Medical physics最新文献

{"title":"CBCT projection domain metal segmentation for metal artifact reduction using hessian-inspired dual-encoding network with guidance from segment anything model.","authors":"Chen Jiang, Tianling Lyu, Gege Ma, Zhan Wu, Xinyun Zhong, Yan Xi, Yang Chen, Wentao Zhu","doi":"10.1002/mp.17716","DOIUrl":"https://doi.org/10.1002/mp.17716","url":null,"abstract":"<p><strong>Background: </strong>Metal artifact is a prevailing factor reducing the image quality of cone-beam computed tomography (CBCT), which is a widely used medical imaging method. Existing metal artifact reduction (MAR) methods typically contain two steps: segmentation and interpolation. Recent MAR algorithms pay more attention to the interpolation of the metal traces, but metal segmentation is also challenging, especially for CBCT.</p><p><strong>Purpose: </strong>Despite the success of deep learning (DL) in image segmentation, the substantial expense associated with annotating metal traces in the projection domain makes most of these approaches impractical for this task. In this paper, we aim to provide a workflow for DL-based metal-trace segmentation without manually delineated ground truth.</p><p><strong>Methods: </strong>We propose a Hessian-inspired dual-encoding network (HIDE-Net) for CBCT projection-domain metal segmentation with guidance from the segment anything model. Specifically, a Hessian eigenvalue module is designed to incorporate human knowledge about the target metal objects; a dual encoder is designed to better extract marginal information; and an input enhancement module is proposed to enhance the projection domain input for better segmentation. Finally, a SAM-based label preprocessing module is investigated to obtain the training label automatically.</p><p><strong>Results: </strong>The proposed method has been tested on both digital phantom data and clinical CBCT data. Experiments on both datasets demonstrate the efficacy of the proposed method. HIDE-Net achieves improved metal segmentation accuracy than recent segmentation-oriented CNN models. Compared with existing MAR algorithms, the proposed method improves Dice index in projection domain by 3.2 <math><semantics><mo>%</mo> <annotation>$%$</annotation></semantics> </math> , and the RMSE in image domain is reduced by 42 <math><semantics><mo>%</mo> <annotation>$%$</annotation></semantics> </math> .</p><p><strong>Conclusions: </strong>The proposed methods would advance MAR techniques in CBCT and have the potential to push forward the use of intraoperative CBCT in human-handed and robotic-assisted MISS.</p>","PeriodicalId":94136,"journal":{"name":"Medical physics","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2025-02-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143525746","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A workflow to select local tolerance limits by combining statistical process control and error curve model.","authors":"Xin Yi, Yanbo Song, Hanyin Zhang, Haixia Cui, Wenli Lu, Junwu Zhao","doi":"10.1002/mp.17715","DOIUrl":"https://doi.org/10.1002/mp.17715","url":null,"abstract":"&lt;p&gt;&lt;strong&gt;Background: &lt;/strong&gt;Patient-specific quality assurance (QA) is a complicated process specific to personnel, equipment, and procedure. The universal or commonly used tolerance limits may not be applicable to local situations. Therefore, it is a need for a medical physicist to establish appropriate local tolerance limits based on actual situations and quantitatively evaluate the error sensitivity of selected tolerance limits to determine their availability in clinical practice.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Purpose: &lt;/strong&gt;This study aims to develop a comprehensive and scientifically sound methodology for determining appropriate local tolerance limits in patient-specific QA.