher signal-to-noise ratio and better s...
BACKGROUND: Routine clinical magnetic resonance diffusion-weighted imaging (DWI) is generally performed with 2D echo planar sequences. A single thick-slab 3D approach could offer higher signal-to-noise ratio and better slice resolution. This has not been adopted due to the difficulty to avoid motion-induced phase error that interfere with multi-shot spatial encoding. PURPOSE: To introduce a new approach for 3D brain DWI: rather than relying on navigator echoes for phase correction, moment-nulled diffusion encoding gradients are used to minimize phase variations at the source. METHODS: A standard 2D echo planar imaging sequence was modified to incorporate moment-nulled diffusion encoding gradients and a second phase encoding gradient for spatial multi-shot encoding along the slice select direction. The single thick-slab 3D diffusion-weighted imaging sequence was evaluated with brain scans in healthy volunteers on a 3 Tesla scanner. RESULTS: Incorporation of both first and second order moment nulling achieved substantial, albeit not comprehensive, reduction of motion-related ghosting artifacts. Without such motion compensation or with first order moment nulling only, motion-related artifacts were consistently more severe. Even though the approach comes with a penalty in echo time-at a diffusion weighting of 1000 s/mm, 119 ms for the moment-nulled 3D acquisition versus 82 ms for the conventional 2D acquisition-the measured ratio between SNR and SNR for a 92-slice scan was 0.99. CONCLUSIONS: This proof-of-concept shows that first and second order moment nulling may be a viable avenue for enabling 3D diffusion imaging. At higher slice numbers the SNR is expected to clearly surpass corresponding SNR. However, further investigation into echo time reduction and correction of residual phase variations is needed before the 3D approach is viable for translation into a clinical setting. Specifically, with higher gradient strength shorter echo times can be realized. Moreover, this reduces third and higher order gradient moments and associated residual phase shifts.
Correction to "A multi-scale network for thyroid segmentation and dose assessment in hyperthyroid patients with uncertainty quantification".
Med Phys · 2026 Mar · PMID 41761611 · Publisher ↗
Abstract loading — click title to view on PubMed.
Influence of radiotherapy dose rate on gold nanoparticle-induced radiosensitization from high dose-rate brachytherapy and external beam therapy.
Cecchi D, Jackson N, Freeman S … +4 more , O'Neil K, Goharian M, Beckham W, Chithrani DB
Med Phys · 2026 Mar · PMID 41761610 · Full text
BACKGROUND: Associated normal tissue toxicity from current radiotherapy (RT) treatments limits effective dose escalation in the tumor to achieve nominal treatment results. Gold nanoparticles (GNPs) as radiosensitizing ag...
BACKGROUND: Associated normal tissue toxicity from current radiotherapy (RT) treatments limits effective dose escalation in the tumor to achieve nominal treatment results. Gold nanoparticles (GNPs) as radiosensitizing agents to locally increase photoelectron production have gained interest as a safe and viable method to improve therapeutic results. Among many other factors, the dose rate of the incident radiation has been shown to affect the radiosensitizing properties of GNPs significantly. PURPOSE: To evaluate GNP-induced radiosensitization during variable dose rate delivery from the decay of a high dose rate brachytherapy 192-Ir source and a clinical 6MV linear accelerator (LINAC). METHODS: HEC-1A endometrial cancer cells were seeded into 35mm petri dishes with or without 10µg/mL with spherical 11nm GNPs functionalized with polyethylene glycol and integrin binding domain RGD to improve intracellular uptake. Variable dose rate delivery from the 192-Ir source was achieved at two source strengths of 37.95 mGy m/h and 18.97 mGy m/h, corresponding to dose rates of 1.1 and 0.55Gy/min, respectively. For 6MV irradiations, dose rate variability was controlled by adjusting the distance to the target from 91cm to 129cm, yielding identical dose rates of 1.1 and 0.55Gy/min, respectively. Cellular viability was measured using a clonogenic assay after irradiations between 0 and 8Gy, and a DNA double-strand break assay after 2Gy irradiations. RESULTS: GNP-induced radiosensitization was significantly greater with higher dose rates than lower. Clonogenic loss with GNPs was increased from 1.00 to 1.19 (p < 0.001) with higher dose rates from 192-Ir source and from 1.03 to 1.16 (p < 0.001) with higher dose rate LINAC irradiations. DNA damage increase from GNPs was not significant at lower dose rates for both 192-Ir (p > 0.05) and LINAC (p > 0.05) irradiations; however, DNA damage was significantly increased at higher dose rates (192-Ir: p < 0.01; 6MV: p < 0.05). CONCLUSIONS: We have successfully demonstrated in vitro that clinically plausible GNP concentrations can induce variable radiosensitization based on the administered dose rate from both 192-Ir and LINAC irradiations. This work demands future research into the clinical translation of GNPs into high-dose-rate environments.
About
- Frequency
- Sun
- Papers found
- 200
- RSS feed
- Subscribe