Medical physics最新文献

{"title":"On the use of principal component analysis method to optimize sphere packing algorithm for lattice radiotherapy of large/bulky unresectable tumor","authors":"Joshua Misa, James R. Castle, Thomas A. Oldland, William St. Clair, Mark E. Bernard, Damodar Pokhrel","doi":"10.1002/mp.17982","DOIUrl":"https://doi.org/10.1002/mp.17982","url":null,"abstract":"<div>\u0000 \u0000 \u0000 <section>\u0000 \u0000 <h3> Background</h3>\u0000 \u0000 <p>Spatially Fractionated Radiotherapy (SFRT) delivers highly heterogenous dose distribution, characterized by an alternating pattern of high- and low-dose regions within the large tumor volume. Lattice SFRT (LRT) achieves this dose distribution via high-dose spheres arranged in a hexagonal pattern throughout the tumor. A major obstacle in LRT planning is optimizing the number of spheres within the tumor while maintaining the geometric constraints to allow for steep dose gradients and preserving normal tissue dose levels.</p>\u0000 </section>\u0000 \u0000 <section>\u0000 \u0000 <h3> Purpose</h3>\u0000 \u0000 <p>We present a novel strategy for lattice deployment for LRT using principal component analysis (PCA) to address the sphere packing problem. The aim of this report is improving sphere packing for LRT treatments will increase the volume of the peak dose the tumor receives from a given lattice configuration, potentially enhancing patient outcomes.</p>\u0000 </section>\u0000 \u0000 <section>\u0000 \u0000 <h3> Methods</h3>\u0000 \u0000 <p>Three lattice deployment methods were investigated. Our proposed method split-PCA (s-PCA), in which the lattice pattern is oriented split between the first and second principal axes, 1-PCA has the lattice pattern oriented based on the first principal axis, and n-PCA which does not utilize PCA to orientate the lattice pattern within the tumor. Thirty-five previously treated SFRT patients (15 Gy in 1 fraction) were replanned using each PCA method. All plans utilized four full VMAT arcs, offset collimator angles of ± 15°, 6MV-FFF energy, and a lattice diameter of 1.5 cm and spacing of 3 cm. The resulting plans were evaluated based on <i>D</i><sub>50%</sub>, <i>D</i><sub>mean</sub>, <i>V</i><sub>50%</sub>, <i>D</i><sub>10%</sub>, <i>D</i><sub>90%</sub>, peak-to-valley dose ratio (PVDR = <i>D</i><sub>10%</sub> ÷ <i>D</i><sub>90%</sub>), <i>D</i><sub>2cm</sub>, and <i>D</i><sub>max</sub> to nearby critical organs.</p>\u0000 </section>\u0000 \u0000 <section>\u0000 \u0000 <h3> Results</h3>\u0000 \u0000 <p>The s-PCA lattice plans had a statistically significant increase in the number of spheres packed within the tumor compared to the 1-PCA (Δ mean = 0.91, <i>p</i> = 0.019) and n-PCA (Δ mean = 1.43, <i>p</i> < 0.001). In addition, the s-PCA method outperformed the n-PCA method in terms of <i>D</i><sub>50%</sub>, <i>D</i><sub>mean</sub>, <i>V</i><sub>50%</sub>, <i>D</i><sub>10%</sub>, and <i>D</i><sub>90%</sub> and statistically outperformed 1-PCA in terms of <i>D</i><sub>50%</sub>, <i>D</i><sub>10%</sub>, and <i>D</i><sub>90%</sub>. However, the s-PCA gave a statistically significant decrease in PVDR in comparison to the 1-PCA (Δ mean = −0.94, <i>p</i> = 0.036) and the n-PCA (Δ mean = −2.21, <i>p</","PeriodicalId":18384,"journal":{"name":"Medical physics","volume":"52 7","pages":""},"PeriodicalIF":3.2,"publicationDate":"2025-07-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144635407","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Fine-tuning open-source large language models to improve their performance on radiation oncology tasks: A feasibility study to investigate their potential clinical applications in radiation oncology","authors":"Peilong Wang,&nbsp;Zhengliang Liu,&nbsp;Yiwei Li,&nbsp;Jason Holmes,&nbsp;Peng Shu,&nbsp;Lian Zhang,&nbsp;Xiang Li,&nbsp;Quanzheng Li,&nbsp;Brady S. Laughlin,&nbsp;Diego Santos Toesca,&nbsp;Carlos E. Vargas,&nbsp;Sujay A. Vora,&nbsp;Samir H. Patel,&nbsp;Terence T. Sio,&nbsp;Tianming Liu,&nbsp;Wei Liu","doi":"10.1002/mp.17985","DOIUrl":"https://doi.org/10.1002/mp.17985","url":null,"abstract":"<div>\u0000 \u0000 \u0000 <section>\u0000 \u0000 <h3> Background</h3>\u0000 \u0000 <p>The radiation oncology clinical practice involves many steps relying on the dynamic interplay of abundant text data. Large language models have displayed remarkable capabilities in processing complex text information. But their direct applications in specific fields like radiation oncology remain underexplored.