Helical TomoTherapy provides highly conformal dose distributions for breast irradiation but is limited by extensive low-dose spillage (“low-dose bath”), contributing to increased integral dose and potential long-term toxicities. Complete blocks can suppress low-dose spread, but at the cost of prolonged treatment times on legacy TomoTherapy systems.
Purpose
To develop and validate a deformable image registration (DIR)-based workflow that predicts patient-specific low-dose distributions and generates personalized complete blocks for TomoTherapy, aiming to reduce low-dose exposure and integral dose. A secondary objective was to determine whether Radixact, a modern helical platform, could mitigate treatment-time penalties while preserving dosimetric benefits.
Methods
Twenty-eight patients were retrospectively analyzed (18 tangential partial-arc volumetric modulated arc therapy [t-VMAT], 10 TomoTherapy). DIR-based dose prediction derived from t-VMAT atlases was used to construct complete blocks for replanning on Hi-Art (TOMO_RE) and Radixact (TOMO_FA). Dosimetric endpoints included target conformity, homogeneity, organ-at-risk (OAR) doses, and integral dose (ID). Statistical analyses used Mann–Whitney U test for independent cohorts and Friedman/Wilcoxon tests for paired TomoTherapy plans with Holm–Bonferroni correction.
Results
TOMO_FA significantly reduced low-dose exposure compared with TOMO_ORI, including lower contralateral lung mean dose (0.79 vs. 3.13 Gy, p < 0.01) and reduced Heart V5 (12.81% vs. 20.94%, p = 0.027). Body-PTV ID decreased meaningfully (103.14 vs. 114.52 Gy·L, p = 0.012). High-dose cardiac parameters (V25, V40) remained within clinically acceptable limits and comparable to t-VMAT. Treatment time improved substantially on Radixact (587.2 ± 44.3 s vs. 1118.0 ± 135.5 s).
Conclusions
The proposed DIR-based complete block workflow effectively reduces low-dose exposure and integral dose in helical TomoTherapy without compromising delivery efficiency when implemented on Radixact. TOMO_FA represents a practical, personalized planning option, particularly for patients requiring stringent low-dose sparing.
背景：螺旋断层扫描治疗为乳腺照射提供了高度适形的剂量分布，但受到广泛的低剂量溢出（“低剂量浴”）的限制，导致总剂量增加和潜在的长期毒性。完全阻断可以抑制低剂量扩散，但代价是延长传统TomoTherapy系统的治疗时间。目的：开发和验证基于可变形图像配准（DIR）的工作流程，预测患者特定的低剂量分布，并为TomoTherapy生成个性化的完整块，旨在减少低剂量暴露和积分剂量。第二个目标是确定Radixact，一个现代螺旋平台，是否可以减轻治疗时间的损失，同时保持剂量学的好处。方法：回顾性分析28例患者（18例切向部分弧线体积调节弧线治疗[t-VMAT]， 10例TomoTherapy）。从t-VMAT图谱中获得的基于dir的剂量预测用于在Hi-Art （TOMO_RE）和Radixact （TOMO_FA）上构建完整的重新规划块。剂量学终点包括目标符合性、均匀性、器官危险（OAR）剂量和积分剂量（ID）。统计分析采用独立队列的Mann-Whitney U检验和配对TomoTherapy计划的Friedman/Wilcoxon检验，并采用Holm-Bonferroni校正。结果：与TOMO_ORI相比，TOMO_FA显著减少了低剂量暴露，包括更低的对侧肺平均剂量（0.79 Gy vs. 3.13 Gy, p）。结论：提出的基于ir的完全阻断工作流程有效地减少了螺旋tomo_治疗中的低剂量暴露和整体剂量，而在Radixact上实施时不会影响递送效率。TOMO_FA代表了一种实用的、个性化的计划选择，特别是对于需要严格的低剂量保留的患者。
{"title":"Reducing low-dose exposure in helical TomoTherapy for locally advanced left-sided breast cancer with a deformable image registration–based dose-mimicking workflow","authors":"Chih-Chieh Chang, Jo-Ting Tsai, Shih-Ming Hsu","doi":"10.1002/acm2.70470","DOIUrl":"10.1002/acm2.70470","url":null,"abstract":"<div>\u0000 \u0000 \u0000 <section>\u0000 \u0000 <h3> Background</h3>\u0000 \u0000 <p>Helical TomoTherapy provides highly conformal dose distributions for breast irradiation but is limited by extensive low-dose spillage (“low-dose bath”), contributing to increased integral dose and potential long-term toxicities. Complete blocks can suppress low-dose spread, but at the cost of prolonged treatment times on legacy TomoTherapy systems.</p>\u0000 </section>\u0000 \u0000 <section>\u0000 \u0000 <h3> Purpose</h3>\u0000 \u0000 <p>To develop and validate a deformable image registration (DIR)-based workflow that predicts patient-specific low-dose distributions and generates personalized complete blocks for TomoTherapy, aiming to reduce low-dose exposure and integral dose. A secondary objective was to determine whether Radixact, a modern helical platform, could mitigate treatment-time penalties while preserving dosimetric benefits.</p>\u0000 </section>\u0000 \u0000 <section>\u0000 \u0000 <h3> Methods</h3>\u0000 \u0000 <p>Twenty-eight patients were retrospectively analyzed (18 tangential partial-arc volumetric modulated arc therapy [t-VMAT], 10 TomoTherapy). DIR-based dose prediction derived from t-VMAT atlases was used to construct complete blocks for replanning on Hi-Art (TOMO_RE) and Radixact (TOMO_FA). Dosimetric endpoints included target conformity, homogeneity, organ-at-risk (OAR) doses, and integral dose (ID). Statistical analyses used Mann–Whitney <i>U</i> test for independent cohorts and Friedman/Wilcoxon tests for paired TomoTherapy plans with Holm–Bonferroni correction.</p>\u0000 </section>\u0000 \u0000 <section>\u0000 \u0000 <h3> Results</h3>\u0000 \u0000 <p>TOMO_FA significantly reduced low-dose exposure compared with TOMO_ORI, including lower contralateral lung mean dose (0.79 vs. 3.13 Gy, <i>p</i> < 0.01) and reduced Heart V5 (12.81% vs. 20.94%, <i>p</i> = 0.027). Body-PTV ID decreased meaningfully (103.14 vs. 114.52 Gy·L, <i>p</i> = 0.012). High-dose cardiac parameters (V25, V40) remained within clinically acceptable limits and comparable to t-VMAT. Treatment time improved substantially on Radixact (587.2 ± 44.3 s vs. 1118.0 ± 135.5 s).</p>\u0000 </section>\u0000 \u0000 <section>\u0000 \u0000 <h3> Conclusions</h3>\u0000 \u0000 <p>The proposed DIR-based complete block workflow effectively reduces low-dose exposure and integral dose in helical TomoTherapy without compromising delivery efficiency when implemented on Radixact. TOMO_FA represents a practical, personalized planning option, particularly for patients requiring stringent low-dose sparing.</p>\u0000 </section>\u0000 </div>","PeriodicalId":14989,"journal":{"name":"Journal of Applied Clinical Medical Physics","volume":"27 2","pages":""},"PeriodicalIF":2.2,"publicationDate":"2026-01-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12836287/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146052149","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
