Andrew Lian, Trevor Ketcherside, An Liu, Chunhui Han, Karine A Al Feghali, Arjun Maniyedath, Arya Amini, Colton Ladbury
{"title":"Evaluation of the stability of radiomic features of non-irradiated organs utilizing fan-beam kilovoltage computed tomography.","authors":"Andrew Lian, Trevor Ketcherside, An Liu, Chunhui Han, Karine A Al Feghali, Arjun Maniyedath, Arya Amini, Colton Ladbury","doi":"10.1002/mp.17914","DOIUrl":"https://doi.org/10.1002/mp.17914","url":null,"abstract":"<p><strong>Background: </strong>Although radiation oncologists obtain non-contrast computed tomography (CT) images in every treatment, their quality is often insufficient for radiomic analyses for monitoring response over the course of treatment. Newer linear accelerators, such as RefleXion X1, have higher-quality imaging, creating new opportunities for radiomic analysis throughout treatment.</p><p><strong>Purpose: </strong>To best utilize the high-quality kilovoltage computed tomography (kVCT) scans of RefleXion X1 for radiomic analyses of cancerous tissue, radiomic features that remain consistent through treatment in normal organs must first be identified. Stable features can be used for normalization to calculate radiomic features in tumors, enabling monitoring of response during treatment and early adaptation if required.</p><p><strong>Methods: </strong>The kVCT localization images acquired over the course of treatment on RefleXion X1 were analyzed for a total of five patients. A total of five patients were scanned using the RefleXion X1 throughout treatment. The imaging used standardized acquisition parameters for all treatments to minimize variation. Images were acquired using \"Body/Medium Dose/Slow Couch\" parameters. The regions of interest (ROIs) for each organ were automatically segmented using an auto-segmentation system Medical Mind Inc. Daily CT images and structure files were imported into an Image Biomarker Standardization Initiative (IBSI) compliant radiomic software package (LifeX) to extract radiomic features. Four non-irradiated organs were used to analyze the repeatability of normal tissues: liver, spleen, heart, and spinal cord. The intraclass correlation coefficient (ICC) using a 2-way mixed-effects model was used to measure repeatability, while the concordance correlation coefficient (CCC) was used to measure reproducibility.</p><p><strong>Results: </strong>Cutoff values were applied to the average ICC across patients and the average CCC across both patients and fractions. Forty features were identified with a cutoff value of 0.8, accounting for 82% of the original features. Using a cutoff value of 0.9, the subset of stable features was further reduced to 29, representing 59% of the original features.</p><p><strong>Conclusions: </strong>A subset of several radiomic features extracted remained stable throughout treatment. Thus, radiomic analyses of cancerous tissue using RefleXion X1 imaging throughout treatment would be feasible as an ongoing assessment of response during treatment for personalized adaptive approaches.</p>","PeriodicalId":94136,"journal":{"name":"Medical physics","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2025-05-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144176268","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}
Edward R Criscuolo, Zhendong Zhang, Yao Hao, Deshan Yang
{"title":"A vessel bifurcation landmark pair dataset for abdominal CT deformable image registration (DIR) validation.","authors":"Edward R Criscuolo, Zhendong Zhang, Yao Hao, Deshan Yang","doi":"10.1002/mp.17907","DOIUrl":"10.1002/mp.17907","url":null,"abstract":"<p><strong>Purpose: </strong>Deformable image registration (DIR) is an enabling technology in many diagnostic and therapeutic tasks. Despite this, DIR algorithms have limited clinical use, largely due to a lack of benchmark datasets for quality assurance during development. DIRs of intra-patient abdominal CTs are among the most challenging registration scenarios due to significant organ deformations and inconsistent image content. To support future algorithm development, here we introduce our first-of-its-kind abdominal CT DIR benchmark dataset, comprising large numbers of highly accurate landmark pairs on matching blood vessel bifurcations.