</p>\u0000 </section>\u0000 \u0000 <section>\u0000 \u0000 <h3> Purpose</h3>\u0000 \u0000 <p>This study aims to investigate whether fine-tuning LLMs with domain knowledge can improve the performance on Task (1) treatment regimen generation, Task (2) treatment modality selection (photon, proton, electron, or brachytherapy), and Task (3) ICD-10 code prediction in radiation oncology.</p>\u0000 </section>\u0000 \u0000 <section>\u0000 \u0000 <h3> Methods</h3>\u0000 \u0000 <p>Data for 15 724 patient cases were extracted. Cases where patients had a single diagnostic record, and a clearly identifiable primary treatment plan were selected for preprocessing and manual annotation to have 7903 cases of the patient diagnosis, treatment plan, treatment modality, and ICD-10 code. Each case was used to construct a pair consisting of patient diagnostics details and an answer (treatment regimen, treatment modality, or ICD-10 code, respectively) for the supervised fine-tuning of these three tasks. Open source LLaMA2-7B and Mistral-7B models were utilized for the fine-tuning with the Low-Rank Approximations method. Accuracy and ROUGE-1 score were reported for the fine-tuned models and original models. Clinical evaluation was performed on Task (1) by radiation oncologists, while precision, recall, and F-1 score were evaluated for Task (2) and (3). One-sided Wilcoxon signed-rank tests were used to statistically analyze the results.</p>\u0000 </section>\u0000 \u0000 <section>\u0000 \u0000 <h3> Results</h3>\u0000 \u0000 <p>Fine-tuned LLMs outperformed original LLMs across all tasks with <i>p</i> value ≤ 0.001. Clinical evaluation demonstrated that over 60% of the fine-tuned LLMs-generated treatment regimens were clinically acceptable. Precision, recall, and F1-score showed improved performance of fine-tuned LLMs.</p>\u0000 </section>\u0000 \u0000 <section>\u0000 \u0000 <h3> Conclusion</h3>\u0000 \u0000 <p>Fine-tuned LLMs demonstrated statistically significant improvements over original LLMs upon three clinically important tasks in radiation oncology. This study explored the feasibility of applying fine-tuned LLMs in radiation oncology, inspiring further development of utilizing LLMs to assist with radiation oncology tasks.</p>\u0000 </section>\u0000 </div>","PeriodicalId":18384,"journal":{"name":"Medical physics","volume":"52 7","pages":""},"PeriodicalIF":3.2,"publicationDate":"2025-07-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144635422","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Explainable AI for raising confidence in deep learning-based tumor tracking models","authors":"Dragos Grama,&nbsp;Max Dahele,&nbsp;Ben Slotman,&nbsp;Wilko F. A. R. Verbakel","doi":"10.1002/mp.17940","DOIUrl":"https://doi.org/10.1002/mp.17940","url":null,"abstract":"<div>\u0000 \u0000 \u0000 <section>\u0000 \u0000 <h3> Background</h3>\u0000 \u0000 <p>Recently, tumor position monitoring using fluoroscopic images acquired during volumetric modulated arc therapy (VMAT) delivery has become available in a research setting. Accurate tracking during stereotactic body radiotherapy (SBRT), using VMAT, for lung tumors can help to ensure that the tumor is only irradiated when it is inside the planning target volume.</p>\u0000 </section>\u0000 \u0000 <section>\u0000 \u0000 <h3> Purpose</h3>\u0000 \u0000 <p>Traditionally, template matching is used to determine the tumor position, but with low tracking rates. A deep learning-based approach has the potential to improve this, but as deep learning is considered a “black box,” it would be desirable to know when to trust the predictions made by the model.