<div> <section> <h3> Background</h3> <p>Using a constant relative biological effectiveness (RBE = 1.1) in proton therapy may underestimate the RBE-weighted dose in high linear energy transfer (LET) regions at the distal end of the beam, thereby limiting the ability to accurately predict clinical outcomes.</p> </section> <section> <h3> Purpose</h3> <p>To commission and validate a Monte Carlo (MC) model incorporating variable RBE for breast cancer proton therapy, enabling improved RBE-weighted dose calculation.</p> </section> <section> <h3> Methods</h3> <p>A FLUKA-based MC model of a raster scanning proton beamline was commissioned and benchmarked against the clinically employed treatment planning system (TPS) (Siemens Syngo) and physical measurements. Dose-averaged LET (LET<sub>d</sub>) and variable RBE-weighted dose distributions were computed using McMahon (McM), McNamara (McN), and Wedenberg (Wed) models. Treatment plans for four representative breast cancer cases were recalculated to compare TPS and MC results using dose-volume histograms (DVH) and three-dimensional gamma (γ) analysis. LET<sub>d</sub>-volume histograms (LVH) and variable RBE-weighted dose distributions were analyzed to compare cases without adverse effects versus those presenting rib fractures or radiation pneumonitis.</p> </section> <section> <h3> Results</h3> <p>The FLUKA-MC model showed good agreement with both the TPS results and the measured data, exhibiting proton range deviations within ±0.1 mm. The γ pass rates for the four patients are 94.0%, 92.2%, 92.6%, and 86.7%, respectively. LET<sub>d</sub> analysis of 0.5 cm<sup>3</sup> volumes of rib revealed numerical differences (fracture cases: 11.1 and 10.8 keV/µm; non-fracture: 9.2 and 10.0 keV/µm). The RBE-weighted dose to 0.5 cm<sup>3</sup> of the ribs was consistently elevated in fracture cases across all models (RBE = 1.1: 46.2–49.0 Gy; McM: 54.6–56.5 Gy; McN: 51.0–53.3 Gy; Wed: 50.6–52.5 Gy) versus non-fracture cases (RBE = 1.1: 44.0–45.3 Gy; McM: 52.2–53.8 Gy; McN: 48.6–50.1 Gy; Wed: 48.3–49.8 Gy). The estimated RBE values in the rib region were 1.60 (McM), 1.38 (McN), and 1.44 (Wed), which were derived from the mean LET<sub>d</sub> within 0.5 cm<sup>3</sup> rib volumes. The RBE-weighted lung V20 was elevated in pneumonitis patients across all models. All variable RBE models predicted elevated RBE-weighted doses in distal proton beam regions across cases.</p> </section> <section> <h3> Conclusions</h3> <p>The commissio
{"title":"Commissioning of a Monte Carlo-based scanning proton beam for breast cancer: Incorporating LETd calculations and variable RBE models","authors":"Zhen Cao, Qing Zhang, Jingfang Zhao","doi":"10.1002/acm2.70477","DOIUrl":"10.1002/acm2.70477","url":null,"abstract":"<div>\u0000 \u0000 \u0000 <section>\u0000 \u0000 <h3> Background</h3>\u0000 \u0000 <p>Using a constant relative biological effectiveness (RBE = 1.1) in proton therapy may underestimate the RBE-weighted dose in high linear energy transfer (LET) regions at the distal end of the beam, thereby limiting the ability to accurately predict clinical outcomes.</p>\u0000 </section>\u0000 \u0000 <section>\u0000 \u0000 <h3> Purpose</h3>\u0000 \u0000 <p>To commission and validate a Monte Carlo (MC) model incorporating variable RBE for breast cancer proton therapy, enabling improved RBE-weighted dose calculation.</p>\u0000 </section>\u0000 \u0000 <section>\u0000 \u0000 <h3> Methods</h3>\u0000 \u0000 <p>A FLUKA-based MC model of a raster scanning proton beamline was commissioned and benchmarked against the clinically employed treatment planning system (TPS) (Siemens Syngo) and physical measurements. Dose-averaged LET (LET<sub>d</sub>) and variable RBE-weighted dose distributions were computed using McMahon (McM), McNamara (McN), and Wedenberg (Wed) models. Treatment plans for four representative breast cancer cases were recalculated to compare TPS and MC results using dose-volume histograms (DVH) and three-dimensional gamma (γ) analysis. LET<sub>d</sub>-volume histograms (LVH) and variable RBE-weighted dose distributions were analyzed to compare cases without adverse effects versus those presenting rib fractures or radiation pneumonitis.</p>\u0000 </section>\u0000 \u0000 <section>\u0000 \u0000 <h3> Results</h3>\u0000 \u0000 <p>The FLUKA-MC model showed good agreement with both the TPS results and the measured data, exhibiting proton range deviations within ±0.1 mm. The γ pass rates for the four patients are 94.0%, 92.2%, 92.6%, and 86.7%, respectively. LET<sub>d</sub> analysis of 0.5 cm<sup>3</sup> volumes of rib revealed numerical differences (fracture cases: 11.1 and 10.8 keV/µm; non-fracture: 9.2 and 10.0 keV/µm). The RBE-weighted dose to 0.5 cm<sup>3</sup> of the ribs was consistently elevated in fracture cases across all models (RBE = 1.1: 46.2–49.0 Gy; McM: 54.6–56.5 Gy; McN: 51.0–53.3 Gy; Wed: 50.6–52.5 Gy) versus non-fracture cases (RBE = 1.1: 44.0–45.3 Gy; McM: 52.2–53.8 Gy; McN: 48.6–50.1 Gy; Wed: 48.3–49.8 Gy). The estimated RBE values in the rib region were 1.60 (McM), 1.38 (McN), and 1.44 (Wed), which were derived from the mean LET<sub>d</sub> within 0.5 cm<sup>3</sup> rib volumes. The RBE-weighted lung V20 was elevated in pneumonitis patients across all models. All variable RBE models predicted elevated RBE-weighted doses in distal proton beam regions across cases.</p>\u0000 </section>\u0000 \u0000 <section>\u0000 \u0000 <h3> Conclusions</h3>\u0000 \u0000 <p>The commissio","PeriodicalId":14989,"journal":{"name":"Journal of Applied Clinical Medical Physics","volume":"27 2","pages":""},"PeriodicalIF":2.2,"publicationDate":"2026-01-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12836373/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146052099","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Laura E. Dinwiddie, Jared M. Baggett, James M. Kofler, Cameron B. Kofler, Daniel J. Long, Robert J. Dawson, Stefan K. Wehmeier, Yitian Wang, Juan C. Ocampo-Ramos, Lukas M. Carter, Harry Marquis, Gunjan Kayal, Adam L. Kesner, Wesley E. Bolch