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Methods and materials: &lt;/strong&gt;A total of 214 RapidArc plans for cervical cancer were selected. Systematic multi-leaf collimator (MLC) positional errors were simulated across eighteen offsets ranging from ± 0.2 to ± 5 mm. Dose verification was conducted on 808 RapidArc plans, and a retrospective review was carried out. Firstly, six commonly used QA metrics in gamma and DVH analysis were extracted from the QA results of 196 error-free RapidArc plans. These QA metrics included GP&lt;sub&gt;10&lt;/sub&gt; (gamma passing rates [GPRs] at 3%/2mm, 10% dose threshold), GP&lt;sub&gt;50&lt;/sub&gt; (GPRs at 3%/2mm, 50% dose threshold), µGI&lt;sub&gt;50&lt;/sub&gt; (mean gamma index at 3%/2mm, 50% dose threshold), PTV&lt;sub&gt;95&lt;/sub&gt; (dose received by 95% of PTV), PTV&lt;sub&gt;5&lt;/sub&gt; (dose received by 5% of PTV) and PTV&lt;sub&gt;mean&lt;/sub&gt; (mean dose received by PTV). Secondly, the statistical process control was used to establish the corresponding tolerance limits for each metric. Then, six error curve models were created based on 360 error-introduced plans to record changes in QA metrics under different magnitudes of MLC positional error. The error range of theoretical detection limits for systematic MLC positional errors was investigated to assess error sensitivity quantitatively using the error curve model. Finally, the process-based tolerance limits of six single QA metrics and four combined QA metrics were validated by using 252 sets of test data. The binary classification performance (error-free/error-introduced) was assessed based on detection rate, accuracy, precision, recall, and f1-score.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Results: &lt;/strong&gt;The theoretical detection limits for process-based tolerance limits of GP&lt;sub&gt;10&lt;/sub&gt;, GP&lt;sub&gt;50&lt;/sub&gt;, µGI&lt;sub&gt;50&lt;/sub&gt;, PTV&lt;sub&gt;95&lt;/sub&gt;, PTV&lt;sub&gt;mean&lt;/sub&gt;, and PTV&lt;sub&gt;5&lt;/sub&gt; were 2.19 mm, 2.71 mm, 3.52 mm, 1.93 mm, 3.20 mm, and 2.15 mm, respectively. In the validation phase, the process-based tolerance limits for PTV&lt;sub&gt;95&lt;/sub&gt; effectively identified systematic MLC positional errors exceeding 0.6 mm with a detection rate of 76.19%, displaying superior performance in binary classification among six single metrics. Regarding combined metrics, the joint evaluation of process-based tolerance limits for GP&lt;sub&gt;10&lt;/sub&gt; and PTV&lt;sub&gt;95&lt;/sub&gt; showed a higher detection rate of 80.16% for systemat","PeriodicalId":94136,"journal":{"name":"Medical physics","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2025-02-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143525742","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":"Radiotherapy dose prediction using off-the-shelf segmentation networks: A feasibility study with GammaPod planning.","authors":"Qingying Wang, Mingli Chen, Mahdieh Kazemimoghadam, Zi Yang, Kangning Zhang, Xuejun Gu, Weiguo Lu","doi":"10.1002/mp.17711","DOIUrl":"https://doi.org/10.1002/mp.17711","url":null,"abstract":"&lt;p&gt;&lt;strong&gt;Background: &lt;/strong&gt;Radiotherapy requires precise, patient-specific treatment planning to achieve high-quality dose distributions that improve patient outcomes. Traditional manual planning is time-consuming and clinically impractical for performing necessary plan trade-off comparisons, including treatment modality selection, prescription dose settings, and organ at risk (OAR) constraints. A time-efficient dose prediction tool could accelerate the planning process by guiding clinical plan optimization and adjustments. While the deep convolutional neural networks (CNNs) are prominent in radiotherapy dose prediction tasks, most studies have attempted to customize network architectures for different diseases and treatment modalities.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Purpose: &lt;/strong&gt;This study proposes a universal and efficient strategy, Seg2Dose, leveraging a state-of-the-art segmentation network for radiotherapy dose prediction without the need for model architecture modifications. We aim to provide a convenient off-the-shelf dose prediction tool that simplifies the dose prediction process, enhancing planning speed, and plan quality while minimizing the need for extensive coding and customization.