</p><p><strong>Acquisition and validation methods: </strong>Abdominal CT image pairs of 30 patients were acquired from several publicly available repositories as well as the authors' institution with IRB approval. The two CTs of each pair were originally acquired for the same patient but on different days. An image processing workflow was developed and applied to each CT image pair: (1) Abdominal organs were segmented with a deep learning model, and image intensity within organ masks was overwritten. (2) Matching image patches were manually identified between two CTs of each image pair. (3) Vessel bifurcation landmarks were labeled on one image of each image patch pair. (4) Image patches were deformably registered, and landmarks were projected onto the second image. (5) Landmark pair locations were refined manually or with an automated process. This workflow resulted in 1895 total landmark pairs, or 63 per case on average. Estimates of the landmark pair accuracy using digital phantoms were 0.7 mm ± 1.2 mm.</p><p><strong>Data format and usage notes: </strong>The data are published in Zenodo at https://doi.org/10.5281/zenodo.14362785. Instructions for use can be found at https://github.com/deshanyang/Abdominal-DIR-QA.</p><p><strong>Potential applications: </strong>This dataset is a first-of-its-kind for abdominal DIR validation. The number, accuracy, and distribution of landmark pairs will allow for robust validation of DIR algorithms with precision beyond what is currently available.</p>","PeriodicalId":94136,"journal":{"name":"Medical physics","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2025-05-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144176261","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}
Yang Lei, Ming Chao, Kaida Yang, Vishal Gupta, Emi J Yoshida, Tingyu Wang, Xiaofeng Yang, Tian Liu
{"title":"A novel network architecture for post-applicator placement CT auto-contouring in cervical cancer HDR brachytherapy.","authors":"Yang Lei, Ming Chao, Kaida Yang, Vishal Gupta, Emi J Yoshida, Tingyu Wang, Xiaofeng Yang, Tian Liu","doi":"10.1002/mp.17908","DOIUrl":"https://doi.org/10.1002/mp.17908","url":null,"abstract":"<p><strong>Background: </strong>High-dose-rate brachytherapy (HDR-BT) is an integral part of treatment for locally advanced cervical cancer, requiring accurate segmentation of the high-risk clinical target volume (HR-CTV) and organs at risk (OARs) on post-applicator CT (pCT) for precise and safe dose delivery. Manual contouring, however, is time-consuming and highly variable, with challenges heightened in cervical HDR-BT due to complex anatomy and low tissue contrast. An effective auto-contouring solution could significantly enhance efficiency, consistency, and accuracy in cervical HDR-BT planning.</p><p><strong>Purpose: </strong>To develop a machine learning-based approach that improves the accuracy and efficiency of HR-CTV and OAR segmentation on pCT images for cervical HDR-BT.</p><p><strong>Methods: </strong>The proposed method employs two sequential deep learning models to segment target and OARs from planning CT data. The intuitive model, a U-Net, initially segments simpler structures such as the bladder and HR-CTV, utilizing shallow features and iodine contrast agents. Building on this, the sophisticated model targets complex structures like the sigmoid, rectum, and bowel, addressing challenges from low contrast, anatomical proximity, and imaging artifacts. This model incorporates spatial information from the intuitive model and uses total variation regularization to improve segmentation smoothness by applying a penalty to changes in gradient. This dual-model approach improves accuracy and consistency in segmenting high-risk clinical target volumes and organs at risk in cervical HDR-BT. To validate the proposed method, 32 cervical cancer patients treated with tandem and ovoid (T&O) HDR brachytherapy (3-5 fractions, 115 CT images) were retrospectively selected. The method's performance was assessed using four-fold cross-validation, comparing segmentation results to manual contours across five metrics: Dice similarity coefficient (DSC), 95% Hausdorff distance (HD<sub>95</sub>), mean surface distance (MSD), center-of-mass distance (CMD), and volume difference (VD). Dosimetric evaluations included D90 for HR-CTV and D2cc for OARs.