<div> <section> <h3> Background</h3> <p>Computed tomography (CT) is an essential imaging modality for disease diagnosis, treatment efficacy, and image-based guidance of various medical procedures. The locally deposited radiation dose in tissues, as estimated by the computed tomography dose index (CTDI), can vary considerably across exposures delivered by CT scanners from different vendors, even if the scans are performed using similar technique factors, such as tube potential and tube current. The volumetric CTDI (CTDI<sub>vol</sub>) is a common dose metric that reports an average radiation dose (in mGy) delivered to a specific volume within a test phantom. The CTDI<sub>vol</sub> is important in dosimetry applications as the organ absorbed dose within the patient has been shown to scale in near-linear proportion, creating a basis for comparing organ doses across different scan protocols and scanner models.</p> </section> <section> <h3> Purpose</h3> <p>To develop a database of tube current-time product (mAs) normalized CTDI<sub>vol</sub> values for currently utilized CT scanner models for each of the four primary CT vendors for use in the MIRDct organ dosimetry software available at MIRDsoft.org. This data forms the basis of the MIRDct code, which reports organ doses across a range of computational phantoms based upon axial organ dose coefficient libraries generated through Monte Carlo radiation transport for a reference CT scanner. Organ doses delivered by alternate CT scanner vendors and models may then be reported using ratios of normalized CTDI<sub>vol</sub> values under similar technique factors.</p> </section> <section> <h3> Methods</h3> <p>Scanners were selected from four major CT manufacturers: Philips Healthcare, GE Healthcare, Canon Medical Systems, and Siemens Healthineers. Technique parameters were also selected for each scanner that closely matched values used in the generation of an equivalent CT source term (small to large bowtie filters; 80–140-kVp tube voltage; and 10-mm to 40-mm beam collimation). For each scanner chosen, the appropriate technique factors and protocols were selected, and the console-reported CTDI<sub>vol</sub> values were recorded and normalized to a set value of 100 mAs. The normalized CTDI<sub>vol</sub> data collected for use within the MIRDct code were analyzed for noticeable patterns, features, and trends, and were compared to similar normalized CTDI<sub>vol</sub> datasets used within the National Cancer Institute NCICT software and the Virtual Phantoms, Inc. VirtualDose software.</p> </section> <section> <h3> Results</h3>
背景：计算机断层扫描（CT）是疾病诊断、治疗效果和基于图像指导各种医疗程序的基本成像方式。根据计算机断层扫描剂量指数（CTDI）估计，组织中局部沉积的辐射剂量可能因不同供应商的CT扫描仪所提供的照射而有很大差异，即使使用类似的技术因素（如管电位和管电流）进行扫描。体积CTDI （CTDIvol）是一种常用的剂量度量，它报告了在测试模体中传递到特定体积的平均辐射剂量（以mGy为单位）。CTDIvol在剂量学应用中很重要，因为患者体内的器官吸收剂量已显示成近线性比例，为比较不同扫描方案和扫描仪模型的器官剂量创造了基础。目的：为目前使用的四家主要CT供应商的CT扫描仪模型开发一个管电流时间产物（mAs）标准化CTDIvol值数据库，用于MIRDsoft.org上提供的MIRDct器官剂量学软件。该数据构成了MIRDct代码的基础，该代码基于通过蒙特卡罗辐射传输为参考CT扫描仪生成的轴向器官剂量系数库，在一系列计算幻象中报告器官剂量。在类似的技术因素下，可使用归一化CTDIvol值的比值来报告由不同的CT扫描仪供应商和型号提供的器官剂量。方法：选用Philips Healthcare、GE Healthcare、Canon Medical Systems和Siemens Healthineers四家主要CT制造商的扫描仪。为每台扫描仪选择的技术参数与生成等效CT源项时使用的值（从小到大的领结滤波器，80-140 kvp的管电压，10-mm到40-mm的光束准直）密切匹配。对于选择的每个扫描仪，选择适当的技术因素和方案，并记录控制台报告的CTDIvol值并归一化为100 ma的设定值。对收集的用于MIRDct代码的规范化CTDIvol数据进行分析，以发现明显的模式、特征和趋势，并将其与国家癌症研究所NCICT软件和Virtual phantom， Inc.中使用的类似规范化CTDIvol数据集进行比较。VirtualDose软件。结果：对于所有给定的CT扫描仪和技术因素组合，所有三种代码的归一化CTDIvol值都非常一致：比较扫描仪的差异在0%到12%之间。不同CT扫描仪供应商和型号的CTDIvol值与MIRDct参考扫描仪（Cannon Aquilion One Genesis）相应的CTDIvol值的比率也在16厘米头部PMMA幻影或32厘米身体PMMA幻影的基础上进行了比较。这些归一化CTDIvol比值（头部与身体比值）的平均商约为1.06，因此任何比值都可以应用于MIRDct报告患者器官剂量。结论：我们建立了一个归一化CTDIvol （mGy/100 mAs）数据库，适用于来自四家制造商的17种不同管电位、准直、x射线领结滤光片和幻象尺寸的CT扫描仪，用于MIRDct软件。
{"title":"Survey of normalized CTDIvol values across four major computed tomography vendors for use in the MIRDct software","authors":"Laura E. Dinwiddie, Jared M. Baggett, James M. Kofler, Cameron B. Kofler, Daniel J. Long, Robert J. Dawson, Stefan K. Wehmeier, Yitian Wang, Juan C. Ocampo-Ramos, Lukas M. Carter, Harry Marquis, Gunjan Kayal, Adam L. Kesner, Wesley E. Bolch","doi":"10.1002/acm2.70473","DOIUrl":"10.1002/acm2.70473","url":null,"abstract":"<div>\u0000 \u0000 \u0000 <section>\u0000 \u0000 <h3> Background</h3>\u0000 \u0000 <p>Computed tomography (CT) is an essential imaging modality for disease diagnosis, treatment efficacy, and image-based guidance of various medical procedures. The locally deposited radiation dose in tissues, as estimated by the computed tomography dose index (CTDI), can vary considerably across exposures delivered by CT scanners from different vendors, even if the scans are performed using similar technique factors, such as tube potential and tube current. The volumetric CTDI (CTDI<sub>vol</sub>) is a common dose metric that reports an average radiation dose (in mGy) delivered to a specific volume within a test phantom. The CTDI<sub>vol</sub> is important in dosimetry applications as the organ absorbed dose within the patient has been shown to scale in near-linear proportion, creating a basis for comparing organ doses across different scan protocols and scanner models.</p>\u0000 </section>\u0000 \u0000 <section>\u0000 \u0000 <h3> Purpose</h3>\u0000 \u0000 <p>To develop a database of tube current-time product (mAs) normalized CTDI<sub>vol</sub> values for currently utilized CT scanner models for each of the four primary CT vendors for use in the MIRDct organ dosimetry software available at MIRDsoft.org. This data forms the basis of the MIRDct code, which reports organ doses across a range of computational phantoms based upon axial organ dose coefficient libraries generated through Monte Carlo radiation transport for a reference CT scanner. Organ doses delivered by alternate CT scanner vendors and models may then be reported using ratios of normalized CTDI<sub>vol</sub> values under similar technique factors.</p>\u0000 </section>\u0000 \u0000 <section>\u0000 \u0000 <h3> Methods</h3>\u0000 \u0000 <p>Scanners were selected from four major CT manufacturers: Philips Healthcare, GE Healthcare, Canon Medical Systems, and Siemens Healthineers. Technique parameters were also selected for each scanner that closely matched values used in the generation of an equivalent CT source term (small to large bowtie filters; 80–140-kVp tube voltage; and 10-mm to 40-mm beam collimation). For each scanner chosen, the appropriate technique factors and protocols were selected, and the console-reported CTDI<sub>vol</sub> values were recorded and normalized to a set value of 100 mAs. The normalized CTDI<sub>vol</sub> data collected for use within the MIRDct code were analyzed for noticeable patterns, features, and trends, and were compared to similar normalized CTDI<sub>vol</sub> datasets used within the National Cancer Institute NCICT software and the Virtual Phantoms, Inc. VirtualDose software.</p>\u0000 </section>\u0000 \u0000 <section>\u0000 \u0000 <h3> Results</h3>\u0000 \u0000 ","PeriodicalId":14989,"journal":{"name":"Journal of Applied Clinical Medical Physics","volume":"27 2","pages":""},"PeriodicalIF":2.2,"publicationDate":"2026-01-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12835190/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146052161","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Steve D. Mann, Donald L. Miller, Grant Fong, Allen R. Goode, Emily L. Marshall, Thomas Nishino, Pavlina Boxx, Liqiang Ren, Celalettin Topbas, Alan H. Schoenfeld, Vivek Singh, Jie Zhang
