Medical physics最新文献

{"title":"Characterization of commercial detectors for absolute proton UHDR dosimetry on a compact clinical proton synchrocyclotron.","authors":"Jufri Setianegara, Aoxiang Wang, Nicolas Gerard, Jarrick Nys, H Harold Li, Ronald C Chen, Hao Gao, Yuting Lin","doi":"10.1002/mp.17847","DOIUrl":"https://doi.org/10.1002/mp.17847","url":null,"abstract":"<p><strong>Background: </strong>Modern compact proton synchrocyclotrons can achieve ultra-high dose rates ( <math><semantics><mo>≥</mo> <annotation>$ ge $</annotation></semantics> </math> 40 Gy/s) to support ultra-high-dose-rate (UHDR) preclinical experiments utilizing pencil beam scanning (PBS) protons. Unique to synchrocyclotrons is a pulsed proton time structure as compared to the quasi-continuous nature of other proton accelerators like isochronous cyclotrons. Thus, high instantaneous proton currents in the order of several µA must be generated to achieve UHDRs. This will lead to high doses-per-pulse (DPP), which may cause significant charge recombination for ionization chambers, which must be characterized for accurate UHDR dosimetry programs.</p><p><strong>Purpose: </strong>In this work, we investigate the suitability of various commercial radiation detectors for accurate proton UHDR dosimetry using PBS proton beams from a compact proton synchrocyclotron (IBA ProteusONE). This is achieved by cross-calibrating them with conventional dose rates, measuring UHDR recombination (P<sub>ion</sub>) and polarity correction factors (P<sub>pol</sub>) for ionization chambers, and determining the absorbed proton UHDR dose delivered for all detectors.</p><p><strong>Methods: </strong>An IBA ProteusONE synchrocyclotron was initially tuned to achieve UHDRs with 228 MeV protons at 0° gantry angle. Various detectors, including Razor Chamber, Razor Nano Chamber, Razor Diode, and microDiamond, were cross-calibrated against a PPC05 plane-parallel ionization chamber (PPIC) that had an ADCL calibration coefficient of 59.23 cGy/nC. Then, all ionization chambers were exposed to UHDR protons with the P<sub>pol</sub> and P<sub>ion</sub> subsequently calculated. P<sub>ion</sub> was calculated using two methods: TRS-398 methods and Niatel's model. Finally, the absolute UHDR proton doses delivered were determined for all detectors and cross-compared.</p><p><strong>Results: </strong>Faraday cup measurements were performed for a single spot proton UHDR beam, and the nozzle current at the isocenter was determined to be 129.5 nA during UHDR irradiations at 98.61% of the maximum theoretical dose rate. Repeated Faraday cup measurements of the UHDR beam yielded a percentage standard deviation of 0.8%, which was higher than 0.120% when similar repeated measurements were performed with conventional proton beams. P<sub>pol</sub> was found to be relatively dose-rate independent for all ionization chambers investigated. P<sub>ion</sub> was found to be the lowest for the PPC05 ionization chamber (1.0097) compared to corresponding values of 1.0214 and 1.0294 for the Razor and Razor Nano detectors, respectively, for UHDRs. P<sub>ion</sub> values calculated using Niatel's model closely matched values from TRS-398 if the V<sub>H</sub>/V<sub>L</sub> ratio were kept at 2.5 for the PPC05 and Razor detectors and 2.0 for the Razor Nano detector. Absolute proton UHDR doses determined using cros","PeriodicalId":94136,"journal":{"name":"Medical physics","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2025-04-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144065554","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Interpretable AI-assisted clinical decision making for treatment selection for brain metastases in radiation therapy.","authors":"Yufeng Cao, Hua-Ren Cherng, Dan Kunaprayoon, Mark V Mishra, Lei Ren","doi":"10.1002/mp.17844","DOIUrl":"https://doi.org/10.1002/mp.17844","url":null,"abstract":"<p><strong>Background: </strong>AI modeling CDM can improve the quality and efficiency of clinical practice or provide secondary opinion consultations for patients with limited medical resources to address healthcare disparities.</p><p><strong>Purpose: </strong>In this study, we developed an interpretable AI model to select radiotherapy treatment options, that is, whole-brain radiation therapy (WBRT) versus stereotactic radiosurgery (SRS), for patients with brain metastases.