</p><p><strong>Results: </strong>The proposed method demonstrates high segmentation accuracy for HR-CTV, bladder, and rectum, achieving DSC values of 0.79 ± 0.06, 0.83 ± 0.10, and 0.76 ± 0.15, MSD values of 1.92 ± 0.77 mm, 2.24 ± 1.20 mm, and 4.18 ± 3.74 mm, and absolute VD values of 5.34 ± 4.85 cc, 17.16 ± 17.38 cc, and 18.54 ± 16.83 cc, respectively. Despite challenges in bowel and sigmoid segmentation due to poor soft tissue contrast in CT and variability in manual contouring (ground truth volumes of 128.48 ± 95.9 cc and 51.87 ± 40.67 cc), the method significantly outperforms two state-of-the-art methods on DSC, MSD, and CMD metrics (p-value < 0.05). For HR-CTV, the mean absolute D90 difference was 0.42 ± 1.17 Gy (p-value > 0.05), less than 5% of the prescription dose. Over 75% of cases showed changes within ± 0.5 G","PeriodicalId":94136,"journal":{"name":"Medical physics","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2025-05-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144145309","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}
Remo Cristoforetti, Jennifer Josephine Hardt, Niklas Wahl
{"title":"Scenario-free robust optimization algorithm for IMRT and IMPT treatment planning.","authors":"Remo Cristoforetti, Jennifer Josephine Hardt, Niklas Wahl","doi":"10.1002/mp.17905","DOIUrl":"https://doi.org/10.1002/mp.17905","url":null,"abstract":"<p><strong>Background: </strong>Robust treatment planning algorithms for intensity modulated proton therapy (IMPT) and intensity modulated radiation therapy (IMRT) allow for uncertainty reduction in the delivered dose distributions through explicit inclusion of error scenarios. Due to the curse of dimensionality, application of such algorithms can easily become computationally prohibitive.</p><p><strong>Purpose: </strong>This work proposes a scenario-free probabilistic robust optimization algorithm that overcomes both the runtime and memory limitations typical of traditional robustness algorithms.</p><p><strong>Methods: </strong>The scenario-free approach minimizes cost-functions evaluated on expected-dose distributions and total variance. Calculation of these quantities relies on precomputed expected-dose-influence and total-variance-influence matrices, such that no scenarios need to be stored for optimization. The algorithm is developed within matRad and tested in several optimization configurations for photon and proton irradiation plans. A traditional robust optimization algorithm and a margin-based approach are used as a reference to benchmark the performance of the scenario-free algorithm in terms of plan quality, robustness, and computational workload.</p><p><strong>Results: </strong>The implemented scenario-free approach achieves plan quality similar to traditional robust optimization algorithms, and it reduces the distribution of standard deviation within selected structures when variance reduction objectives are defined. Avoiding the storage of individual scenario information allows for the solution of treatment plan optimization problems, including an arbitrary number of error scenarios. The observed computational time required for optimization is close to a nominal, non-robust algorithm and substantially lower compared to the traditional robust approach. Estimated gains in relative runtime range from approximately <math> <semantics><mrow><mn>5</mn></mrow> <annotation>$hskip.001pt 5$</annotation></semantics> </math> - <math> <semantics><mrow><mn>600</mn></mrow> <annotation>$hskip.001pt 600$</annotation></semantics> </math> times with respect to the traditional approach.</p><p><strong>Conclusion: </strong>This work introduces a novel scenario-free optimization approach relying on the precomputation of probabilistic quantities while preserving compatibility with state-of-the-art uncertainty modeling. The measured runtime and memory footprint are independent of the number of included error scenarios and similar to those of non-robust margin-based optimization algorithms, while achieving the required dose and robustness specifications under multiple different optimization conditions. These properties make the scenario-free approach suitable and beneficial for 3D and 4D robust optimization involving a high number of error scenarios and/or CT phases.</p>","PeriodicalId":94136,"journal":{"name":"Medical physics","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2025-05-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144145320","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}