</p>\u0000 </section>\u0000 \u0000 <section>\u0000 \u0000 <h3> Methods</h3>\u0000 \u0000 <p>We investigate the reliability and effectiveness of four explainable AI (XAI) methods (Guided Backpropagation (GBP), Layer-wise Relevance Propagation (LRP), DeepLIFT and PatternAttribtuion) to highlight the most relevant features for a deep learning-based 2D markerless lung tumor tracking model. The experiments are conducted on two phantoms and six clinical patients with small lung tumors (0.23–2.93 <span></span><math>\u0000 <semantics>\u0000 <msup>\u0000 <mtext>cm</mtext>\u0000 <mn>3</mn>\u0000 </msup>\u0000 <annotation>${text{cm}}^3$</annotation>\u0000 </semantics></math>). Both quantitative and qualitative evaluation is conducted to assess the suitability of the selected XAI methods for tumor tracking.</p>\u0000 </section>\u0000 \u0000 <section>\u0000 \u0000 <h3> Results</h3>\u0000 \u0000 <p>Our findings suggest that out of the four selected XAI methods, only GBP and DeepLIFT demonstrate a reliable and consistent behavior across all patients and phantoms; LRP shows good performance in the phantom setting but has lower qualitative results on the clinical data.</p>\u0000 </section>\u0000 \u0000 <section>\u0000 \u0000 <h3> Conclusions</h3>\u0000 \u0000 <p>Based on our results, we argue that GBP and DeepLIFT can be used out-of-the-box to explain deep learning-based tracking models for SBRT using VMAT. Further investigation is needed to develop a robust measure of the model's reliability in clinical practice during treatment delivery.</p>\u0000 </section>\u0000 </div>","PeriodicalId":18384,"journal":{"name":"Medical physics","volume":"52 7","pages":""},"PeriodicalIF":3.2,"publicationDate":"2025-07-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://onlinelibrary.wiley.com/doi/epdf/10.1002/mp.17940","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144624394","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Anti-jamming thermoacoustic imaging based on fiber Bragg grating ultrasonic detection and photoelectric conversion triggering","authors":"Yulong Guo,&nbsp;Lin Huang,&nbsp;Tiantao Hu,&nbsp;Wanting Peng,&nbsp;Kaili Jin,&nbsp;Liuge Cui,&nbsp;Ziqi Zhou,&nbsp;Yan Luo,&nbsp;Wenwu Ling","doi":"10.1002/mp.17944","DOIUrl":"https://doi.org/10.1002/mp.17944","url":null,"abstract":"&lt;div&gt;\u0000 \u0000 \u0000 &lt;section&gt;\u0000 \u0000 &lt;h3&gt; Background&lt;/h3&gt;\u0000 \u0000 &lt;p&gt;Thermoacoustic Imaging (TAI) combines the high contrast of microwave imaging with the high resolution of ultrasound imaging, establishing itself as a novel, non-invasive medical diagnostic technique. However, the high-power (peak power: 5∼100 kW; pulse width: 10∼1000 ns) used in TAI often cause remarkable interference with thermoacoustic signal acquisition systems and traditional piezoelectric ultrasound transducers. This interference leads to degraded thermoacoustic image quality and a pronounced “near-field dead zone” (NFDZ), which severely restricts the clinical application and development of TAI.&lt;/p&gt;\u0000 &lt;/section&gt;\u0000 \u0000 &lt;section&gt;\u0000 \u0000 &lt;h3&gt; Purpose&lt;/h3&gt;\u0000 \u0000 &lt;p&gt;To address these challenges, this paper proposes using fiber Bragg grating (FBG) for thermoacoustic signal detection. The high sensitivity and strong resistance to electromagnetic interference offered by FBG are leveraged to achieve interference-resistant signal acquisition. Additionally, a photoelectric conversion synchronous triggering method is adopted to prevent strong coupling of spatially distributed pulse microwave signals to the data acquisition system due to excessively long coaxial cables. This approach ensures more reliable and interference-resistant data acquisition for TAI.