<div> <section> <h3> Background</h3> <p>The ACR Diagnostic Fluoroscopy Dose Index Registry (DIR-Fluoro) is expanding to include diagnostic fluoroscopy. Variations in dose reference points and overhead radiography events may introduce unique challenges for benchmarking.</p> </section> <section> <h3> Purpose</h3> <p>To survey the technological status and dosimetric performance of fluoroscopes participating in the DIR-Fluoro pilot project, focusing on longitudinal stability and variability of fluoroscopic dose reporting accuracy across multiple institutions and vendors.</p> </section> <section> <h3> Methods</h3> <p>Sixty-six fluoroscopic systems from nine institutions (24 facilities) were surveyed for facility type, fluoroscope type, image receptor type, age, dose reporting capabilities, and other key features. Of these, 56 were evaluable. Semi-annual measurements assessed reference air kerma (K<sub>a,r</sub>) and air kerma area product (P<sub>KA</sub>) accuracy. Linear mixed-effects models evaluated changes in dose accuracy over time, incorporating system-specific random effects; models were compared using likelihood ratio testing. Radiation Dose Structured Reports (RDSR) contents were investigated to understand the challenges in benchmarking diagnostic fluoroscopy dose indices.</p> </section> <section> <h3> Results</h3> <p>Nearly 80% of units were tube-under-table fluoroscopes. Average age was 9.6 ± 5.2 years. Sixty-four percent of the units produced RDSRs. Median deviations for K<sub>a,r</sub> and P<sub>KA</sub> were 1%–4%. Accuracy of P<sub>KA</sub> and K<sub>a,r</sub> remained stable, with no significant time-dependent drift for RDSR-capable systems (<i>p</i> > 0.05). Incorporating detector type significantly improved performance for P<sub>KA</sub> measurements (<i>p</i> < 0.05 for all datasets); K<sub>a,r</sub> models were generally best fit by simpler models (<i>p</i> > 0.05 for 3 of 4 datasets). Major discrepancies in RDSRs were observed, including differences in K<sub>a,r</sub> reference point definitions and in event-level data. Overhead radiography exposures were not well distinguished from fluoroscope exposures. These issues resulted in inconsistencies in reported K<sub>a,r</sub> values.</p> </section> <section> <h3> Conclusion</h3> <p>Fluoroscopic dose indices were accurate and stable over time. Differences in RDSR availability result in data biased to newer systems with flat panel detectors. Discrepancies in RDSR content and inc
{"title":"The American college of radiology diagnostic fluoroscopy dose index registry pilot: Dosimetric performance and benchmarking challenges","authors":"Steve D. Mann, Donald L. Miller, Grant Fong, Allen R. Goode, Emily L. Marshall, Thomas Nishino, Pavlina Boxx, Liqiang Ren, Celalettin Topbas, Alan H. Schoenfeld, Vivek Singh, Jie Zhang","doi":"10.1002/acm2.70458","DOIUrl":"10.1002/acm2.70458","url":null,"abstract":"<div>\u0000 \u0000 \u0000 <section>\u0000 \u0000 <h3> Background</h3>\u0000 \u0000 <p>The ACR Diagnostic Fluoroscopy Dose Index Registry (DIR-Fluoro) is expanding to include diagnostic fluoroscopy. Variations in dose reference points and overhead radiography events may introduce unique challenges for benchmarking.</p>\u0000 </section>\u0000 \u0000 <section>\u0000 \u0000 <h3> Purpose</h3>\u0000 \u0000 <p>To survey the technological status and dosimetric performance of fluoroscopes participating in the DIR-Fluoro pilot project, focusing on longitudinal stability and variability of fluoroscopic dose reporting accuracy across multiple institutions and vendors.</p>\u0000 </section>\u0000 \u0000 <section>\u0000 \u0000 <h3> Methods</h3>\u0000 \u0000 <p>Sixty-six fluoroscopic systems from nine institutions (24 facilities) were surveyed for facility type, fluoroscope type, image receptor type, age, dose reporting capabilities, and other key features. Of these, 56 were evaluable. Semi-annual measurements assessed reference air kerma (K<sub>a,r</sub>) and air kerma area product (P<sub>KA</sub>) accuracy. Linear mixed-effects models evaluated changes in dose accuracy over time, incorporating system-specific random effects; models were compared using likelihood ratio testing. Radiation Dose Structured Reports (RDSR) contents were investigated to understand the challenges in benchmarking diagnostic fluoroscopy dose indices.</p>\u0000 </section>\u0000 \u0000 <section>\u0000 \u0000 <h3> Results</h3>\u0000 \u0000 <p>Nearly 80% of units were tube-under-table fluoroscopes. Average age was 9.6 ± 5.2 years. Sixty-four percent of the units produced RDSRs. Median deviations for K<sub>a,r</sub> and P<sub>KA</sub> were 1%–4%. Accuracy of P<sub>KA</sub> and K<sub>a,r</sub> remained stable, with no significant time-dependent drift for RDSR-capable systems (<i>p</i> > 0.05). Incorporating detector type significantly improved performance for P<sub>KA</sub> measurements (<i>p</i> < 0.05 for all datasets); K<sub>a,r</sub> models were generally best fit by simpler models (<i>p</i> > 0.05 for 3 of 4 datasets). Major discrepancies in RDSRs were observed, including differences in K<sub>a,r</sub> reference point definitions and in event-level data. Overhead radiography exposures were not well distinguished from fluoroscope exposures. These issues resulted in inconsistencies in reported K<sub>a,r</sub> values.</p>\u0000 </section>\u0000 \u0000 <section>\u0000 \u0000 <h3> Conclusion</h3>\u0000 \u0000 <p>Fluoroscopic dose indices were accurate and stable over time. Differences in RDSR availability result in data biased to newer systems with flat panel detectors. Discrepancies in RDSR content and inc","PeriodicalId":14989,"journal":{"name":"Journal of Applied Clinical Medical Physics","volume":"27 2","pages":""},"PeriodicalIF":2.2,"publicationDate":"2026-01-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12833689/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146046752","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