</p><p><strong>Materials/methods: </strong>A total of 232 patients with brain metastases treated by radiation therapy from 2018 to 2023 were obtained. CT/MR images with contoured target lesions and organs-at-risk (OARs) as well as non-image-based clinical parameters were extracted and digitized as inputs to the model. These parameters included (1) tumor size, shape, location, and proximity of lesions to OARs; (2) age; (3) the number of brain metastases; (4) Eastern Cooperative Oncology Group (ECOG) performance status; (5) presence of neurologic symptoms; (6) if surgery was performed (either pre/post-op RT); (7) newly diagnosed cancer with brain metastases (de-novo) versus re-treatment (either local or distant in the brain); (8) primary cancer histology; (9) presence of extracranial metastases; (10) extent of extracranial disease (progression vs. stable); and (11) receipt of systemic therapy. One vanilla and two interpretable 3D convolutional neural networks (CNN) models were developed. The vanilla one-path model (VM-1) uses only images as input, while the two interpretable models use both images and clinical parameters as inputs with two (IM-2) and 11 (IM-11) independent paths, respectively. This novel design allowed the model to calculate a class activation score for each input to interpret its relative weighting and importance in decision-making. The actual radiotherapy treatment (WBRT or SRS) used for the patients was used as ground truth for model training. The model performance was assessed by Stratified-10-fold cross-validation, with each fold consisting of selected 184 training, 24 validation, and 24 testing subjects.</p><p><strong>Result: </strong>A total of 232 brain metastases patients treated by WBRT or SRS were evaluated, including 80 WBRT and 152 SRS patients. Based on the images alone, the VM-1 model prescribed correctly for 143 (94%) SRS and 67 (84%) WBRT cases. Based on both images and clinical parameters, the IM-2 model prescribed correctly for 149 (98%) SRS and 74 (93%) WBRT cases. IM-11 provided the most interpretability with a relative weighting for each input as follows: CT image (59.5%), ECOG performance status (7.5%), re-treatment (5%), extracranial metastases (1.5%), number of brain metastases (9.5%), neurologic symptoms (3%), pre/post-surgery (2%), primary cancer histology (2%), age (1%), progressive extracranial disease (6%), and receipt of systemic therapy (4.5%), reflecting the importance of all these inputs in clinical decision-making.</p><p><stron","PeriodicalId":94136,"journal":{"name":"Medical physics","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2025-04-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144055963","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Sensitivity analysis of fluorescent nuclear track detectors for fast and high-energy mono-energetic neutron dosimetry.","authors":"Stefan Schmidt, Jeppe B Christensen, Benjamin Lutz, Alberto Stabilini, Eduardo G Yukihara, José Vedelago","doi":"10.1002/mp.17799","DOIUrl":"https://doi.org/10.1002/mp.17799","url":null,"abstract":"<p><strong>Background: </strong>In ion beam radiotherapy, treatment radiation fields are inevitably contaminated with secondary neutrons. The energies of these neutrons can reach several hundreds of MeV. Fluorescent nuclear track detectors (FNTDs) offer a promising solution for dosimetry of fast and high-energy neutrons, particularly given their low linear energy transfer in water (LET) detection threshold.</p><p><strong>Purpose: </strong>This study presents an experimental FNTD sensitivity analysis in six fast mono-energetic neutron fields, comparing the response to poly allyl diglycol carbonate (PADC) neutron detectors, and investigates the feasibility of estimating ambient dose equivalent for neutrons, <math> <semantics><msup><mi>H</mi> <mo>∗</mo></msup> <annotation>$H^{*}$</annotation></semantics> </math> (10). Moreover, it investigates the impact of converter thickness on the detector signal for both fast and high-energy neutrons and analyzes the resulting differences in signal.</p><p><strong>Methods: </strong>FNTDs and PADCs were exposed to mono-energetic neutron fields with energies of 1.2 MeV, 2.5 MeV, 5 MeV, 6.5 MeV, 14.8 MeV, and 19 MeV and evaluated based on the track density. The <math> <semantics><msup><mi>H</mi> <mo>∗</mo></msup> <annotation>$H^{*}$</annotation></semantics> </math> (10) values for FNTDs were determined by applying energy calibration factors, <math> <semantics><mrow><mi>k</mi> <mo>(</mo> <mi>E</mi> <mo>)</mo></mrow> <annotation>$k(E)$</annotation></semantics> </math> , which were determined through Monte Carlo (MC) simulations. The benchmarked MC model is employed to investigate the sensitivity of FNTDs to high-energy neutrons up to 200 MeV for various polyethylene (PE) converter thicknesses and to analyze the detector signal, including the particle type and the recoil proton LET.