&lt;/p&gt;\u0000 &lt;/section&gt;\u0000 \u0000 &lt;section&gt;\u0000 \u0000 &lt;h3&gt; Methods&lt;/h3&gt;\u0000 \u0000 &lt;p&gt;Utilizing FBG with their high sensitivity and electromagnetic interference resistance, the system replaces traditional piezoelectric transducers for thermoacoustic signal detection. Additionally, a photoelectric conversion triggering method was implemented to avoid microwave pulse coupling interference caused by long coaxial cables. Experimental validation included comparisons between FBGs and traditional transducers in terms of signal-to-noise ratio (SNR), as well as noise measurements using both coaxial bayonet nut connector (BNC) cable and photoelectric trigger configurations to contrast test noise levels. Furthermore, the distance between the microwave antenna and data acquisition system was adjusted to validate the attenuation pattern of electromagnetic interference. TAI experiments used soy sauce tubes for their controllable composition/morphology. Systems imaging the tube also apply to human arms and ex vivo porcine liver (similar microwave absorption to tumors). This standardized, cost-efficient, and ethical approach provides a reliable basis for biological tissue imaging research.&lt;/p&gt;\u0000 &lt;/section&gt;\u0000 \u0000 &lt;section&gt;\u0000 \u0000 &lt;h3&gt; Results&lt;/h3&gt;\u0000 \u0000 &lt;p&gt;Experimental results show FBG with photoelectric triggering achieves SNR values of 28.32 (70.2% improvement) and 30.74 dB (66.3% improvement) compared to traditional transducers at 16.64 and ","PeriodicalId":18384,"journal":{"name":"Medical physics","volume":"52 7","pages":""},"PeriodicalIF":3.2,"publicationDate":"2025-07-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144624395","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"An unsupervised sparse-view CT reconstruction framework using combination of iterative deep image prior and ADMM","authors":"Jiahao Chang,&nbsp;Shuo Xu,&nbsp;Jintao Fu,&nbsp;Zirou Jiang,&nbsp;Zhentao Wang,&nbsp;Xincheng Xiang,&nbsp;Peng Cong,&nbsp;Yuewen Sun","doi":"10.1002/mp.17933","DOIUrl":"https://doi.org/10.1002/mp.17933","url":null,"abstract":"<div>\u0000 \u0000 \u0000 <section>\u0000 \u0000 <h3> Background</h3>\u0000 \u0000 <p>Sparse views and low-dose scanning reduce radiation exposure computed tomography (CT), but the reconstructed images exhibit severe artifacts and noise due to inadequate view sampling and diminished ray intensity. Recent advances in supervised deep learning (DL) methods have achieved remarkable success in CT reconstruction. However, their reliance on large datasets of paired high-quality and degraded images has constrained their applicability.</p>\u0000 </section>\u0000 \u0000 <section>\u0000 \u0000 <h3> Purpose</h3>\u0000 \u0000 <p>In this work, we introduce an unsupervised DL framework called ADMM-DRP, which integrates an untrained neural network with alternating direction method of multipliers (ADMM) iterative reconstruction algorithm.</p>\u0000 </section>\u0000 \u0000 <section>\u0000 \u0000 <h3> Methods</h3>\u0000 \u0000 <p>Specifically, the method employs an untrained neural network as an image generator to optimize the data inconsistency in the Radon domain. To avoid the overfitting phenomenon of traditional deep image prior (DIP)-based methods, we further utilize the ADMM with total variation (TV) regularization continuously update the input of the neural network during the training process.</p>\u0000 </section>\u0000 \u0000 <section>\u0000 \u0000 <h3> Results</h3>\u0000 \u0000 <p>Experiments on sparse-view and low-dose CT reconstruction tasks demonstrate that the proposed framework outperforms conventional supervised and iterative reconstruction methods in terms of metric and visual quality.</p>\u0000 </section>\u0000 \u0000 <section>\u0000 \u0000 <h3> Conclusions</h3>\u0000 \u0000 <p>ADMM-DRP reduces the algorithm's reliance on training data, achieves excellent performance in sparse-view and low-dose CT reconstruction, and demonstrates substantial potential for further development in medical imaging.</p>\u0000 </section>\u0000 </div>","PeriodicalId":18384,"journal":{"name":"Medical physics","volume":"52 7","pages":""},"PeriodicalIF":3.2,"publicationDate":"2025-07-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144624397","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Library of realistic 4D digital beating heart models based on patient CT data","authors":"Ethan J. Malin,&nbsp;Can Ceritoglu,&nbsp;Ehsan Abadi,&nbsp;J. Tilak Ratnanather,&nbsp;Ehsan Samei,&nbsp;W. Paul Segars","doi":"10.1002/mp.17945","DOIUrl":"https://doi.org/10.1002/mp.17945","url":null,"abstract":"<div>\u0000 \u0000 \u0000 <section>\u0000 \u0000 <h3> Background</h3>\u0000 \u0000 <p>The prevalence of cardiovascular disease (CVD) has risen alongside new medical imaging technologies designed for its diagnosis and treatment. Computational phantoms play a crucial role in imaging research, supporting applications ranging from basic simulation studies to larger-scale virtual imaging trials (VITs).