<div> <section> <h3> Background</h3> <p>The measurement of computed tomography dose index 100 (<span></span><math> <semantics> <mrow> <mi>C</mi> <mi>T</mi> <mi>D</mi> <msub> <mi>I</mi> <mn>100</mn> </msub> </mrow> <annotation>$CTD{{I}_{100}}$</annotation> </semantics></math>), which is feasible only through axial scanning, requires that the clinical spiral protocols be replaced with those for axial scanning. The real-time ionization chamber detects the integral of radiation dose rate profile, enabling the direct verification of the volume CTDI on spiral CT scanning (<span></span><math> <semantics> <mrow> <mi>C</mi> <mi>T</mi> <mi>D</mi> <msubsup> <mi>I</mi> <mrow> <mi>v</mi> <mi>o</mi> <mi>l</mi> </mrow> <mrow> <mi>S</mi> <mi>p</mi> <mi>i</mi> <mi>r</mi> <mi>a</mi> <mi>l</mi> </mrow> </msubsup> </mrow> <annotation>$CTDI_{vol}^{Spiral}$</annotation> </semantics></math>).</p> </section> <section> <h3> Purpose</h3> <p>This study aimed to develop a direct measurement technique for <span></span><math> <semantics> <mrow> <mi>C</mi> <mi>T</mi> <mi>D</mi> <msubsup> <mi>I</mi> <mrow> <mi>v</mi> <mi>o</mi> <mi>l</mi> </mrow> <mrow> <mi>S</mi> <mi>p</mi> <mi>i</mi> <mi>r</mi> <mi>a</mi> <mi>l</mi>
背景：计算机断层扫描剂量指数100 （CTD I 100 $CTD{{I}_{100}}$）的测量只有通过轴向扫描才能实现，需要将临床螺旋方案替换为轴向扫描方案。实时电离室检测辐射剂量率曲线的积分，实现螺旋CT扫描体积CTDI的直接验证（CTDI vol Sp I I I al $CTDI_{vol}^{spiral}$）。目的:本研究旨在开发一个直接测量技术C T D I v o l S p I r l $ CTDI_{卷}^{螺旋}$,比较其准确性和测量使用轴向扫描(C T D我v o l x l $ CTDI_{卷}^{轴}$),或者显示在控制台(C T D I v o y l D S p l e D $ CTDI_{卷}^{显示}$)。方法：将带实时电离室的CTDI假体放置于头枕或检查台上。CTDI_{100}^{Axial}$的测量参数为：管电压= 120 kV，有效mAs = 100，旋转时间= 1.00 s。测量C - T - D - 100 S - p - I - r - 1 $CTDI_{100}^{螺旋}$的参数设置与轴向扫描参数相同，除了旋转次数= 0.33、0.50和1.00，节距= 0.35、0.50、0.75、1.00、1.25和1.50，扫描范围= 15 cm。CTDI 100 Sp I ral $CTDI_{100}^{螺旋}$从辐射剂量率曲线积分中提取。随后计算了C - T - D - I v / l S - p - I - 1 $CTDI_{vol}^{螺旋}$，并与C - T - D - I - I - I - I - 1 $CTDI_{vol}^{轴向}$和C - D - I - I - I - I - a - I - 1 $CTDI_{vol}^{显示}$进行了比较。最后,C T D I v o l S p I r l $ CTDI_{卷}^{螺旋}$ 10临床测量协议并与C T D I v o y l D S p l e D $ CTDI_{卷}^{显示}$。结果:C T D之间的差异我v o l x l $ CTDI_{卷}^{轴}$ l和C T D I v o D I y s p l e D $ CTDI_{卷}^{显示}$,C T D I v o l s p I r l $ CTDI_{卷}^{螺旋}$ l和C T D I v o D I y s p l e D $ CTDI_{卷}^{显示}$,和C T D我v o l x l $ CTDI_{卷}^{轴}$ l和C T D I v o S p I r l $ CTDI_{卷}^{螺旋}$头部和身体的幻影都C T D I v o l D I y S p l e D $ CTDI_{卷}^{显示}$ l和C T D I v o S p I r l $ CTDI_{卷}^{螺旋}$对临床协议扫描并没有自动曝光控制的结论:结果显示，C - T - D - v - 1 - A - x - 1 - CTDI_{vol}}^{Axial}$与C - T - D - v - 1 - S - p - 1 - 1 - CTDI_{vol}}^{Spiral}$具有很好的一致性，支持临床螺旋CT扫描使用实时电离室验证C - T - D - v - 1 - D - 1 - D - C - D - 1 - D - 1 - C - D - 1 - D - 1 - C - D - 1 - D - C - D - 1 - vol}}^{display}$。
{"title":"Measurement of the volume CT dose index on spiral CT scanning with a real-time ionization chamber","authors":"Atsushi Fukuda, Nao Ichikawa, Takuma Hayashi, Ayaka Hirosawa, Kosuke Matsubara","doi":"10.1002/acm2.70469","DOIUrl":"10.1002/acm2.70469","url":null,"abstract":"<div>\u0000 \u0000 \u0000 <section>\u0000 \u0000 <h3> Background</h3>\u0000 \u0000 <p>The measurement of computed tomography dose index 100 (<span></span><math>\u0000 <semantics>\u0000 <mrow>\u0000 <mi>C</mi>\u0000 <mi>T</mi>\u0000 <mi>D</mi>\u0000 <msub>\u0000 <mi>I</mi>\u0000 <mn>100</mn>\u0000 </msub>\u0000 </mrow>\u0000 <annotation>$CTD{{I}_{100}}$</annotation>\u0000 </semantics></math>), which is feasible only through axial scanning, requires that the clinical spiral protocols be replaced with those for axial scanning. The real-time ionization chamber detects the integral of radiation dose rate profile, enabling the direct verification of the volume CTDI on spiral CT scanning (<span></span><math>\u0000 <semantics>\u0000 <mrow>\u0000 <mi>C</mi>\u0000 <mi>T</mi>\u0000 <mi>D</mi>\u0000 <msubsup>\u0000 <mi>I</mi>\u0000 <mrow>\u0000 <mi>v</mi>\u0000 <mi>o</mi>\u0000 <mi>l</mi>\u0000 </mrow>\u0000 <mrow>\u0000 <mi>S</mi>\u0000 <mi>p</mi>\u0000 <mi>i</mi>\u0000 <mi>r</mi>\u0000 <mi>a</mi>\u0000 <mi>l</mi>\u0000 </mrow>\u0000 </msubsup>\u0000 </mrow>\u0000 <annotation>$CTDI_{vol}^{Spiral}$</annotation>\u0000 </semantics></math>).</p>\u0000 </section>\u0000 \u0000 <section>\u0000 \u0000 <h3> Purpose</h3>\u0000 \u0000 <p>This study aimed to develop a direct measurement technique for <span></span><math>\u0000 <semantics>\u0000 <mrow>\u0000 <mi>C</mi>\u0000 <mi>T</mi>\u0000 <mi>D</mi>\u0000 <msubsup>\u0000 <mi>I</mi>\u0000 <mrow>\u0000 <mi>v</mi>\u0000 <mi>o</mi>\u0000 <mi>l</mi>\u0000 </mrow>\u0000 <mrow>\u0000 <mi>S</mi>\u0000 <mi>p</mi>\u0000 <mi>i</mi>\u0000 <mi>r</mi>\u0000 <mi>a</mi>\u0000 <mi>l</mi>\u0000 ","PeriodicalId":14989,"journal":{"name":"Journal of Applied Clinical Medical Physics","volume":"27 2","pages":""},"PeriodicalIF":2.2,"publicationDate":"2026-01-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12835228/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146052082","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Low-dose CT (LDCT) is increasingly being adopted as a preferred method for lung cancer screening. However, the accompanying rise in image noise necessitates robust denoising strategies. Therefore, this study compared LDCT images with their denoised counterparts using objective image quality metrics and key nodule-related features.
Methods
The dataset utilized in this study was chest CT scans for lung cancer screening, sourced from the LDCT and Projection Data collection. Seven deep learning-based image denoising methods were used in this work. The denoising performance was evaluated using root-mean-square error (RMSE), peak signal-to-noise ratio (PSNR), structural similarity index measure (SSIM), nodule size, CT density, and Lung-RADS classification.
Results
For solid nodules, denoising improved SSIM from 51% to 60%–64%, reduced RMSE from 137.13 HU to 62.40–78.30 HU, and increased PSNR from 23.91 dB to 28.59–30.51 dB. It also reduced the percent difference in diameter (PDdia) from 2.05% to 1.44%–1.52%, in volume (PDvol) from 5.95% to 4.43%–4.70%, in mean HU value (PDHU) from 24.40% to 8.54%–15.33%. For subsolid nodules, denoising improved SSIM from 47% to 57%–61%, reduced RMSE from 110.87 HU to 54.62–63.96 HU, and increased PSNR from 25.78 dB to 30.53–31.61 dB. Before denoising, the PDdia, PDvol and PDHU were 15.41%, 40.16% and 10.69%, respectively, which were 7.54%–15.94%, 17.54%–29.29%, and 6.10%–8.25% after denoising. These improvements led to higher Lung-RADS categorization accuracy for solid nodules, while subsolid nodules remained more affected by noise and denoising-induced bias.
Conclusion
The integration of denoising techniques into LDCT workflows could potentially enhance early lung cancer detection without increasing radiation exposure. Nonetheless, validating their influence on diagnostic performance remains crucial for clinical adoption.