</p><p><strong>Results: </strong>The sensitivity values revealed an energy dependence for FNTDs, with variations by a factor of up to 23, whereas PADC detectors showed a smaller variation, ranging from 3 to 12. Accurate <math> <semantics><msup><mi>H</mi> <mo>∗</mo></msup> <annotation>$H^{*}$</annotation></semantics> </math> (10) estimation can be achieved employing MC-derived <math> <semantics><mrow><mi>k</mi> <mo>(</mo> <mi>E</mi> <mo>)</mo></mrow> <annotation>$k(E)$</annotation></semantics> </math> factors, with deviations not exceeding <math> <semantics><mrow><mn>10</mn> <mspace></mspace> <mo>%</mo></mrow> <annotation>$10, %$</annotation></semantics> </math> . The sensitivity values increased almost continuously up to <math> <semantics><mrow><mn>200</mn> <mspace></mspace> <mi>MeV</mi></mrow> <annotation>$200 ,mathrm{MeV}$</annotation></semantics> </math> for PE converter thicknesses above <math> <semantics><mrow><mn>2</mn> <mspace></mspace> <mi>mm</mi></mrow> <annotation>$2 ,mathrm{mm}$</annotation></semantics> </math> , whereas plateaued for thinner PE converters above 10 MeV to 15 MeV. For neutrons above <math> <semantics><mrow><mn>20</","PeriodicalId":94136,"journal":{"name":"Medical physics","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2025-04-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144061449","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Spatiotemporal coarse-to-fine diffusion model for automatic brain network generation.","authors":"Qiankun Zuo, Jiaojiao Yu, Conghuan Ye, Ling Chen, Hao Tian, Yixian Wu, Yudong Zhang","doi":"10.1002/mp.17833","DOIUrl":"https://doi.org/10.1002/mp.17833","url":null,"abstract":"<p><strong>Background: </strong>Functional magnetic resonance imaging (fMRI) has emerged as a transformative tool in analyzing and understanding brain diseases. It is a challenge to learn effective features from the high-dimensional fMRI. Most studies have focused on extracting connectivity-based features for disease analysis. However, they heavily rely on the software toolboxes to construct connectivity-based features, which may suffer from large errors because of different manual parameter settings and thus lead to bad performance in brain disorder analysis.</p><p><strong>Purpose: </strong>A novel brain denoiser model is proposed to transform four-dimensional fMRI (4D fMRI) into a brain network in a unified framework for brain disease analysis.</p><p><strong>Methods: </strong>By introducing anatomical knowledge, the proposed model first reduces the 4D fMRI into a 2D coarse region-of-interest(ROI)-based time series and then diffuses it into noisy status by gradually adding Gaussian noise. Moreover, the coarse-to-fine transformer refinement is designed to capture multi-scale temporal dynamics and iteratively remove unrelated multi-frequency noise. Besides, the low-frequency preservation module is devised to enhance the effective signal at low frequencies during the denoising process. This can improve the signal-to-noise ratio at each timestep, which ensures accurate restoration of ROI time series and improves the performance of brain network construction.</p><p><strong>Results: </strong>We evaluate the performance of the Brain Denoiser on the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset and the Autism Brain Imaging Data Exchange (ABIDE) dataset, demonstrating its ability to effectively suppress noise while preserving the underlying neural signals. Comparative analyses with related competing methods demonstrate the superiority of the proposed model.</p><p><strong>Conclusions: </strong>Generally, the proposed model presents a robust and innovative solution for brain network generation, paving the way for efficient analysis of brain disease.