Medical physics最新文献

{"title":"Experimental validation of a comprehensive fluoroscopy peak skin dose model using four different computational phantoms.","authors":"Daniel Vergara, Rasha S Makkia, Zhimin Li, Matthew Hoerner, Muhannad N Fadhel","doi":"10.1002/mp.17737","DOIUrl":"https://doi.org/10.1002/mp.17737","url":null,"abstract":"<p><strong>Background: </strong>Accurately determining the Peak Skin Dose (PSD) delivered to the patient during Fluoroscopically Guided Interventional Procedures (FGIP) is crucial for assessing potential radiation-induced skin injuries and determining the necessary follow-up care for exposed patients.</p><p><strong>Purpose: </strong>This study evaluates the accuracy of PSD estimation model for FGIPs using mathematical and anthropomorphic computational phantoms that mimic the dimensions of the imaged patient and provides their description.</p><p><strong>Methods: </strong>The modeling of the FGIP and calculation of peak skin dose involved extracting geometric parameters like primary and secondary angulation, fields sizes, and table shifts, as well as dosimetric parameters such as tube voltage, Air Kerma, Kerma Area Product, and additional filtration of the FGIP stored in a dose tracking system. Computational phantoms were employed to represent the patient anatomy and their axes scaled to match the patient dimensions. The first, a hybrid computational human phantom (HCHP) was developed using Rhinoceros 6.0, generating a 3D surface skin model derived from an adult International Commission on Radiological Protection (ICRP) reference voxel-male-phantom. Three other computational (mathematical) phantoms with cylindrical, ellipsoidal, and semi-ellipsoidal geometries were created using MATLAB software and employed to calculate PSD. Dose-distribution mapping was performed on all phantoms using MATLAB software, following the guidelines outlined in the summary of a joint report by AAPM TG357 and EFOMP by Andersson et al. To improve the PSD model accuracy, measured Kerma correction factors (KCF) that account for backscatter and table attenuation were incorporated for all radiation fields. Two FGIPs were conducted utilizing a male anthropomorphic phantom. Thermoluminescent Dosimeters (TLD) were strategically positioned in a grid pattern on the posterior surface of the phantom to serve as reference measurements. Traditional methods, in which all fields overlap, intersect the table and phantom, were also used to calculate the PSD. The resulting skin doses, derived from the HCHP, mathematical phantoms, and traditional methodologies, were then compared against the corresponding reference measurements for a comprehensive evaluation.</p><p><strong>Results: </strong>The results showed that the PSD calculations obtained through the HCHP, cylindrical, ellipsoidal, and semi-ellipsoidal phantoms were 3.163, 3.085, 2.952, and 3.095 Gy, respectively, for the first FGIP and 3.728, 3.722, 3.598, 3.720 Gy, respectively, for the second FGIP. In comparison, the measured PSD using TLDs was 3.161 and 3.713 Gy for the first and second FGIP. The use of the HCHP-introduced PSD differences of 0.1% and 0.4%, and the mathematical phantoms yielded differences of -2.4% and 0.3%, 6.6% and -3.1%, and -2.1% and 0.2% for the cylindrical, ellipsoidal, and semi-ellipsoidal phantoms, respective","PeriodicalId":94136,"journal":{"name":"Medical physics","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2025-03-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143652824","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":"Analytical model for pulse pileup spectra and count statistics in photon counting detectors with seminonparalyzable behavior.","authors":"Yirong Yang, Norbert J Pelc, Adam S Wang","doi":"10.1002/mp.17746","DOIUrl":"https://doi.org/10.1002/mp.17746","url":null,"abstract":"<p><strong>Background: </strong>Photon counting detectors (PCDs) with energy discriminating capabilities enable quantitative imaging of materials. However, the accuracy of estimates may be substantially degraded due to pulse pileup effects (PPEs) at high count rates. Accurate description of the output spectrum and count rate behavior of a PCD subject to pulse pileup is crucial to the development of photon counting computed tomography (PCCT).</p><p><strong>Purpose: </strong>This study presents a fully analytical model to accurately predict the pulse pileup spectrum and count statistics (mean and covariance of energy-binned counts) for a non-paralyzable detector with nonzero pulse length and, therefore, seminonparalyzable behavior, that is, retriggering of dead time by pulses incident during the previous dead time.</p><p><strong>Methods: </strong>We recursively computed the probability density function (PDF) of pulse pileup spectra at different pulse pileup orders. To do this, we considered the following factors: the unipolar pulse shape, the incident pulse spectrum, the distribution of time intervals between incident pulses, and the trigger threshold. We then derived the count rate and spectrum-dependent expression of total count statistics (mean and variance of total counts) based on renewal theory. We simulated a non-paralyzable PCD using Monte Carlo simulation to separately validate the spectrum and count statistics model outputs. Finally, we investigated the model accuracy in predicting material decomposition noise using the Cramér-Rao lower bound (CRLB) and a multibin system model. A comparison between predictions of the proposed model and Monte Carlo simulation is presented.