{"title":"Evaluating the impact of deep learning-based image denoising on low-dose CT for lung cancer screening","authors":"Shih-Sheng Chen, Hsiao-Hua Liu, Ching-Ching Yang","doi":"10.1002/acm2.70480","DOIUrl":"10.1002/acm2.70480","url":null,"abstract":"<div>\u0000 \u0000 \u0000 <section>\u0000 \u0000 <h3> Purpose</h3>\u0000 \u0000 <p>Low-dose CT (LDCT) is increasingly being adopted as a preferred method for lung cancer screening. However, the accompanying rise in image noise necessitates robust denoising strategies. Therefore, this study compared LDCT images with their denoised counterparts using objective image quality metrics and key nodule-related features.</p>\u0000 </section>\u0000 \u0000 <section>\u0000 \u0000 <h3> Methods</h3>\u0000 \u0000 <p>The dataset utilized in this study was chest CT scans for lung cancer screening, sourced from the LDCT and Projection Data collection. Seven deep learning-based image denoising methods were used in this work. The denoising performance was evaluated using root-mean-square error (RMSE), peak signal-to-noise ratio (PSNR), structural similarity index measure (SSIM), nodule size, CT density, and Lung-RADS classification.</p>\u0000 </section>\u0000 \u0000 <section>\u0000 \u0000 <h3> Results</h3>\u0000 \u0000 <p>For solid nodules, denoising improved SSIM from 51% to 60%–64%, reduced RMSE from 137.13 HU to 62.40–78.30 HU, and increased PSNR from 23.91 dB to 28.59–30.51 dB. It also reduced the percent difference in diameter (PD<sub>dia</sub>) from 2.05% to 1.44%–1.52%, in volume (PD<sub>vol</sub>) from 5.95% to 4.43%–4.70%, in mean HU value (PD<sub>HU</sub>) from 24.40% to 8.54%–15.33%. For subsolid nodules, denoising improved SSIM from 47% to 57%–61%, reduced RMSE from 110.87 HU to 54.62–63.96 HU, and increased PSNR from 25.78 dB to 30.53–31.61 dB. Before denoising, the PD<sub>dia</sub>, PD<sub>vol</sub> and PD<sub>HU</sub> were 15.41%, 40.16% and 10.69%, respectively, which were 7.54%–15.94%, 17.54%–29.29%, and 6.10%–8.25% after denoising. These improvements led to higher Lung-RADS categorization accuracy for solid nodules, while subsolid nodules remained more affected by noise and denoising-induced bias.</p>\u0000 </section>\u0000 \u0000 <section>\u0000 \u0000 <h3> Conclusion</h3>\u0000 \u0000 <p>The integration of denoising techniques into LDCT workflows could potentially enhance early lung cancer detection without increasing radiation exposure. Nonetheless, validating their influence on diagnostic performance remains crucial for clinical adoption.</p>\u0000 </section>\u0000 </div>","PeriodicalId":14989,"journal":{"name":"Journal of Applied Clinical Medical Physics","volume":"27 2","pages":""},"PeriodicalIF":2.2,"publicationDate":"2026-01-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12831282/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146041048","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Yifan Lei, Han Bai, Chengshu Gong, Yaoxiong Xia, Yu Hou, Ruiling Yang, Jinhui Yu, Zhe Zhang, Li Wang, Bo Li, Li Wang, Lan Li
<div> <section> <h3> Objective</h3> <p>Dosiomics and radiomics elaborate the low-and high-order features extracted from images to predict clinical outcomes. Whole-brain radiotherapy (WBRT) has been widely used in patients with diffuse brain metastases of small cell lung cancer (SCLC). The objective of this study is to ascertain the predictors of treatment response in patients with SCLC treated with WBRT. Furthermore, the study seeks to develop accurate machine learning models to predict the radiotherapy response of WBRT.</p> </section> <section> <h3> Materials and methods</h3> <p>This study retrospectively enrolled BM patients who received whole brain irradiation in Yunnan Cancer Hospital from January 2020 to June 2024. Radiomics features and dosiomics features were extracted from pre-treatment CT images and dose images of TPS using 3D slicer software, features were screened by Least Absolute Shrinkage and Selection Operator (LASSO) regression, and Logistic Regression (LR) models assessed the association of the features with WBRT reaction. Patients who showed complete response (CR) or partial response (PR) were classified as the Radiation Response Group, while those with stable disease (SD) or progressive disease (PD) were categorized as the Radiation Non-Response Group. A total of seven classification models were constructed, clinic factors (CFM)), radiomics features (RFM), dosiomics features (DFM), clinical factors combined with radiomics features (FM + RFM), clinical factors combined with dosiomics features (CFM + DFM), radiomics combined with dosiomics features (RFM + DFM), and the hybrid features combining clinical factors, radiomics, and dosiomics features (HFM). The HFM was our focus, evaluated the prediction performance of the model, used nomograms to visualize individualized Radiation Therapy (RT) response prediction, and prospectively collected a subset of patients for external validation set.</p> </section> <section> <h3> Result</h3> <p>Based on univariate analysis combined with LASSO regression, three dosiomics features and four radiomics features related to the therapeutic effect were respectively selected from 851 dosiomics and radiomic features. Multivariate analysis indicated that concurrent chemoradiotherapy (CCRT), conformal boost radiotherapy (CBRT), radiomics, and dosiomics were independent predictors of the radiotherapy response of WBRT. The multicomponent model based on dosiomics, radiomics and clinical factors showed optimal predictive power in the patient cohort, with a mean AUC = 0.792 (95% CI 0.708–0.852), AUC of external validation set = 0.711 (95%CI 0.487–0.934) and the constructed nomogram charts have good clinical valu
{"title":"Multi-omics predicts radiotherapy response in small cell lung cancer patients receiving whole brain irradiation","authors":"Yifan Lei, Han Bai, Chengshu Gong, Yaoxiong Xia, Yu Hou, Ruiling Yang, Jinhui Yu, Zhe Zhang, Li Wang, Bo Li, Li Wang, Lan Li","doi":"10.1002/acm2.70466","DOIUrl":"10.1002/acm2.70466","url":null,"abstract":"<div>\u0000 \u0000 \u0000 <section>\u0000 \u0000 <h3> Objective</h3>\u0000 \u0000 <p>Dosiomics and radiomics elaborate the low-and high-order features extracted from images to predict clinical outcomes. Whole-brain radiotherapy (WBRT) has been widely used in patients with diffuse brain metastases of small cell lung cancer (SCLC). The objective of this study is to ascertain the predictors of treatment response in patients with SCLC treated with WBRT. Furthermore, the study seeks to develop accurate machine learning models to predict the radiotherapy response of WBRT.</p>\u0000 </section>\u0000 \u0000 <section>\u0000 \u0000 <h3> Materials and methods</h3>\u0000 \u0000 <p>This study retrospectively enrolled BM patients who received whole brain irradiation in Yunnan Cancer Hospital from January 2020 to June 2024. Radiomics features and dosiomics features were extracted from pre-treatment CT images and dose images of TPS using 3D slicer software, features were screened by Least Absolute Shrinkage and Selection Operator (LASSO) regression, and Logistic Regression (LR) models assessed the association of the features with WBRT reaction. Patients who showed complete response (CR) or partial response (PR) were classified as the Radiation Response Group, while those with stable disease (SD) or progressive disease (PD) were categorized as the Radiation Non-Response Group. A total of seven classification models were constructed, clinic factors (CFM)), radiomics features (RFM), dosiomics features (DFM), clinical factors combined with radiomics features (FM + RFM), clinical factors combined with dosiomics features (CFM + DFM), radiomics combined with dosiomics features (RFM + DFM), and the hybrid features combining clinical factors, radiomics, and dosiomics features (HFM). The HFM was our focus, evaluated the prediction performance of the model, used nomograms to visualize individualized Radiation Therapy (RT) response prediction, and prospectively collected a subset of patients for external validation set.</p>\u0000 </section>\u0000 \u0000 <section>\u0000 \u0000 <h3> Result</h3>\u0000 \u0000 <p>Based on univariate analysis combined with LASSO regression, three dosiomics features and four radiomics features related to the therapeutic effect were respectively selected from 851 dosiomics and radiomic features. Multivariate analysis indicated that concurrent chemoradiotherapy (CCRT), conformal boost radiotherapy (CBRT), radiomics, and dosiomics were independent predictors of the radiotherapy response of WBRT. The multicomponent model based on dosiomics, radiomics and clinical factors showed optimal predictive power in the patient cohort, with a mean AUC = 0.792 (95% CI 0.708–0.852), AUC of external validation set = 0.711 (95%CI 0.487–0.934) and the constructed nomogram charts have good clinical valu","PeriodicalId":14989,"journal":{"name":"Journal of Applied Clinical Medical Physics","volume":"27 2","pages":""},"PeriodicalIF":2.2,"publicationDate":"2026-01-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12826989/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146029491","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Deep learning has advanced breast tumor prediction research, but traditional single-modality models limit feature diversity and accuracy.