</p>","PeriodicalId":94136,"journal":{"name":"Medical physics","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2025-04-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144048953","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Effect of lung inflation states on chest CT image quality and pulmonary nodule detection with visualized respiratory Indicator.","authors":"Chengxin Kang, Tong Su, Binjie Fu, Yineng Zheng, Zhigang Chu, Guoshu Wang, Fajin Lv","doi":"10.1002/mp.17826","DOIUrl":"https://doi.org/10.1002/mp.17826","url":null,"abstract":"<p><strong>Background: </strong>Parts of lung cancer screening guidelines describe the specific scanning protocol of low dose CT (LDCT), among which the requirement for respiratory state is full inspiration end-breath hold. The main focus of lung cancer screening is to evaluate and follow-up pulmonary nodule (PN), so the display and detection of PNs are important. To achieve full inspiration, strict breathing training is required for patients. In clinical scans, the lung inflation state of patient is not visualized and the possibility of incomplete inspiration exists. Thus, the image quality and nodule detection of chest CT in different lung inflation states need to be explored.</p><p><strong>Methods: </strong>Fifty-six participants (32 females, 24 males) were included in this prospective study. Each participant underwent non-contrast chest CT scanned three times continually with different lung inflation state, including deep inspiration end-breath hold, calm breath hold, and deep expiration end-breath hold. A respiratory indicator was used to monitor the state of lung inflation visually. Subjective and objective image quality and nodule detection among these lung inflation states were analyzed in this study.</p><p><strong>Results: </strong>The images of deep inspiration end-breath hold yielded the best, with superior subjective ratings and objective image quality, including the lowest image noise and the best signal-to-noise ratio. PN detection was most accurate in the inflation state of deep inspiration end-breath hold, particularly for nodules ≤ 5 mm, while fewer nodules detected in the inflation state of calm breath hold and deep expiration end-breath hold.</p><p><strong>Conclusions: </strong>Lung inflation states significantly impact both image quality and PN detection in chest CT. Deep inspiration end-breath hold provided optimal image quality and nodule detection, while non-fully inflated states reduced diagnostic accuracy, especially for PNs≤5 mm. In clinical application, deep inspiration end-breath hold is recommended as the best inflation state of chest CT.</p>","PeriodicalId":94136,"journal":{"name":"Medical physics","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2025-04-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144061447","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Deep learning-based hippocampus asymmetry assessment for Alzheimer's disease diagnosis.","authors":"Fan Zhang, Yifan Wang, Xinhong Zhang","doi":"10.1002/mp.17831","DOIUrl":"https://doi.org/10.1002/mp.17831","url":null,"abstract":"<p><strong>Background: </strong>The symmetry of the brain hippocampus may be disrupted by natural aging and neurodegenerative diseases.</p><p><strong>Purpose: </strong>Currently, clinical studies on hippocampus asymmetry are limited to subjective visual evaluation and rough volume measurements, lacking quantitative standards.</p><p><strong>Methods: </strong>This paper proposes a quantitative assessment method of the hippocampus asymmetry based on deep learning, named DeepHAA (Deep Learning-based Hippocampus Asymmetry Assessment). The DeepHAA model extracts feature representations of left and right hippocampus structures in MRI images and achieved feature fusion through a cross-attention mechanism. A quantitative assessment method is proposed based on the distance between the multimodal embedding of the input sample and the reference embedding space.</p><p><strong>Results: </strong>The experimental dataset of this paper included MRI scans of 199 subjects, including 53 normal cognition (NC), 71 mild cognitive impairment (MCI) and 33 Alzheimer's disease (AD). The experimental results show that DeepHAA model can effectively identify and distinguish the NC, MCI, and AD.</p><p><strong>Conclusions: </strong>The proposed deep learning method integrates asymmetric information about hippocampus structure into the diagnosis of AD and has potential clinical application value.