Iymad R Mansour, Christian Valdes-Cortez, David Santiago Ayala Alvarez, Francisco Berumen, Jean-Simon Côte, Gaël Ndoutoume-Paquet, Peter G F Watson, Jan Seuntjens, Facundo Ballester, Ernesto Mainegra-Hing, Rowan M Thomson, Luc Beaulieu, Javier Vijande
{"title":"Reference datasets for commissioning of model-based dose calculation algorithms for electronic brachytherapy.","authors":"Iymad R Mansour, Christian Valdes-Cortez, David Santiago Ayala Alvarez, Francisco Berumen, Jean-Simon Côte, Gaël Ndoutoume-Paquet, Peter G F Watson, Jan Seuntjens, Facundo Ballester, Ernesto Mainegra-Hing, Rowan M Thomson, Luc Beaulieu, Javier Vijande","doi":"10.1002/mp.17872","DOIUrl":"https://doi.org/10.1002/mp.17872","url":null,"abstract":"<p><strong>Purpose: </strong>This work provides the first two clinical test cases for commissioning electronic brachytherapy (eBT) model-based dose calculation algorithms (MBDCAs) for skin irradiation using surface applicators.</p><p><strong>Acquisition and validation methods: </strong>The test cases utilize the INTRABEAM 30 mm surface applicator. Test Case I: water phantom is used to evaluate the algorithm's performance in a uniform medium consisting of a voxelized water cube surrounded by air. Test Case II: Surface eBT represents a heterogeneous medium with four distinct layers: skin tissue, adipose tissue, cortical bone, and soft tissue. Treatment plans for both cases were created and exported into the Radiance treatment planning system (TPS). Dose-to-medium calculations were then performed using this Monte Carlo (MC)-based TPS and compared with MC simulations conducted independently by three different groups using two codes: EGSnrc and PENELOPE. The results agreed within expected Type A and B statistical uncertainties.</p><p><strong>Data format and usage notes: </strong>The dataset is available online at https://doi.org/10.52519/00005. A proprietary file designed for use within Radiance containing CT images and the treatment plan for both test cases, the LINAC modeling, and the CT calibration are included, as well as reference MC and TPS dose data in RTdose format and all files required to run the MC simulations.</p><p><strong>Potential applications: </strong>This dataset serves as a valuable resource for commissioning eBT MBDCAs and lays the groundwork for developing clinical test cases for other eBT systems. It is also a helpful educational tool for exploring various eBT devices and their advantages and drawbacks. Furthermore, brachytherapy researchers seeking a benchmark for dosimetric calculations in the low-energy domain will find this dataset indispensable.</p>","PeriodicalId":94136,"journal":{"name":"Medical physics","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2025-05-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144145315","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}
Jan Sebek, Pinyo Taeprasartsit, Chanok Pathomparai, Damian E Dupuy, Henky Wibowo, Punit Prakash
{"title":"Computational modeling of microwave ablation of lung tumors: Assessment of model-predictions against post-treatment imaging.","authors":"Jan Sebek, Pinyo Taeprasartsit, Chanok Pathomparai, Damian E Dupuy, Henky Wibowo, Punit Prakash","doi":"10.1002/mp.17897","DOIUrl":"https://doi.org/10.1002/mp.17897","url":null,"abstract":"<p><strong>Background: </strong>Percutaneous microwave ablation is a clinically established method for treatment of unresectable lung nodules. When planning the intervention, the size of ablation zone, which should encompass the nodule as well as a surrounding margin of normal tissue, is predicted via manufacturer-provided geometric models, which do not consider patient-specific characteristics. However, the size and shape of ablation is dependent on tissue composition and properties and can vary between patients.</p><p><strong>Purpose: </strong>To comparatively assess performance of a computational model-based approach and manufacturer geometric model for predicting extent of ablation zones following microwave lung ablation procedures on a retrospectively collected clinical imaging dataset.