Purpose
To develop and validate a multimodal deep learning approach that combines mammography and ultrasound imaging for improved breast tumor classification and enhanced clinical decision-making.
Methods
This retrospective study analyzed 663 female patients with breast lesions from 2018 to 2021, including 384 benign and 279 malignant cases. The two-stage prediction model employed improved modality-specific attention mechanisms: efficient channel attention (ECA-Net) for ultrasound and convolutional block attention module (CBAM) for mammography. The fused features were input into a stacking ensemble module with logistic regression (LR), support vector machine (SVM), random forest (RF), and Extra-Trees (ET) as base learners, and multilayer perceptron (MLP) neural network as meta-learner. Data was divided into training (464), validation (133), and test (66) sets with a 7:2:1 ratio.
Results
The proposed multimodal prediction model—mammography ultrasound (MPM-MU) achieved superior performance with an area under the receiver operating characteristic (ROC) Curve (AUC) of 87.9 ± 0.21%, representing improvements of 13.4% and 15.6% over attention-enhanced mammography (74.5%) and ultrasound (72.3%) models, respectively. Ablation studies confirmed the effectiveness of both multimodal feature fusion and attention mechanisms in enhancing diagnostic performance.
Conclusions
The multimodal prediction model—mammography ultrasound (MPM-MU) with modality-specific attention mechanisms demonstrated superior performance in distinguishing between benign and malignant breast tumors compared to single-modality approaches. This approach assists radiologists in improving breast lesion classification accuracy and enhancing clinical decision-making, potentially reducing unnecessary biopsies and improving diagnostic consistency.
{"title":"Multimodal deep learning for breast tumor classification: Integrating mammography and ultrasound for enhanced diagnostic accuracy","authors":"Yu Yan, Yichen Xu, Ge Fang, Xu He, Yifei Qian, Wenwen Zhu","doi":"10.1002/acm2.70464","DOIUrl":"10.1002/acm2.70464","url":null,"abstract":"<div>\u0000 \u0000 \u0000 <section>\u0000 \u0000 <h3> Background</h3>\u0000 \u0000 <p>Deep learning has advanced breast tumor prediction research, but traditional single-modality models limit feature diversity and accuracy.</p>\u0000 </section>\u0000 \u0000 <section>\u0000 \u0000 <h3> Purpose</h3>\u0000 \u0000 <p>To develop and validate a multimodal deep learning approach that combines mammography and ultrasound imaging for improved breast tumor classification and enhanced clinical decision-making.</p>\u0000 </section>\u0000 \u0000 <section>\u0000 \u0000 <h3> Methods</h3>\u0000 \u0000 <p>This retrospective study analyzed 663 female patients with breast lesions from 2018 to 2021, including 384 benign and 279 malignant cases. The two-stage prediction model employed improved modality-specific attention mechanisms: efficient channel attention (ECA-Net) for ultrasound and convolutional block attention module (CBAM) for mammography. The fused features were input into a stacking ensemble module with logistic regression (LR), support vector machine (SVM), random forest (RF), and Extra-Trees (ET) as base learners, and multilayer perceptron (MLP) neural network as meta-learner. Data was divided into training (464), validation (133), and test (66) sets with a 7:2:1 ratio.</p>\u0000 </section>\u0000 \u0000 <section>\u0000 \u0000 <h3> Results</h3>\u0000 \u0000 <p>The proposed multimodal prediction model—mammography ultrasound (MPM-MU) achieved superior performance with an area under the receiver operating characteristic (ROC) Curve (AUC) of 87.9 ± 0.21%, representing improvements of 13.4% and 15.6% over attention-enhanced mammography (74.5%) and ultrasound (72.3%) models, respectively. Ablation studies confirmed the effectiveness of both multimodal feature fusion and attention mechanisms in enhancing diagnostic performance.</p>\u0000 </section>\u0000 \u0000 <section>\u0000 \u0000 <h3> Conclusions</h3>\u0000 \u0000 <p>The multimodal prediction model—mammography ultrasound (MPM-MU) with modality-specific attention mechanisms demonstrated superior performance in distinguishing between benign and malignant breast tumors compared to single-modality approaches. This approach assists radiologists in improving breast lesion classification accuracy and enhancing clinical decision-making, potentially reducing unnecessary biopsies and improving diagnostic consistency.</p>\u0000 </section>\u0000 </div>","PeriodicalId":14989,"journal":{"name":"Journal of Applied Clinical Medical Physics","volume":"27 1","pages":""},"PeriodicalIF":2.2,"publicationDate":"2026-01-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://aapm.onlinelibrary.wiley.com/doi/epdf/10.1002/acm2.70464","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145998251","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Odette Rios-Ibacache, James Manalad, Kayla O'Sullivan-Steben, Emily Poon, Luc Galarneau, Julia Khriguian, George Shenouda, John Kildea
Background
Head and neck cancer (HNC) patients undergoing radiotherapy (RT) may experience anatomical changes during treatment, which can compromise the validity of the initial treatment plan, necessitating replanning. However, ad hoc replanning disrupts clinical workflows and increases workload. Currently, no standardized method exists to quantify anatomical variation that necessitates replanning.
Purpose
This project aimed to create geometrical metrics to describe anatomical changes in HNC patients during RT. The usefulness of these metrics was evaluated by a univariate analysis and through machine learning (ML) models to predict the need for replanning.
Methods
A cohort of 150 HNC patients treated at McGill University Health Centre was analyzed. Based on the shapes of the RT structures (body, PTV, mandible, neck, and submandibular contours), we developed 43 metrics and automatically calculated them through a Python pipeline that we called HNGeoNatomyX. Univariate analysis using linear regression was conducted to obtain the rate of change of each metric. We also obtained the relative variation of each metric between the pre-treatment and replanning-requested scans. Fraction-specific ML models (incorporating information available up to and including the specific fraction) for fractions 5, 10, and 15 were built using metrics, clinical data, and feature selection techniques. Model performance was estimated with repeated stratified 5-fold cross-validation resampling technique and the area under the curve (AUC) of the receiver operating characteristic (ROC) curve.
Results
Univariate analysis showed that body- and neck-related metrics were most predictive of replanning need. Our best specific multivariate models for fractions 5, 10, and 15 yielded testing scores of 0.82, 0.70, and 0.79, respectively. Our models early predicted replanning for 76% of the true positives.
Conclusions
The created metrics have the potential to characterize and distinguish which patients will necessitate RT replanning. They show promise in guiding clinicians to evaluate RT replanning for HNC patients and streamline workflows.