</p>\u0000 </section>\u0000 \u0000 <section>\u0000 \u0000 <h3> Purpose</h3>\u0000 \u0000 <p>In this work, we develop a population of detailed, anatomically variable 4D beating heart models for medical imaging research.</p>\u0000 </section>\u0000 \u0000 <section>\u0000 \u0000 <h3> Methods</h3>\u0000 \u0000 <p>32 sets of 4D CT data from the PROspective Multicenter Imaging Study for Evaluation of Chest Pain (PROMISE) national clinical trial served as the basis for the cardiac library. Each dataset was electrocardiogram-gated, containing 10 lower-resolution frames over the cardiac cycle and one high-resolution frame at mid-diastole. The 4D data for each patient was segmented using the AI-based Automatic Segmentation (AS) Cardio tool from Synopsys Simpleware. The segmented high-resolution frame was used to define the initial instance of the heart, each structure defined as a polygon mesh. The multi-channel Large Deformation Diffeomorphic Metric Mapping (MC-LDDMM) image registration algorithm was then used to calculate the frame-to-frame motion of the heart from the low-resolution segmentations. The motion was applied to the cardiac model, creating a time-changing mesh model. Cubic spline curves were fit to the time-changing vertex locations, creating a 4D continuous model from which any number of time points can be generated. An example heart model was imported into a whole-body XCAT computational phantom and imaged with the DukeSim CT simulator for demonstration.</p>\u0000 </section>\u0000 \u0000 <section>\u0000 \u0000 <h3> Results</h3>\u0000 \u0000 <p>Compared to reference values, the image-based cardiac models mimic the twisting, contracting motion of the heart for anatomically variable subjects. When combined with DukeSim, realistic virtual cardiac imaging data can be produced.</p>\u0000 </section>\u0000 \u0000 <section>\u0000 \u0000 <h3> Conclusions</h3>\u0000 \u0000 <p>4D beating heart models were successfully created combining AI-based segmentation and image registration. The library of realistic cardiac models can be a vital tool for 4D cardiac imaging studies.</p>\u0000 </section>\u0000 </div>","PeriodicalId":18384,"journal":{"name":"Medical physics","volume":"52 7","pages":""},"PeriodicalIF":3.2,"publicationDate":"2025-07-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144624399","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"TrackRAD2025 challenge dataset: real-time tumor tracking for MRI-guided radiotherapy","authors":"Yiling Wang,&nbsp;Elia Lombardo,&nbsp;Adrian Thummerer,&nbsp;Tom Blöcker,&nbsp;Yu Fan,&nbsp;Yue Zhao,&nbsp;Christianna Iris Papadopoulou,&nbsp;Coen Hurkmans,&nbsp;Rob H. N. Tijssen,&nbsp;Pia A. W. Görts,&nbsp;Shyama U. Tetar,&nbsp;Davide Cusumano,&nbsp;Martijn PW Intven,&nbsp;Pim Borman,&nbsp;Marco Riboldi,&nbsp;Denis Dudáš,&nbsp;Hilary Byrne,&nbsp;Lorenzo Placidi,&nbsp;Marco Fusella,&nbsp;Michael Jameson,&nbsp;Miguel Palacios,&nbsp;Paul Cobussen,&nbsp;Tobias Finazzi,&nbsp;Cornelis J. A. Haasbeek,&nbsp;Paul Keall,&nbsp;Christopher Kurz,&nbsp;Guillaume Landry,&nbsp;Matteo Maspero","doi":"10.1002/mp.17964","DOIUrl":"https://doi.org/10.1002/mp.17964","url":null,"abstract":"<div>\u0000 \u0000 \u0000 <section>\u0000 \u0000 <h3> Purpose</h3>\u0000 \u0000 <p>Magnetic resonance imaging (MRI) to visualize anatomical motion is becoming increasingly important when treating cancer patients with radiotherapy. Hybrid MRI-linear accelerator (MRI-linac) systems allow real-time motion management during irradiation. This paper presents a multi-institutional real-time MRI time series dataset from different MRI-linac vendors. The dataset is designed to support developing and evaluating real-time tumor localization (tracking) algorithms for MRI-guided radiotherapy within the TrackRAD2025 challenge (\u0000https://trackrad2025.grand-challenge.org/).