</p>","PeriodicalId":94136,"journal":{"name":"Medical physics","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2025-04-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144033624","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"The self-supervised fitting method based on similar neighborhood information of voxels for intravoxel incoherent motion diffusion-weighted MRI.","authors":"Lingfeng Luo, Chen Ye, Tianxian Li, Ming Zhong, Lihui Wang, Yuemin Zhu","doi":"10.1002/mp.17825","DOIUrl":"https://doi.org/10.1002/mp.17825","url":null,"abstract":"<p><strong>Background: </strong>The intravoxel incoherent motion (IVIM) parameter estimation is affected by noise, while existing CNN-based fitting methods utilize neighborhood spatial features around voxels to obtain more robust parameters. However, due to the heterogeneity of tissue, neighborhood features with low similarity can lead to excessively smooth parameter maps and even loss of tissue details.</p><p><strong>Purpose: </strong>To propose a novel neural network fitting approach, IVIM-CNN_similar, which utilizes similar neighborhood information of voxels to assist in the estimation of IVIM parameters in diffusion-weighted imaging (DWI).</p><p><strong>Methods: </strong>The proposed fitting model is based on convolutional neural network (CNN), which first identifies the similar neighborhoods of voxels through cluster analysis and then uses CNN to learn the spatial features of similar neighborhoods to reduce the impact of noise on the parameter estimation of the voxel. To evaluate the performance of the proposed method, comparisons were conducted with the least squares (LSQ), Bayesian, PI-DNN, and IVIM-CNN_unet algorithms on both simulated and in vivo brains, including 23 healthy brains and three brain tumors, in terms of root mean square error (RMSE) of IVIM parameters and the parameter contrast ratio between the tumor and normal regions.</p><p><strong>Results: </strong>The CNN-based methods, such as IVIM-CNN_similar and IVIM-CNN_unet, yield smoother parameter maps compared to voxel-based methods like nonlinear least squares, segmented nonlinear least squares, Bayesian, and PI-DNN. Additionally, the IVIM-CNN_similar retains more local tissue details while maintaining smoothness of parameter maps compared to the IVIM-CNN_unet. In simulated experiments, IVIM-CNN_similar outperforms IVIM-CNN_unet in terms of parameter estimation accuracy (SNR = 30; RMSE [ <math><semantics><mi>D</mi> <annotation>$D$</annotation></semantics> </math> ] = 0.0168 vs. 0.0253; RMSE ( <math><semantics><mi>F</mi> <annotation>$F$</annotation></semantics> </math> ) = 0.0001 vs. 0.0002; RMSE [ <math> <semantics><msup><mi>D</mi> <mo>∗</mo></msup> <annotation>$D^{*}$</annotation></semantics> </math> ] = 0.0266 vs. 0.0416). In addition, compared with other methods, the proposed IVIM-CNN_similar is more robust to noise, which is reflected in the lower RMSE of each parameter at different SNRs. For in vivo brains, compared to other methods, IVIM-CNN_similar achieved the highest PCR for most parameters when comparing the normal and tumor regions.</p><p><strong>Conclusions: </strong>The IVIM-CNN_similar method uses similar neighborhood information to assist IVIM parameter fitting by reducing the impact of noise on voxel parameter estimation, thereby improving the accuracy of parameter estimation and increasing the potential for IVIM clinical application.</p>","PeriodicalId":94136,"journal":{"name":"Medical physics","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2025-04-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144059498","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A CNN-transformer-based hybrid U-shape model with long-range relay for esophagus 3D CT image gross tumor volume segmentation.","authors":"Songli Yu, Yunxiang Li, Pengfei Jiao, Yixiu Liu, Jianxiang Zhao, Chenggang Yan, Qifeng Wang, Shuai Wang","doi":"10.1002/mp.17818","DOIUrl":"https://doi.org/10.1002/mp.17818","url":null,"abstract":"<p><strong>Background: </strong>Accurate and reliable segmentation of esophageal gross tumor volume (GTV) in computed tomography (CT) is beneficial for diagnosing and treating. However, this remains a challenging task because the esophagus has a variable shape and extensive vertical range, resulting in tumors potentially appearing at any position within it.</p><p><strong>Purpose: </strong>This study introduces a novel CNN-transformer-based U-shape model (LRRM-U-TransNet) designed to enhance the segmentation accuracy of esophageal GTV. By leveraging advanced deep learning techniques, we aim to address the challenges posed by the variable shape and extensive range of the esophagus, ultimately improving diagnostic and treatment outcomes.