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Methods: &lt;/strong&gt;The proposed Seg2Dose consists of three modules: the Adapter, the segmentation network, and the Smoother. Prior to model training, the Adapter processes dose distributions into dose level map with an adjustable interval, which serves as the ground truth of the segmentation network, and generates two input channels: weighted avoidance image and normalized prescribed dose image. The segmentation network predicts dose levels from input channels using the nnU-Net, which was trained, validated and tested on 304, 77, and 64 breast cancer GammaPod treatment plans from 90 patients. The Smoother converts the predicted dose levels into continuous dose distribution with a Gaussian filter. The performance of Seg2Dose models with two different dose level intervals, 2% (Seg2Dose 2%) and 5% (Seg2Dose 5%), was evaluated by the Dice similarity coefficients (DSCs), voxel-based mean absolute percent error (MAPE), dose-volume histogram (DVH) metrics, global 3%/2 mm and 3%/1 mm gamma passing rate (GPR), and a case study including normal and worst cases. Additionally, Seg2Dose was compared with an exciting cutting-edge Cascade 3D (C3D) dose prediction model, which was trained on continuous dose distributions, to investigate the impact of using dose level map.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Results: &lt;/strong&gt;For dose level prediction, Seg2Dose achieved average DSCs of 0.94 and 0.93 for the 2% and 5% intervals, respectively. For dose distribution prediction, both Seg2Dose 2% and Seg2Dose 5% achieved MAPEs within 6% for targets and most OARs, with the exception of the skin, which had the highest MAPE at 8.58% for Seg2Dose 2% and 15.25% for Seg2Dose 5%. The DVH metrics showed consistent findings. The C3D model has a better performance in GPR than Seg2Dose models. However,","PeriodicalId":94136,"journal":{"name":"Medical physics","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2025-02-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143525672","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":"Accuracy of a whole-body single-photon emission computed tomography with a thallium-bromide detector: Verification via Monte Carlo simulations.","authors":"Toshimune Ito, Keitaro Hitomi, Michael Ljungberg, Sousei Kawasaki, Yuka Katayama, Akane Kato, Hirotatsu Tsuchikame, Kentaro Suzuki, Kyosuke Miyazaki, Ritsushi Mogi","doi":"10.1002/mp.17724","DOIUrl":"https://doi.org/10.1002/mp.17724","url":null,"abstract":"<p><strong>Background: </strong>Single-photon emission computed tomography (SPECT) devices equipped with cadmium-zinc-telluride (CZT) detectors achieve high contrast resolution because of their enhanced energy resolution. Recently, thallium bromide (TlBr) has gained attention as a detector material because of its high atomic number and density.</p><p><strong>Purpose: </strong>This study evaluated the clinical applicability of a SPECT system equipped with TlBr detectors using Monte Carlo simulations, focusing on 99mTc and 177Lu imaging.</p><p><strong>Methods: </strong>This study used the Simulation of Imaging Nuclear Detectors Monte Carlo program to compare the imaging characteristics between a whole-body SPECT system equipped with TlBr (T-SPECT) and a system equipped with CZT detectors (C-SPECT). The simulations were performed using a three-dimensional brain phantom and a National Electrical Manufacturers Association body phantom to evaluate 99mTc and 177Lu imaging. The simulation parameters were accurately set by comparing them with the actual measurements.</p><p><strong>Results: </strong>The T-SPECT system demonstrated improved energy resolution and higher detection efficiency than the C-SPECT system. In 99mTc imaging, T-SPECT demonstrated 1.71 times higher photopeak counts and improved contrast resolution. T-SPECT exhibited a significantly lower impact of hole tailing and higher-energy resolution (4.50% for T-SPECT vs. 7.34% for C-SPECT). Furthermore, T-SPECT showed higher peak signal-to-noise ratio (PSNR) and structural similarity (SSIM) values, indicating better image quality. In 177Lu imaging, T-SPECT showed 2.76 times higher photopeak counts and improved energy resolution (3.94% for T-SPECT vs. 5.20% for C-SPECT). T-SPECT demonstrated a higher contrast recovery coefficient (CRC) and contrast-to-noise ratio (CNR) across all acquisition times, maintaining sufficient counts even with shorter acquisition times. Moreover, T-SPECT acquired higher low-frequency values in power spectrum density (PSD), indicating more accurate internal image reproduction.