</p><p><strong>Methods: </strong>A retrospective computed-tomography (CT) imaging dataset was assembled of 50 patients treated with microwave ablation of lung tumors at a single institution. For each case, the dataset consisted of a pre-procedure CT acquired without the ablation applicator, a peri-procedure CT scan with the ablation applicator in position, and post-procedure CT scan to assess the ablation zone extent acquired on the first follow-up visit. A physics-based computational model of microwave absorption and bioheat transfer was implemented using the finite element method, with the model geometry incorporating key tissue types within 2 cm of the applicator as informed by imaging data. The model-predicted extent of the ablation zone was estimated using the Arrhenius thermal damage model. The ablation zone predicted by the manufacturer geometric model consisted of an ellipsoid registered with the applicator position and dimensions provided by instructions for use documentation. Both ablation estimates were compared to ground truth ablation zone segmented from post-procedure CT via Dice similarity coefficient (DSC) and average absolute error (AAE). The statistically significant difference at level 0.05 in performance between both ablation prediction methods was tested with permutation test on DSC as well as AAE datasets with applied Bonferroni multiple-comparison correction.</p><p><strong>Results: </strong>Receiver operating characteristic analysis of the predictive power of the volume of insufficient coverage (i.e. tumor volume which did not receive an ablative thermal dose) as an indicator of local tumor recurrence yielded an area under the curve of 0.84, illustrating the clinical significance of accurate prediction of ablation zone extents. Across all cases, AAEs were 3.65 ± 1.12 mm, and 5.11 ± 1.93 mm for patient-specific computational and manufacturer geometric models respectively. Similarly, average DSCs were 0.55 ± 0.14, and 0.46 ± 0.19 for computational and manufacturer geometric models respectively. The manufacturer geometric model overpredicted volume of the ablation zone compared to ground truth by 141% on average, whereas the patient-specifi","PeriodicalId":94136,"journal":{"name":"Medical physics","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2025-05-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144129817","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}
Deng-Yuan Chang, Joseph P Speth, Matthew L Scarpelli
{"title":"Evaluating the theranostic potential of ferumoxytol when combined with radiotherapy in a mammary dual tumor mouse model.","authors":"Deng-Yuan Chang, Joseph P Speth, Matthew L Scarpelli","doi":"10.1002/mp.17888","DOIUrl":"https://doi.org/10.1002/mp.17888","url":null,"abstract":"<p><strong>Background: </strong>The radiation-induced abscopal effect (RIAE) is a desirable phenomenon involving radiation-induced activation of the immune system and regression of metastatic disease after local radiotherapy. However, the majority of patients undergoing radiotherapy do not experience abscopal responses. One potential barrier to the RIAE is tumor-associated macrophages (TAMs), which can be recruited to the tumor after radiotherapy and have an immunosuppressive effect on the tumor microenvironment (TME).</p><p><strong>Purpose: </strong>We aim to evaluate the dual capabilities of the FDA-approved iron nanoparticle ferumoxytol for (1) enhancing the RIAE and (2) measuring TAMs by magnetic resonance imaging (MRI). We hypothesized that (1) the immunomodulating effect of ferumoxytol could enhance the RIAE by repolarizing the M2 TAMs to M1 TAMs, and (2) the TAMs could be non-invasively imaged by ferumoxytol-MRI.</p><p><strong>Methods: </strong>Twenty-eight BALB/c mice were subcutaneously implanted with 4T1 primary orthotopic tumor (mammary fat pad) and flank tumor (abscopal tumor). At 14 days post-implantation, mice were separated into four groups: control (Ctrl), radiotherapy (RT) only (8-Gy×3), ferumoxytol only (FMX; 40 mg/kg) and combined (Comb) group (a single dose of 40 mg/kg FMX 24 h prior to 8-Gy×3) (n = 7 mice per group; 56 tumors). At 23- and 24-day post-implantation the pre- and post-FMX injection MRI was performed for mice in FMX and Comb group. The percent change in transverse relaxation time (%T2*) from pre to post ferumoxytol injection was calculated from MR images for both tumors and lymph nodes (LNs). At 25 days post-implantation, both tumors were harvested, and the TAMs were analyzed by flow cytometry.