</p><p><strong>Methods: </strong>Specifically, we propose a long-range relay mechanism to converge all layer feature information by progressively passing adjacent layer feature maps in the pixel and semantic pathways. Moreover, we propose two ready-to-use blocks to implement this mechanism concretely. The Dual FastViT block interacts with feature maps from two paths to enhance feature representation capabilities. The Dual AxialViT block acts as a secondary auxiliary bottleneck to acquire global information for more precise feature map reconstruction.</p><p><strong>Results: </strong>We build a new esophageal tumor dataset with 1665 real-world patient CT samples annotated by five expert radiologists and employ multiple evaluation metrics to validate our model. Results of a five-fold cross-validation on this dataset show that LRRM-U-TransNet achieves a Dice coefficient of 0.834, a Jaccard coefficient of 0.730, a Precision of 0.840, a HD95 of 3.234 mm, and a Volume Similarity of 0.143.</p><p><strong>Conclusions: </strong>We propose a CNN-Transformer hybrid deep learning network to improve the segmentation effect of esophageal tumors. We utilize the local and global information between shallower and deeper layers to prevent early information loss and enhance the cross-layer communication. To validate our model, we collect a dataset composed of 1665 CT images of esophageal tumors from Sichuan Tumor Hospital. The results show that our model outperforms the state-of-the-art models. It is of great significance to improve the accuracy and clinical application of esophageal tumor segmentation.</p>","PeriodicalId":94136,"journal":{"name":"Medical physics","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2025-04-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144028456","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Automatic, machine-agnostic, convolution-based beam, and fluence modeling for Monte Carlo independent dose calculation.","authors":"Mingli Chen, Jingying Lin, Yang Park, Mu-Han Lin, Arnold Pompos, Andrew Godley, Weiguo Lu","doi":"10.1002/mp.17822","DOIUrl":"https://doi.org/10.1002/mp.17822","url":null,"abstract":"<p><strong>Background: </strong>Monte Carlo (MC)-based independent dose calculation is increasingly sought after for plan- and delivery-specific quality assurance (QA) in modern radiotherapy because of its high accuracy. It is particularly valuable for online adaptive radiotherapy, where measurement-based QA solutions are impractical. However, challenges related to beam modeling, commissioning, and plan/delivery-specific fluence calculation have hindered its widespread clinical adoption.</p><p><strong>Purpose: </strong>We propose a generic, automated, convolution-based beam and fluence modeling method for MC dose calculation, assuming zero or very limited knowledge of the linear accelerator (LINAC) head, with all necessary information derived from water phantom measurements. Instead of conventional particle transport through beam modulation devices (the phase space-based approach), we developed a direct convolution-based method to model the effects of beam modulation devices on output factors and fluence for downstream particle transport in the patient's body.</p><p><strong>Methods: </strong>The measurement data necessary for the beam model include the percent depth dose (PDD) profile of a reference field (typically 10 × 10 cm²), the diagonal profile of the largest field at the depth of maximum dose, and the output factors for representative field sizes formed by beam modulation devices (jaws/MLCs). The beam modeling process involves adjusting the energy spectrum to match the reference field PDD, optimizing the weighting factor for electron contamination, and encoding the output factors in a fluence convolution kernel. The fluence is calculated by convolving the intensity map defined by beam modulation devices and monitor units with the kernel, and the dose is calculated through a point source model with initial particles sampled from the fluence. This approach was demonstrated using an in-house developed general-purpose MC dose engine for various clinical LINACs, including those integrated with magnetic resonance imaging.</p><p><strong>Results: </strong>Compared to reference beam data, our calculations achieved average gamma passing rates of over 97% using the 2%/2 mm criteria. Compared to a sample of 20 clinical plans calculated by the treatment planning systems (TPS) across different beam modalities and treatment machines, our calculated dose achieved gamma passing rates of over 97% using the 3%/2 mm criteria with an average calculation time of less than 1 min.