</p><p><strong>Conclusions: </strong>T-SPECT offers superior energy resolution and detection efficiency than C-SPECT. Moreover, T-SPECT can provide higher contrast resolution and sensitivity in clinical imaging with 99mTc and 177Lu. Furthermore, the Monte Carlo simulations are confirmed to be a valuable guide for the development of T-SPECT.</p>","PeriodicalId":94136,"journal":{"name":"Medical physics","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2025-02-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143525743","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":"Hyperpolarized ¹²⁹Xe diffusion-weighted MRI of the lung with 3D golden-angle radial sampling and keyhole reconstruction.","authors":"Luyang Shen, Haidong Li, Yuan Fang, Ming Luo, Yecheng Li, Qian Zhou, Qiuchen Rao, Ming Zhang, Xiuchao Zhao, Lei Shi, Yeqing Han, Fumin Guo, Xin Zhou","doi":"10.1002/mp.17719","DOIUrl":"https://doi.org/10.1002/mp.17719","url":null,"abstract":"<p><strong>Background: </strong>Hyperpolarized (HP) ¹²⁹Xe multiple b-values diffusion-weighted imaging (DWI) facilitates the assessment of pulmonary morphology. However, conventional DWI method, such as 2D GRE DWI, is limited in its application due to the long acquisition time and relatively thick slice.</p><p><strong>Purpose: </strong>To develop a method combining 3D golden-angle radial sampling with keyhole reconstruction (GRSK) for accelerating ¹²⁹Xe multiple b-values DWI and obtaining thinner slice.</p><p><strong>Methods: </strong>For 3D GRSK DWI, 3D kooshball golden-angle radial sampling was used for image acquisition, with each spoke assigned to 1 of 4 b-values, and keyhole method was applied for reconstructing DW images under different b-values. 2D fully sampled GRE DWI and 3D GRSK DWI were obtained in five healthy young volunteers (HYV; 25 [24-26] years). Lung morphological parameter maps, including mean linear intercept (L_m) and surface-to-volume ratio (SVR), were generated using a cylinder model (CM) from the four b-values DW images, and apparent diffusion coefficient (ADC) maps were derived through mono-exponential fitting. Spearman correlation and Bland-Altman analysis were performed to compare L_m, SVR and ADC from 2D GRE and 3D GRSK DWI. In addition, 3D GRSK DWI was applied in five emphysema patients (66 [60-69] years) and five age-matched healthy controls (AMC; 59 [54-68] years), and L_m, SVR and ADC maps were also derived. Wilcoxon rank-sum test was utilized to contrast L_m, SVR and ADC from patients with those from AMC.</p><p><strong>Results: </strong>DW images with an isotropic resolution (5 mm) were obtained with 3D GRSK DWI within 11.4 s. In comparison, 2D GRE DWI acquired four slices with 30 mm thickness in 15.9 s. For L_m, SVR and ADC from 2D GRE and 3D GRSK DWI, the Spearman correlation coefficients were 0.975, 0.900, and 1.000, with corresponding p-values of 0.005, 0.037, and < 0.001, and the Bland-Altman analysis had biases of -3.19%, 1.44%, and -3.71%, respectively. Furthermore, L_m and ADC in patients were significantly higher (p = 0.008 and p = 0.008) than those in AMC, while SVR was notably reduced (p = 0.008).</p><p><strong>Conclusion: </strong>The proposed method could obtain isotropic resolution DW images with four b-values within an 11.4 s breath-hold duration, mitigating the problems of long scan time and large slice thickness in conventional DWI.</p>","PeriodicalId":94136,"journal":{"name":"Medical physics","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2025-02-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143525671","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":"Real-time radiation beam imaging on an MR linear accelerator using quantitative T₁ mapping.","authors":"Brandon T T Tran, Liam S P Lawrence, Shawn Binda, Ryan T Oglesby, Brige P Chugh, Angus Z Lau","doi":"10.1002/mp.17720","DOIUrl":"https://doi.org/10.1002/mp.17720","url":null,"abstract":"&lt;p&gt;&lt;strong&gt;Background: &lt;/strong&gt;Direct three-dimensional imaging of radiation beams could enable more accurate radiation dosimetry. It has been previously reported that changes in T&lt;sub&gt;1&lt;/sub&gt;-weighted magnetic resonance imaging (MRI) intensity could be observed during radiation due to radiochemical oxygen depletion. Quantitative T&lt;sub&gt;1&lt;/sub&gt; mapping could increase sensitivity for dosimetry applications.