</p>\u0000 </section>\u0000 \u0000 <section>\u0000 \u0000 <h3> Acquisition and validation methods</h3>\u0000 \u0000 <p>The dataset consists of sagittal 2D cine MRIs (20-20543 frames per scan) in 585 patients from six centers (3 Dutch, 1 German, 1 Australian, and 1 Chinese). Tumors in the thorax, abdomen, and pelvis acquired on two commercially available MRI-linacs (0.35 T and 1.5 T) were included. For 108 cases, irradiation targets or tracking surrogates were manually segmented on each temporal frame. The dataset was randomly split into a public training set of 527 cases (477 unlabeled and 50 labeled) and a private testing set of 58 cases (all labeled).</p>\u0000 </section>\u0000 \u0000 <section>\u0000 \u0000 <h3> Data format and usage notes</h3>\u0000 \u0000 <p>The data is publicly available under the TrackRAD2025 collection: https://doi.org/10.57967/hf/4539. Both the images and segmentations for each patient are available in metadata format.</p>\u0000 </section>\u0000 \u0000 <section>\u0000 \u0000 <h3> Potential applications</h3>\u0000 \u0000 <p>This novel clinical dataset will enable the development and evaluation of real-time tumor localization algorithms for MRI-guided radiotherapy. By enabling more accurate motion management and adaptive treatment strategies, this dataset has the potential to advance the field of radiotherapy significantly.</p>\u0000 </section>\u0000 </div>","PeriodicalId":18384,"journal":{"name":"Medical physics","volume":"52 7","pages":""},"PeriodicalIF":3.2,"publicationDate":"2025-07-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://onlinelibrary.wiley.com/doi/epdf/10.1002/mp.17964","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144624410","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Assessment and mitigation of geometric distortions in Cartesian MR images at 15.2T for preclinical radiation research","authors":"Silvia Stocchiero,&nbsp;Isselmou Abdarahmane,&nbsp;Esaú Poblador Rodríguez,&nbsp;Vanessa Fröhlich,&nbsp;Markus Zeilinger,&nbsp;Dietmar Georg","doi":"10.1002/mp.17963","DOIUrl":"https://doi.org/10.1002/mp.17963","url":null,"abstract":"&lt;div&gt;\u0000 \u0000 \u0000 &lt;section&gt;\u0000 \u0000 &lt;h3&gt; Background&lt;/h3&gt;\u0000 \u0000 &lt;p&gt;Ultra-high field (UHF) magnetic resonance (MR) systems are advancing in preclinical imaging offering the potential to enhance radiation research. However, system-dependent factors, such as magnetic field inhomogeneities (&lt;span&gt;&lt;/span&gt;&lt;math&gt;\u0000 &lt;semantics&gt;\u0000 &lt;mrow&gt;\u0000 &lt;mi&gt;Δ&lt;/mi&gt;\u0000 &lt;msub&gt;\u0000 &lt;mi&gt;B&lt;/mi&gt;\u0000 &lt;mn&gt;0&lt;/mn&gt;\u0000 &lt;/msub&gt;\u0000 &lt;/mrow&gt;\u0000 &lt;annotation&gt;$Delta B_{0}$&lt;/annotation&gt;\u0000 &lt;/semantics&gt;&lt;/math&gt;) and gradient non-linearity (GNL), induce geometric distortions compromising the sub-millimeter accuracy required for radiation research.&lt;/p&gt;\u0000 &lt;/section&gt;\u0000 \u0000 &lt;section&gt;\u0000 \u0000 &lt;h3&gt; Purpose&lt;/h3&gt;\u0000 \u0000 &lt;p&gt;This study tackles system-dependent distortions in 15.2T MR images by prospective shimming strategies optimization and comparing two imaging methods for voxel displacement correction. The methods were evaluated on a 3D-printed grid phantom and validated on in vivo mouse brain MR images. Additionally, a phantom-based displacement map was tested for GNL correction in mouse brain images.