{"title":"Quantification of head and neck cancer patients' anatomical changes during radiotherapy: Toward the prediction of replanning need","authors":"Odette Rios-Ibacache, James Manalad, Kayla O'Sullivan-Steben, Emily Poon, Luc Galarneau, Julia Khriguian, George Shenouda, John Kildea","doi":"10.1002/acm2.70465","DOIUrl":"10.1002/acm2.70465","url":null,"abstract":"<div>\u0000 \u0000 \u0000 <section>\u0000 \u0000 <h3> Background</h3>\u0000 \u0000 <p>Head and neck cancer (HNC) patients undergoing radiotherapy (RT) may experience anatomical changes during treatment, which can compromise the validity of the initial treatment plan, necessitating replanning. However, ad hoc replanning disrupts clinical workflows and increases workload. Currently, no standardized method exists to quantify anatomical variation that necessitates replanning.</p>\u0000 </section>\u0000 \u0000 <section>\u0000 \u0000 <h3> Purpose</h3>\u0000 \u0000 <p>This project aimed to create geometrical metrics to describe anatomical changes in HNC patients during RT. The usefulness of these metrics was evaluated by a univariate analysis and through machine learning (ML) models to predict the need for replanning.</p>\u0000 </section>\u0000 \u0000 <section>\u0000 \u0000 <h3> Methods</h3>\u0000 \u0000 <p>A cohort of 150 HNC patients treated at McGill University Health Centre was analyzed. Based on the shapes of the RT structures (body, PTV, mandible, neck, and submandibular contours), we developed 43 metrics and automatically calculated them through a Python pipeline that we called <i>HNGeoNatomyX</i>. Univariate analysis using linear regression was conducted to obtain the rate of change of each metric. We also obtained the relative variation of each metric between the pre-treatment and replanning-requested scans. Fraction-specific ML models (incorporating information available up to and including the specific fraction) for fractions 5, 10, and 15 were built using metrics, clinical data, and feature selection techniques. Model performance was estimated with repeated stratified 5-fold cross-validation resampling technique and the area under the curve (AUC) of the receiver operating characteristic (ROC) curve.</p>\u0000 </section>\u0000 \u0000 <section>\u0000 \u0000 <h3> Results</h3>\u0000 \u0000 <p>Univariate analysis showed that body- and neck-related metrics were most predictive of replanning need. Our best specific multivariate models for fractions 5, 10, and 15 yielded testing scores of 0.82, 0.70, and 0.79, respectively. Our models early predicted replanning for 76% of the true positives.</p>\u0000 </section>\u0000 \u0000 <section>\u0000 \u0000 <h3> Conclusions</h3>\u0000 \u0000 <p>The created metrics have the potential to characterize and distinguish which patients will necessitate RT replanning. They show promise in guiding clinicians to evaluate RT replanning for HNC patients and streamline workflows.</p>\u0000 </section>\u0000 </div>","PeriodicalId":14989,"journal":{"name":"Journal of Applied Clinical Medical Physics","volume":"27 1","pages":""},"PeriodicalIF":2.2,"publicationDate":"2026-01-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12811973/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145989318","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
<div> <section> <h3> Background</h3> <p>Brain metastases (BMs) are manually contoured during stereotactic radiosurgery (SRS) treatment planning, which is both time-consuming and potentially inconsistent. To address these challenges, researchers have been actively developing deep learning-based approaches for the detection and segmentation of BMs. However, a comprehensive comparative analysis of deep learning models across different frameworks remains largely absent in the current literature. This study aimed to evaluate and compare deep learning models based on different frameworks for the detection and segmentation of BMs in T1-contrast MRI.</p> </section> <section> <h3> Materials and Methods</h3> <p>Eight deep learning models, based on CNN, Transformer, or Mamba architectures, were trained and validated for the task of detecting and segmenting brain metastatic lesions in T1-contrast MRI. A total of 934 patients were included, with 667 cases from publicly available datasets and 267 cases from our institution, designated for training and testing, respectively. Data were retrospectively collected and organized at our institution, and GTV defined as the total BM tumor volume delineated by the physician at the time of stereotactic radiosurgery (SRS). Additionally, labels in the publicly available dataset were modified under clinician guidance to create a BM GTV that met clinical criteria to improve ground-truth accuracy. A BM was considered detected if the ground-truth contour overlapped with a predicted structure. Sensitivity at both the patient-level (proportion of patients with at least one lesion detected) and lesion-level (proportion of ground-truth lesions detected) were used to evaluate BM detection. Segmentation performance was assessed using several metrics: dice similarity coefficient (DSC), positive predictive value (PPV), surface DSC (sDSC), and Hausdorff distance 95% (HD95). The performance across different BM diameters was also evaluated.</p> </section> <section> <h3> Results</h3> <p>Among the eight deep learning models, the U-Mamba (Bot) achieved a lesion-level sensitivity of 0.796 (95% CI: 0.779–0.812) for all sizes of BM, which was significantly higher than that of the other models, with a false positive rate of 2.46 ± 4.96 per patient. Further stratification by metastasis diameter, the sensitivity was 0.505 for BMs < 3 mm, 0.797 for BMs between 3 and 6 mm, and 0.885 for BMs between 6 and 9 mm. Moreover, U-Mamba (Enc) demonstrated significantly higher lesion-level segmentation performance, with DSC value of 0.632 ± 0.224. In terms of tumor boundary segmentation, nnU-Netv2 achieved the best performance, with Surface DSC and HD95 val
{"title":"Evaluation of deep learning-based methods for automatic detection and segmentation of brain metastases in T1-contrast MRI for stereotactic radiosurgery","authors":"Zhifeng Xu, Yuqi Yang, Guanjie Wang, Yi Xue, Xinyang Zhang, Yang Dong, Liming Xu, Qi Wang, Wei Wang, Zhiyong Yuan, Sheng Huang","doi":"10.1002/acm2.70459","DOIUrl":"10.1002/acm2.70459","url":null,"abstract":"<div>\u0000 \u0000 \u0000 <section>\u0000 \u0000 <h3> Background</h3>\u0000 \u0000 <p>Brain metastases (BMs) are manually contoured during stereotactic radiosurgery (SRS) treatment planning, which is both time-consuming and potentially inconsistent. To address these challenges, researchers have been actively developing deep learning-based approaches for the detection and segmentation of BMs. However, a comprehensive comparative analysis of deep learning models across different frameworks remains largely absent in the current literature. This study aimed to evaluate and compare deep learning models based on different frameworks for the detection and segmentation of BMs in T1-contrast MRI.</p>\u0000 </section>\u0000 \u0000 <section>\u0000 \u0000 <h3> Materials and Methods</h3>\u0000 \u0000 <p>Eight deep learning models, based on CNN, Transformer, or Mamba architectures, were trained and validated for the task of detecting and segmenting brain metastatic lesions in T1-contrast MRI. A total of 934 patients were included, with 667 cases from publicly available datasets and 267 cases from our institution, designated for training and testing, respectively. Data were retrospectively collected and organized at our institution, and GTV defined as the total BM tumor volume delineated by the physician at the time of stereotactic radiosurgery (SRS). Additionally, labels in the publicly available dataset were modified under clinician guidance to create a BM GTV that met clinical criteria to improve ground-truth accuracy. A BM was considered detected if the ground-truth contour overlapped with a predicted structure. Sensitivity at both the patient-level (proportion of patients with at least one lesion detected) and lesion-level (proportion of ground-truth lesions detected) were used to evaluate BM detection. Segmentation performance was assessed using several metrics: dice similarity coefficient (DSC), positive predictive value (PPV), surface DSC (sDSC), and Hausdorff distance 95% (HD95). The performance across different BM diameters was also evaluated.</p>\u0000 </section>\u0000 \u0000 <section>\u0000 \u0000 <h3> Results</h3>\u0000 \u0000 <p>Among the eight deep learning models, the U-Mamba (Bot) achieved a lesion-level sensitivity of 0.796 (95% CI: 0.779–0.812) for all sizes of BM, which was significantly higher than that of the other models, with a false positive rate of 2.46 ± 4.96 per patient. Further stratification by metastasis diameter, the sensitivity was 0.505 for BMs < 3 mm, 0.797 for BMs between 3 and 6 mm, and 0.885 for BMs between 6 and 9 mm. Moreover, U-Mamba (Enc) demonstrated significantly higher lesion-level segmentation performance, with DSC value of 0.632 ± 0.224. In terms of tumor boundary segmentation, nnU-Netv2 achieved the best performance, with Surface DSC and HD95 val","PeriodicalId":14989,"journal":{"name":"Journal of Applied Clinical Medical Physics","volume":"27 1","pages":""},"PeriodicalIF":2.2,"publicationDate":"2026-01-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12811971/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145989364","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
扫码关注我们
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号