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Purpose: &lt;/strong&gt;We use an MRI linear accelerator (MR-Linac) to visualize radiation delivery through the real-time effects of dose on the spin-lattice magnetic relaxation time (T&lt;sub&gt;1&lt;/sub&gt;) of water. We quantify the relationships between dose, spin-lattice relaxation rates (R&lt;sub&gt;1&lt;/sub&gt;) and dissolved oxygen concentration to further investigate the mechanisms of T&lt;sub&gt;1&lt;/sub&gt; change.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Methods: &lt;/strong&gt;An ultrapure water phantom and a 1% agarose gel phantom were irradiated and imaged on a 1.5 T Elekta Unity MR-Linac. Radiation plans were created using the Monaco treatment planning system. Images were acquired before, during and after radiation. A dual-echo Look-Locker inversion recovery pulse sequence was used for simultaneous dynamic T&lt;sub&gt;1&lt;/sub&gt;/B&lt;sub&gt;0&lt;/sub&gt; mapping. The change in R&lt;sub&gt;1&lt;/sub&gt; with respect to dose (∆R&lt;sub&gt;1&lt;/sub&gt;/∆Dose) and the radiochemical oxygen depletion (ROD = ∆O&lt;sub&gt;2&lt;/sub&gt;/∆Dose) were measured. The relaxivity of oxygen (r&lt;sub&gt;1,O2&lt;/sub&gt; = ∆R&lt;sub&gt;1&lt;/sub&gt;/∆O&lt;sub&gt;2&lt;/sub&gt;) in water was also measured in a separate experiment with samples of various dissolved oxygen concentrations. The minimum measurable dose over a 20-min period was estimated using a single-tailed 99th quantile Student's t-distribution.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Results: &lt;/strong&gt;Changes to R&lt;sub&gt;1&lt;/sub&gt; were found to be spatiotemporally correlated to the predicted delivered radiation dose and persisted for at least 1 h after radiation. A complex dose plan could be imaged in the 1% agarose gel phantom, as the gel limits diffusion and convective mixing. In water, the ∆R&lt;sub&gt;1&lt;/sub&gt;/∆Dose was found to be -1.0 × 10&lt;sup&gt;-4&lt;/sup&gt; s&lt;sup&gt;-1&lt;/sup&gt;/Gy, the r&lt;sub&gt;1,O2&lt;/sub&gt; was found to be 5.4 × 10&lt;sup&gt;-3&lt;/sup&gt; s&lt;sup&gt;-1&lt;/sup&gt;/(mg/L), and the ROD was found to be -0.010 (mg/L)/Gy. Both r&lt;sub&gt;1,O2&lt;/sub&gt; and ROD agree with published values. However, combining these two values yields a predicted ∆R&lt;sub&gt;1&lt;/sub&gt;/∆Dose of -5.4 × 10&lt;sup&gt;-5&lt;/sup&gt; s&lt;sup&gt;-1&lt;/sup&gt;/Gy, indicating that radiochemical oxygen depletion alone under-predicts the MRI effect. The detection limit of R&lt;sub&gt;1&lt;/sub&gt; was 1.1 × 10&lt;sup&gt;-3&lt;/sup&gt; s&lt;sup&gt;-1&lt;/sup&gt; which corresponded to a single-voxel minimum detectable dose of 11.1 Gy for this specific sequence.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Conclusion: &lt;/strong&gt;Quantitative T&lt;sub&gt;1&lt;/sub&gt; mapping was used to image radiation dose patterns in real-time in water and agarose gel. Radiochemical oxygen depletion only partially explains the T&lt;sub&gt;1&lt;/sub&gt; changes measured. Agarose gel could be used as a simple system for three-dimensional patient-specific quality assur","PeriodicalId":94136,"journal":{"name":"Medical physics","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2025-02-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143517774","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":"Generalizability of convolutional neural network-based model observer in breast tomosynthesis across volume glandular fractions and signal sizes.","authors":"Hanjoo Jang, Jongduk Baek","doi":"10.1002/mp.17725","DOIUrl":"https://doi.org/10.1002/mp.17725","url":null,"abstract":"&lt;p&gt;&lt;strong&gt;Background: &lt;/strong&gt;In our previous study, we proposed a convolutional neural network (CNN)-based model observer for signal known statistically (SKS) and background known statistically (BKS) tasks to assess the detection performance of breast tomosynthesis systems by varying acquisition angles and the number of projections at a constant dose level. Despite demonstrating the significant potential of the CNN-based model observer in approximating the ideal observer (IO) performance, further research is required to extend its applicability to clinically relevant tasks and validate its robustness across diverse imaging scenarios.