</p><p><strong>Conclusions: </strong>The proposed machine-agnostic, convolution-based beam, and fluence modeling approach enabled efficient automatic commissioning for a wide range of clinical external photon beam machines. The fluence map-based dose calculation approached sub-minute dose calculation efficiency for arbitrary treatment plans. The proposed method has the potential to accelerate the adoption of MC calculation-based QA for online adaptive radiotherapy.</p>","PeriodicalId":94136,"journal":{"name":"Medical physics","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2025-04-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144061245","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Biological adaptive radiotherapy for short-time dose compensation in lung SBRT patients.","authors":"Daisuke Kawahara, Akito S Koganezawa, Hikaru Yamaguchi, Takuya Wada, Yuji Murakami","doi":"10.1002/mp.17820","DOIUrl":"https://doi.org/10.1002/mp.17820","url":null,"abstract":"&lt;p&gt;&lt;strong&gt;Background: &lt;/strong&gt;Conventional adaptive radiation therapy (ART) primarily focuses on adapting to anatomical changes during radiation therapy but does not account for biological effects such as changes in radiosensitivity and tumor response, particularly during treatment interruptions. These interruptions may allow sublethal damage repair in tumor cells, reducing the effectiveness of stereotactic body radiation therapy (SBRT).&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Purpose: &lt;/strong&gt;The aim of this study was to develop and evaluate a novel biological adaptive radiotherapy (BART) framework to compensate for the biological effects of radiation interruptions during SBRT for lung cancer.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Methods: &lt;/strong&gt;This study involved lung SBRT patients using volumetric modulated arc therapy. We evaluated the biological dose loss using a microdosimetric kinetic model during four interruption durations (30, 60, 90, and 120 min). The reduction in the biological dose due to interruptions was calculated. The physical dose was calculated from the decreased biological dose in the in-house software, which was incorporated into the TPS. The optimization process was conducted for dose compensation in the TPS. To quantitatively assess the impact of BART on dose distribution, we evaluated the differences in target dose coverage and organ-at-risk (OAR) exposure between the original plan (without interruption), the plan with interruption, the BART plan, and the plan summing the dose before the interruption and the physical dose after compensation (compensated PD plan). The compensated PD plan assumed no biological dose reduction before the interruption.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Results: &lt;/strong&gt;Without BART compensation, interruptions of 30, 60, 90, and 120 min resulted in biological dose reductions, ranging from 12.1% to 19.0% for D&lt;sub&gt;50%&lt;/sub&gt; of the gross tumor volume (GTV) and from 16.4% to 24.9% for D&lt;sub&gt;98%&lt;/sub&gt; of the PTV. After applying BART, the differences were minimized to -1.5% to -0.6% for D&lt;sub&gt;50%&lt;/sub&gt; of the GTV and -0.1% to 0.9% for D&lt;sub&gt;98%&lt;/sub&gt; of the PTV. In contrast, the compensated PD plan exhibited larger residual deviations, with dose differences ranging from -9.9% to -14.0% for D&lt;sub&gt;50%&lt;/sub&gt; of the GTV and -12.3% to -7.3% for D&lt;sub&gt;98%&lt;/sub&gt; of the PTV. The volume differences between the BART plan and the plan without interruption remained within -0.8% to -0.4% for V&lt;sub&gt;5Gy&lt;/sub&gt; and -0.2% to 0.0% for V&lt;sub&gt;20Gy&lt;/sub&gt;, while differences between the BART and compensated PD plans were similarly small. The maximum dose to the spinal cord (D&lt;sub&gt;0.1cc&lt;/sub&gt;) also remained within -0.2 to 0.1 Gy for the BART plan relative to the plan without interruption and -0.1 to -0.5 Gy compared to the compensated PD plan. These results confirm that the OAR doses remained within clinically acceptable constraints across all evaluated plans.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Conclusion: &lt;/strong&gt;This study demonstrated that the BART framework effectively compensates for the biological d","PeriodicalId":94136,"journal":{"name":"Medical physics","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2025-04-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144048763","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}