&lt;/p&gt;\u0000 &lt;/section&gt;\u0000 \u0000 &lt;section&gt;\u0000 \u0000 &lt;h3&gt; Methods&lt;/h3&gt;\u0000 \u0000 &lt;p&gt;Phantom MR and CT images were acquired with 200&lt;span&gt;&lt;/span&gt;&lt;math&gt;\u0000 &lt;semantics&gt;\u0000 &lt;mrow&gt;\u0000 &lt;mi&gt;μ&lt;/mi&gt;\u0000 &lt;msup&gt;\u0000 &lt;mi&gt;m&lt;/mi&gt;\u0000 &lt;mn&gt;3&lt;/mn&gt;\u0000 &lt;/msup&gt;\u0000 &lt;/mrow&gt;\u0000 &lt;annotation&gt;$mu m^{3}$&lt;/annotation&gt;\u0000 &lt;/semantics&gt;&lt;/math&gt; resolution. In vivo mouse brain MR and CT images had 140&lt;span&gt;&lt;/span&gt;&lt;math&gt;\u0000 &lt;semantics&gt;\u0000 &lt;mrow&gt;\u0000 &lt;mi&gt;μ&lt;/mi&gt;\u0000 &lt;msup&gt;\u0000 &lt;mi&gt;m&lt;/mi&gt;\u0000 &lt;mn&gt;3&lt;/mn&gt;\u0000 &lt;/msup&gt;\u0000 &lt;/mrow&gt;\u0000 &lt;annotation&gt;$mu m^{3}$&lt;/annotation&gt;\u0000 &lt;/semantics&gt;&lt;/math&gt; and 200&lt;span&gt;&lt;/span&gt;&lt;math&gt;\u0000 &lt;semantics&gt;\u0000 &lt;mrow&gt;\u0000 &lt;mi&gt;μ&lt;/mi&gt;\u0000 &lt;msup&gt;\u0000 &lt;mi&gt;m&lt;/mi&gt;\u0000 &lt;mn&gt;3&lt;/mn&gt;\u0000 &lt;/msup&gt;\u0000 &lt;/mrow&gt;\u0000 &lt;annotation&gt;$mu m^{3}$&lt;/annotation&gt;\u0000 &lt;/semantics&gt;&lt;/math&gt; resolutions, respectively. Three shimming strategies were established to assess ","PeriodicalId":18384,"journal":{"name":"Medical physics","volume":"52 7","pages":""},"PeriodicalIF":3.2,"publicationDate":"2025-07-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://onlinelibrary.wiley.com/doi/epdf/10.1002/mp.17963","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144624411","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"AI-enabled precise brain tumor segmentation by integrating Refinenet and contour-constrained features in MRI images","authors":"Cheng Lv,&nbsp;Xu-Jun Shu,&nbsp;jun Qiu,&nbsp;Zi-cheng Xiong,&nbsp;Jing bo Ye,&nbsp;Shang bo Li,&nbsp;Sheng-Bo Chen,&nbsp;Hong Rao","doi":"10.1002/mp.17958","DOIUrl":"https://doi.org/10.1002/mp.17958","url":null,"abstract":"&lt;div&gt;\u0000 \u0000 \u0000 &lt;section&gt;\u0000 \u0000 &lt;h3&gt; Background&lt;/h3&gt;\u0000 \u0000 &lt;p&gt;Medical image segmentation is a fundamental task in medical image analysis and has been widely applied in multiple medical fields. The latest transformer-based deep learning segmentation model, Segment Anything Model (SAM), has demonstrated outstanding performance in natural image segmentation tasks through large-scale pre-training, achieving zero-shot image semantic understanding and pixel-level segmentation. However, medical images present challenges such as style variability, ill-defined object boundaries, and feature ambiguities, limiting the direct applicability of the SAM to medical image segmentation tasks.&lt;/p&gt;\u0000 &lt;/section&gt;\u0000 \u0000 &lt;section&gt;\u0000 \u0000 &lt;h3&gt; Purpose&lt;/h3&gt;\u0000 \u0000 &lt;p&gt;To enhance the robustness of the SAM in the domain of medical segmentation, we propose the SAM-RCCF framework. This approach aims to enhance the generalizability and precision of segmentation performance across diverse intracranial tumor types, including gliomas, metastatic tumors, and meningiomas.&lt;/p&gt;\u0000 &lt;/section&gt;\u0000 \u0000 &lt;section&gt;\u0000 \u0000 &lt;h3&gt; Materials and methods&lt;/h3&gt;\u0000 \u0000 &lt;p&gt;The study collected 484 axial T1-weighted contrast-enhanced (T1CE) magnetic resonance imaging (MRI) data of brain tumor patients, including 164 cases of glioma, 158 cases of metastatic tumors, and 162 cases of meningioma. All imaging data were randomly divided into training and testing sets. We employed the proposed SAM-RCCF model to perform segmentation experiments on these data, and five-fold cross-validation was adopted to evaluate the model's performance. This framework integrates the RefineNet module and the conditional control field with a conditional controller and Mask generator, enabling precise feature recognition and tailored segmentation for medical images, optimizing segmentation accuracy&lt;/p&gt;\u0000 &lt;/section&gt;\u0000 \u0000 &lt;section&gt;\u0000 \u0000 &lt;h3&gt; Results&lt;/h3&gt;\u0000 \u0000 &lt;p&gt;In the glioma segmentation experiment, the SAM-RCCF model achieved outstanding performance with an IOU of 0.90, DSC of 0.912, and HD of 13.13. For the meningioma segmentation task, it obtained an IOU of 0.9214, DSC of 0.93, and HD of 11.41, significantly outperforming other classic segmentation models.