</p><p><strong>Results: </strong>At 25 days post-implantation, the primary tumor volume in the RT and Comb groups was significantly lower than the Ctrl and FMX groups (p < 0.05). No significant size difference of abscopal tumors was observed among all groups. In addition, there was no significant difference in lung metastasis nodules. A significant decrease in %T2* values of tumors and LNs in the FMX and Comb group 24 h post-ferumoxytol injection was observed, suggesting ferumoxytol uptake in TAMs. The flow cytometry result showed that the CD80<sup>+</sup> CD206<sup>-</sup> M1 macrophage population was similar among all tumors and groups. The CD80<sup>-</sup> CD206<sup>+</sup> M2 macrophage population was also similar in all tumors and groups, with the exception of the FMX group, where the M2 tumor macrophage levels were significantly higher when compared to the Ctrl group (p < 0.05). Tumors in the FMX group had a significant negative Pearson correlation between tumor %T2* change and M1 tumor macrophage levels (r = -0.76, p < 0.05) but this correlation was not significant in any other treatment group.</p><p><strong>Conclusions: </strong>Radiotherapy combined with ferumoxytol led to significant growth delays of irradiated tum","PeriodicalId":94136,"journal":{"name":"Medical physics","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2025-05-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144121819","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":"Diffusion Schrödinger bridge models for high-quality MR-to-CT synthesis for proton treatment planning.","authors":"Muheng Li, Xia Li, Sairos Safai, Antony J Lomax, Ye Zhang","doi":"10.1002/mp.17898","DOIUrl":"https://doi.org/10.1002/mp.17898","url":null,"abstract":"<p><strong>Background: </strong>In recent advancements in proton therapy, magnetic resonance (MR)-based treatment planning is gaining momentum due to its excellent soft tissue contrast and high potential to minimize extra radiation exposure compared to traditional computed tomography (CT)-based methods. This transition underscores the critical need for accurate MR-to-CT image synthesis, which is essential for precise proton dose calculations.</p><p><strong>Purpose: </strong>This study aims to introduce and evaluate the diffusion Schrödinger bridge models (DSBM), an innovative approach for high-quality and efficient MR-to-CT synthesis, in order to improve both the quality and speed of synthetic CT (sCT) image generation.</p><p><strong>Methods: </strong>The DSBM learns the nonlinear diffusion processes between MR and CT data distributions. Unlike traditional diffusion models (DMs), which start synthesis from a Gaussian distribution, DSBM starts from the prior distribution, enabling more direct and efficient synthesis. The model was trained on 46 head-and-neck (HN) MR-CT pairs and 77 brain tumor MR-CT pairs, with 8 and 10 scans used for testing, respectively. Comprehensive evaluations were conducted at both image and dosimetric levels, using metrics such as mean absolute error (MAE), Dice score, voxel-wise proton dose differences, gamma pass rates of clinical plans, and typical dose indices.</p><p><strong>Results: </strong>For the HN dataset, DSBM achieved a lower MAE of 72.42 <math><semantics><mo>±</mo> <annotation>$pm$</annotation></semantics> </math> 9.78 Hounsfield unit (HU) compared to 77.72 <math><semantics><mo>±</mo> <annotation>$pm$</annotation></semantics> </math> 9.11 HU with the best baseline approach, and a higher Dice score for bone of 83.32 <math><semantics><mo>±</mo> <annotation>$pm$</annotation></semantics> </math> 3.25% compared to 82.55 <math><semantics><mo>±</mo> <annotation>$pm$</annotation></semantics> </math> 3.62%, indicating superior anatomical accuracy. Dosimetric evaluations showed a 1%/1 mm gamma pass rate of 95.85 <math><semantics><mo>±</mo> <annotation>$pm$</annotation></semantics> </math> 2.99%, surpassing the 95.25 <math><semantics><mo>±</mo> <annotation>$pm$</annotation></semantics> </math> 3.09% achieved by the baseline model. For the brain tumor dataset, DSBM outperformed the baseline with an MAE