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Purpose: &lt;/strong&gt;Exploring the generalizability of the CNN-based model observer is essential for advancing its practical utility in diagnostic imaging. In this work, we explored the generalizability of a CNN-based model observer for SKS and BKS detection tasks in breast tomosynthesis images with two different volume glandular fractions (i.e., VGFs: 30% and 50%), and two different sizes of spiculated signals (i.e., 1 and 2 mm). These efforts aim to provide deeper insights into the factors that influence network optimization for consistent and robust detection performance.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Methods: &lt;/strong&gt;Five different network architectures were used to verify whether optimizing the match between the receptive field (RF) size and signal size would enhance the detection performance; the networks were designed in terms of theoretical receptive field (TRF) size. The detection performance of the CNN-based model observer was compared to that of the Hotelling observer (HO) under various training and testing schemes to observe the key factors in optimizing the network to enhance its generalizability.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Results: &lt;/strong&gt;Throughout the study, we demonstrated that each network focuses more on discriminating the presence of similarly sized signals over achieving robustness to noise variations during training. The CNN-based model observer showed better detection performance compared to the HO, except when the trained and tested datasets incorporated differently sized signals. Networks trained on datasets involving signals of both sizes resulted in better generalizability compared to those trained on mixed-VGF datasets (i.e., datasets comprising both VGFs). Contrary to our assumption, the match between the TRF size and signal size did not improve the detection performance. This led to exploring the effective receptive field (ERF) size of the network as a descriptive metric of network generalizability, using pixelwise gradient activation mapping (pGrad-CAM). We showed that a relationship between the ERF size and signal size exists, thus presenting its clear relevance to the detection performance of the CNN-based model observer.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Conclusions: &lt;/strong&gt;Networks trained on datasets sharing similarly sized signals exhibited optimal detection performance, but showed limited generalizability whe","PeriodicalId":94136,"journal":{"name":"Medical physics","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2025-02-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143525749","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A robust optimization model for intensity-modulated radiotherapy: Cheap-Minimax.","authors":"Andrés C Sevilla, Gonzalo Cabal, Niklas Wahl, María E Puerta, Juan C Rivera","doi":"10.1002/mp.17709","DOIUrl":"https://doi.org/10.1002/mp.17709","url":null,"abstract":"&lt;p&gt;&lt;strong&gt;Background: &lt;/strong&gt;Over the past three decades, the intensity-modulated radiotherapy (IMRT) has become a standard technique, enabling highly conformal dose distributions tailored to specific clinical objectives. Despite these advancements, IMRT treatment plans are significantly susceptible to uncertainties during both the planning and delivery phases. The most commonly used strategy to address these uncertainties is the margin-based or planning target volume (PTV) approach, which relies on the so-called dose cloud approximation. However, the PTV concept has notable limitations, particularly in complex scenarios where target volumes are superficial or located near critical structures. In contrast, the advent of intensity-modulated particle therapy has driven the development of robust optimization models, which have emerged as a promising alternative for managing uncertainties. Among these, the worst-case scenario or minimax strategy is the most widely employed. While minimax can be directly applied to photon treatments, its use in IMRT often leads to overly conservative plans or plans that are very similar to those obtained using the conventional margin-based PTV approach.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Purpose: &lt;/strong&gt;In this work, we present a robust optimization model particularly suitable for photon treatments. The new approach, called Cheap-Minimax, is a generalization of the minimax strategy used for particle therapy and aims to improve the balance between plan robustness and the price of robustness in terms of dose to organs at risk (OARs), an issue particularly pronounced in photon treatments.