&lt;/p&gt;\u0000 &lt;/section&gt;\u0000 \u0000 &lt;section&gt;\u0000 \u0000 &lt;h3&gt; Conclusion&lt;/h3&gt;\u0000 \u0000 &lt;p&gt;The segmentation experiment results demonstrate that in the segmentation tasks of glioma, metastatic tumors, and meningioma MRI images, the SAM-RCCF algorithm significantly outperformed the original SAM in terms of DSC, HD, and IOU segmentation performance metrics. The experimental results verify the effectiveness of the SAM-RCCF framework in segmenting complex and variable brain tumor images, enhancing segmentation","PeriodicalId":18384,"journal":{"name":"Medical physics","volume":"52 7","pages":""},"PeriodicalIF":3.2,"publicationDate":"2025-07-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144624409","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Structural semantic-guided MR synthesis from PET images via a dual cross-attention mechanism","authors":"Hongyan Tang,&nbsp;Wenbo Li,&nbsp;Zhenxing Huang,&nbsp;Yaping Wu,&nbsp;Jianmin Yuan,&nbsp;Yang Yang,&nbsp;Yan Zhang,&nbsp;Yongfeng Yang,&nbsp;Hairong Zheng,&nbsp;Dong Liang,&nbsp;Meiyun Wang,&nbsp;Zhanli Hu","doi":"10.1002/mp.17957","DOIUrl":"https://doi.org/10.1002/mp.17957","url":null,"abstract":"&lt;div&gt;\u0000 \u0000 \u0000 &lt;section&gt;\u0000 \u0000 &lt;h3&gt; Background&lt;/h3&gt;\u0000 \u0000 &lt;p&gt;Multimodal medical imaging methods, such as positron emission tomography/computed tomography (PET/CT), are widely used for diagnosing diseases because they provide both structural and functional information. However, PET/CT has limitations in terms of visualizing soft tissues, particularly for brain diseases, which highlights the need for magnetic resonance imaging (MRI).&lt;/p&gt;\u0000 &lt;/section&gt;\u0000 \u0000 &lt;section&gt;\u0000 \u0000 &lt;h3&gt; Purpose&lt;/h3&gt;\u0000 \u0000 &lt;p&gt;Given the limited adoption of PET/ magnetic resonance (MR) devices for making MR images available and the discomfort of elderly cancer patients during long-term MR scanning, a promising solution is to develop methods for synthesizing MR images from other modalities. While previous research has focused mainly on structure-to-structure modality transitions, such as CT-to-MR synthesis, our study aims to explore a new function-to-structure transition approach to realize PET-to-MR synthesis. Specifically, we propose a structural semantic-guided deep learning network to synthesize MR images from PET data to simplify medical imaging processes, improving both efficiency and accessibility.&lt;/p&gt;\u0000 &lt;/section&gt;\u0000 \u0000 &lt;section&gt;\u0000 \u0000 &lt;h3&gt; Methods&lt;/h3&gt;\u0000 \u0000 &lt;p&gt;We propose a structural semantic-guided deep learning network with a dual cross-attention (DCA) module to synthesize MR images from PET data for realizing the function-to-structure modality transition. The network introduces a structural semantic loss to preserve structural information and details, and the DCA module utilizes cross-attention to effectively capture the channel and spatial interdependencies among multiscale features. The proposed method was compared with other deep learning-based methods, including 3DUXNET, UNETR, nnFormer, CycleGAN, Pix2pix, edge-aware generative adversarial network (Ea-GAN), and MedNet. Additionally, visual and quantitative analysis was employed to evaluate the model performance. Furthermore, correlation analysis based on pixel averages, semantic assessment, and additional data assessment was performed for the quantitative evaluation of image synthesis results. Additionally, an ablation experiment was conducted to validate the effectiveness of introducing structural semantic loss and the DCA module in enhancing model performance.&lt;/p&gt;\u0000 &lt;/section&gt;\u0000 \u0000 &lt;section&gt;\u0000 \u0000 &lt;h3&gt; Results&lt;/h3&gt;\u0000 \u0000 &lt;p&gt;The experiments demonstrate that the proposed method yields superior visual and quantitative outcomes, with a peak signal-to-noise ratio (PSNR) of 29.09 dB, a structural similarity index measure (SSIM) of 0.8417, and a mean absolute error (MAE) of 0.0296. Additionally, the correlation analysis based on pixel averages shows a fitted slope of 0.957 in the","PeriodicalId":18384,"journal":{"name":"Medical physics","volume":"52 7","pages":""},"PeriodicalIF":3.2,"publicationDate":"2025-07-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144624473","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}