of 91.73 <math><semantics><mo>±</mo> <annotation>$pm$</annotation></semantics> </math> 6.86 HU compared to 103.25 <math><semantics><mo>±</mo> <annotation>$pm$</annotation></semantics> </math> 9.58 HU, and a Dice score for bone of 82.85 <math><semantics><mo>±</mo> <annotation>$pm$</annotation></semantics> </math> 3.88% compared to 81.27 <math><semantics><mo>±</mo> <annotation>$pm$</annotation></semantics> </math> 4.59%. DSBM also demonstrated a higher 1%/1 mm gamma pass rate of 97.93 <math><semantics><mo>±</mo> <annotation>$pm$</annotation></semantics> </math> 1.82%, confirming its robustness across different anatomical regi","PeriodicalId":94136,"journal":{"name":"Medical physics","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2025-05-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144121817","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}
Tonglong Li, Minheng Chen, Mingying Li, Chuanyou Li, Youyong Kong
{"title":"Automatic x-ray to CT registration using embedding reconstruction and lite cross-attention.","authors":"Tonglong Li, Minheng Chen, Mingying Li, Chuanyou Li, Youyong Kong","doi":"10.1002/mp.17896","DOIUrl":"https://doi.org/10.1002/mp.17896","url":null,"abstract":"<p><strong>Background: </strong>The registration of intraoperative x-ray images with preoperative CT images is an important step in image-guided surgery. However, existing regression-based methods lack an interpretable and stable mechanism when fusing information from intraoperative images and preoperative CT volumes. In addition, existing feature extraction and fusion methods limit the accuracy of pose regression.</p><p><strong>Purpose: </strong>The objective of this study is to develop a method that leverages both x-ray and computed tomography (CT) images to rapidly and robustly estimate an accurate initial registration within a broad search space. This approach integrates the strengths of learning-based registration with those of traditional registration methodologies, enabling the acquisition of registration outcomes across a wide search space at an accelerated pace.</p><p><strong>Methods: </strong>We introduce a regression-based registration framework to address the aforementioned issues. We constrain the feature fusion process by training the network to reconstruct the high-dimensional feature representation vector of the preoperative CT volume in the embedding space from the input single-view x-ray, thereby enhancing the interpretability of feature extraction. Also, in order to promote the effective fusion and better extraction of local texture features and global information, we propose a lightweight cross-attention mechanism named lite cross-attention(LCAT). Besides, to meet the intraoperative requirements, we employ the intensity-based registration method CMA-ES to refine the result of pose regression.</p><p><strong>Results: </strong>Our approach is verified on both real and simulated x-ray data. Experimental results show that compared with the existing learning-based registration methods, the median rotation error of our method can reach 1.9 <math> <semantics><msup><mrow></mrow> <mo>∘</mo></msup> <annotation>$^circ$</annotation></semantics> </math> and the median translation error can reach 5.6 mm in the case of a large search range. When evaluated on 52 real x-ray images, we have a median rotation error of 1.6 <math> <semantics><msup><mrow></mrow> <mo>∘</mo></msup> <annotation>$^circ$</annotation></semantics> </math> and a median translation error of 3.8 mm due to the smaller search range. We also verify the role of the LCAT and embedding reconstruction modules in our registration framework. If they are not used, our registration performance will be reduced to approximately random initialization results.</p><p><strong>Conclusions: </strong>During the experiments, our method demonstrates higher accuracy and larger capture range on both simulated images and real x-ray images compared to existing methods. The inspiring experimental results indicate the potential for future clinical application of our method.</p>","PeriodicalId":94136,"journal":{"name":"Medical physics","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2025-05-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144121810","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}