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Methods: &lt;/strong&gt;The c-minimax model was implemented in the MatRad treatment planning system, developed at the German Cancer Research Center (DKFZ). It was applied to 20 clinical cases, comprising 5 prostate cancer cases and 15 breast cancer cases. The results were compared with those obtained using the conventional minimax model and the PTV-based approach.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Results: &lt;/strong&gt;For prostate cancer cases, the c-minimax model maintained a robustness comparable to the PTV approach, while achieving a 20% reduction in &lt;math&gt; &lt;semantics&gt;&lt;msub&gt;&lt;mi&gt;V&lt;/mi&gt; &lt;mrow&gt;&lt;mn&gt;40&lt;/mn&gt; &lt;mspace&gt;&lt;/mspace&gt; &lt;mtext&gt;Gy&lt;/mtext&gt;&lt;/mrow&gt; &lt;/msub&gt; &lt;annotation&gt;$V_{40 , text{Gy}}$&lt;/annotation&gt;&lt;/semantics&gt; &lt;/math&gt; for the rectum and a 10% reduction in &lt;math&gt; &lt;semantics&gt;&lt;msub&gt;&lt;mi&gt;V&lt;/mi&gt; &lt;mrow&gt;&lt;mn&gt;60&lt;/mn&gt; &lt;mspace&gt;&lt;/mspace&gt; &lt;mtext&gt;Gy&lt;/mtext&gt;&lt;/mrow&gt; &lt;/msub&gt; &lt;annotation&gt;$V_{60 , text{Gy}}$&lt;/annotation&gt;&lt;/semantics&gt; &lt;/math&gt; for the bladder compared to the minimax model. In breast cancer cases, the c-minimax model improved robustness by 23.7% relative to the PTV approach and by 18.2% compared to the minimax model. Additionally, the c-minimax model reduced &lt;math&gt; &lt;semantics&gt;&lt;msub&gt;&lt;mi&gt;V&lt;/mi&gt; &lt;mrow&gt;&lt;mn&gt;20&lt;/mn&gt; &lt;mspace&gt;&lt;/mspace&gt; &lt;mtext&gt;Gy&lt;/mtext&gt;&lt;/mrow&gt; &lt;/msub&gt; &lt;annotation&gt;$V_{20 , text{Gy}}$&lt;/annotation&gt;&lt;/semantics&gt; &lt;/math&gt; for the ipsilateral lung by 3.7% and the mean hea","PeriodicalId":94136,"journal":{"name":"Medical physics","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2025-02-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143517762","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":"Attentive feature interaction based persistent homology-augmented network for esophageal cancer lesion detection.","authors":"Chen Huang, Fuce Guo, Shengmei Lin, Yongmei Dai, Qianshun Chen, Shu Zhang, Xunyu Xu","doi":"10.1002/mp.17707","DOIUrl":"https://doi.org/10.1002/mp.17707","url":null,"abstract":"<p><strong>Background: </strong>Detecting lesions in esophageal cancer (EC) is important in guiding subsequent treatment. Deep learning methods based on convolutional neural networks (CNNs) and vision transformer (ViT) have made remarkable strides in the field of medical image analysis due to their powerful representational capabilities. However, without prior knowledge, both traditional CNNs and ViTs are susceptible to disregarding critical anatomical information, including loops and voids. The current methods, combined with persistent homology (PH), have been proposed to address certain limitations, but they neglect the inconsistencies among features caused by the lack of interaction between features, ultimately leading to a reduction in the model's generalization ability.</p><p><strong>Purpose: </strong>To address these challenges, we propose a novel framework, combined with PH and feature interaction, for identifying EC lesions from 3D CT images. The goal is to enhance the predictive capability of existing deep learning models by incorporating both topological information from PH and effective feature interaction mechanisms.</p><p><strong>Methods: </strong>We applied cube-wise classification techniques to improve the detection of lesions associated with EC. The proposed framework consists of two fundamental modules: (1) persistence diagram cross-attention encoder (PDCAE) that completely encodes the persistence diagram (PD) created by PH through cross-attention. (2) recalibration guidance module (RGM) connecting the PH features with the image features efficiently to remove inconsistencies.</p><p><strong>Results: </strong>The experimental results show that the proposed modules significantly enhance the predictive capability of standard backbone networks, and outperform the state-of-the-art classification network.</p><p><strong>Conclusions: </strong>This work highlights the potential of combining topological data analysis with deep learning for medical image analysis tasks. More potential downstream tasks that can utilize topological relationships remain to be explored in the future.</p>","PeriodicalId":94136,"journal":{"name":"Medical physics","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2025-02-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143517764","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":"Evaluation of model uncertainty in AI-based synthetic CT generation from CBCT for abdominal adaptive radiotherapy.","authors":"Paulo Quintero, Laura Cerviño, Hao Zhang, Wendy Harris","doi":"10.1002/mp.17721","DOIUrl":"https://doi.org/10.1002/mp.17721","url":null,"abstract":"&lt;p&gt;&lt;strong&gt;Background: &lt;/strong&gt;The synthesis of CT from CBCT images using AI methods has been explored in radiotherapy to improve adaptive workflows. However, the model training process can be particularly challenging for the abdominal region due to dataset disparities between CT and CBCT images caused by organ motion, low soft tissue contrast, and inconsistencies in air volumes. These factors might impact the implicit prediction uncertainties, which are not actively considered on the synthetized images, overlooking poorly predicted image regions that might lead to inaccuracies in the dose calculation.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Purpose: &lt;/strong&gt;To evaluate the impact of the model uncertainty on the predicted Hounsfield Units (HU) and dose calculation on synthetic CT (sCT) for abdominal patients.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Methods: &lt;/strong&gt;CBCT images from 65 abdominal patients were retrospectively used to generate sCT images. Rigid image registration (RIR) and deformable image registration (DIR) were individually implemented to create two datasets (D1 and D2) to train (80%), validate (10%), and test (10%) three models (M1: Unet, M2: Bayes-Unet, M3: cycle-GAN). Treatment plans were made on the ground truth CT (&lt;sub&gt;GT&lt;/sub&gt;CT) and the sCTs for dose calculation comparison. The model performance was evaluated with mean absolute error (MAE) and root mean square error (RMSE), and the sCT quality was verified with structural similarity index measure (SSIM). Gamma index (2%/2  mm), D95% of PTV, and D&lt;sub&gt;mean&lt;/sub&gt; of liver were evaluated and compared between the plans calculated on the &lt;sub&gt;GT&lt;/sub&gt;CT and the sCT. The voxel-wise uncertainty map for M1 and M3 were generated by calculating the standard variation of each voxel from training the model independently ten times. For M2 the Monte Carlo DropConnect method was implemented with 100 iterations. Finally, the uncertainty was associated with the accuracy of CT numbers and dose calculation.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Results: &lt;/strong&gt;Across the three models {M1, M2, M3} trained with D1 and D2, the MAE were {50.9 ± 13.3} and {40.9 ± 11.5}, respectively, the RMSE were {68.3 ± 13.5} and {62.2 ± 10.7}, respectively, and the SSIM were {0.89 ± 0.05} and {0.94 ± 0.05}, respectively. For D1 and D2, the gamma rates were {96.3 ± 1.04} and {97.3 ± 0.2}, respectively. No major differences in DVH were noticed between &lt;sub&gt;GT&lt;/sub&gt;CT and sCT (p &lt; 00.1). The correlation between the whole sCT uncertainty maps and gamma index was statistically significant (Spearman's coefficient = 0.84, p &lt; 0.001) and weak between the target volume uncertainty and gamma index (Spearman's coefficient = 0.01, p = 0.89).&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Conclusion: &lt;/strong&gt;Using DIR resulted in improved performance across all three models. Metrics used to evaluate synthetic image accuracy might not reflect the uncertainty implications in image quality and dose calculations, which suggests the benefit of displaying uncertainty errors in AI generated sCT as a potential strategy to improve t","PeriodicalId":94136,"journal":{"name":"Medical physics","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2025-02-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143517773","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}