</p><p><strong>Results: </strong>The results show excellent agreement between the proposed model prediction of pulse pileup spectrum and count statistics and Monte Carlo simulation for relative count rates (the average number of counts detected during one dead time, <math> <semantics><mrow><mi>λ</mi> <mi>τ</mi></mrow> <annotation>$lambda tau $</annotation></semantics> </math> ) of up to <math> <semantics><mrow><mn>2.5</mn></mrow> <annotation>$2.5$</annotation></semantics> </math> . The coefficient of variation (CV) values between the spectra from model prediction and Monte Carlo simulation are less than <math> <semantics><mrow><mn>9</mn> <mo>%</mo></mrow> <annotation>$9% $</annotation></semantics> </math> , and the coefficient of determination values, <math> <semantics><msup><mi>R</mi> <mn>2</mn></msup> <annotation>${R}^2$</annotation></semantics> </math> , between the count statistics from model prediction and Monte Carlo simulation are greater than <math> <semantics><mrow><mn>0.99</mn></mrow> <annotation>$0.99$</annotation></semantics> </math> . The proposed model also accurately predicts material decomposition noise for a non-paralyzable PCD for relative count rates of up to <math> <semantics><mrow><mn>1.0</mn></mrow> <annotation>$1.0$</annotation></semantics> </math> ","PeriodicalId":94136,"journal":{"name":"Medical physics","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2025-03-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143652893","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":"Quantifying the dosimetric accuracy of expiration-gated stereotactic lung radiotherapy.","authors":"Daan Hoffmans, Isabel Remmerts de Vries, Max Dahele, Wilko Verbakel","doi":"10.1002/mp.17743","DOIUrl":"https://doi.org/10.1002/mp.17743","url":null,"abstract":"<p><strong>Background: </strong>In stereotactic body radiotherapy, a form of motion management is often applied to mobile lung tumors. Gated radiotherapy is such form of motion management in which the radiation beam is switched on or off depending on the actual tumor position. Compared to inspiration, the tumor position is typically more stable during expiration. Also, the tumor spends more time in expiration position. Therefore, we often consider expiration-gating for patients with relatively large tumor motion.</p><p><strong>Purpose: </strong>We validated dosimetric accuracy of expiration-gated stereotactic lung radiotherapy by means of phantom measurements and modeling the effects of residual motion in patients.</p><p><strong>Methods: </strong>Dose profiles from film measurements in a respiratory-motion phantom were compared to dose calculations, for different expiration gating methods. Fluoroscopic real-time tumor tracking was used to produce a convolution kernel which was applied to the calculated dose distribution to model dosimetric effects of residual motion. This convolution method was validated against film measurements and then retrospectively applied to clinical tumor tracking data of five patients. In addition, clinical tumor motion data was manipulated to simulate the effect of a short breathing period of 2 s and prolonged gating latency of 500 ms.</p><p><strong>Results: </strong>A good agreement between calculated and measured dose was found when amplitude gating was used (100% gamma pass rate, 3%/2 mm). For phase gating, good agreement required a stable breathing period. Measurements showed good performance of the convolution method (gamma pass rate > 99%). For the clinical data, we found a maximal dose shift of 2.4 mm, introduced by residual tumor motion or respiratory drift. For all patients, the size of the ITV was adequate to account for this dose shift. Simulating higher breathing speed in combination with large latency values resulted in dosimetric shifts that were larger than the PTV margin.</p><p><strong>Conclusion: </strong>Amplitude gating is robust for irregular breathing patterns. Expiration-gating is a dosimetrically accurate method of treatment delivery provided that during delivery there is a prompt reaction to respiratory drift and the latency of the gating system is short.</p>","PeriodicalId":94136,"journal":{"name":"Medical physics","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2025-03-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143652898","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":"Flexible and modular PET: Evaluating the potential of TOF-DOI panel detectors.","authors":"Gašper Razdevšek, Georges El Fakhri, Thibault Marin, Rok Dolenec, Matic Orehar, Yanis Chemli, Alberto Giacomo Gola, David Gascon, Stan Majewski, Rok Pestotnik","doi":"10.1002/mp.17741","DOIUrl":"https://doi.org/10.1002/mp.17741","url":null,"abstract":"&lt;p&gt;&lt;strong&gt;Background: &lt;/strong&gt;Panel detectors have the potential to provide a flexible, modular approach to Positron Emission Tomography (PET), enabling customization to meet patient-specific needs and scan objectives. The panel design allows detectors to be positioned close to the patient, aiming to enhance sensitivity and spatial resolution through improved geometric coverage and reduced noncollinearity blurring. Parallax error can be mitigated using depth of interaction (DOI) information.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Purpose: &lt;/strong&gt;One of the key questions the article addresses is: Do panel detectors offer viable clinical imaging capabilities, or does limited angular sampling restrict their utility by causing image distortions and artifacts? Additionally, this article explores the scalability of panel detectors for constructing scanners with a long axial field of view (LAFOV).&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Methods: &lt;/strong&gt;Monte Carlo simulations using GATE software were used to assess the performance of panel detectors with various DOI resolutions and Time-of-Flight (TOF) resolutions as fine as 70 ps. The 30 &lt;math&gt;&lt;semantics&gt;&lt;mo&gt;×&lt;/mo&gt; &lt;annotation&gt;$times$&lt;/annotation&gt;&lt;/semantics&gt; &lt;/math&gt; 30 cm panels comprised pixelated 3 &lt;math&gt;&lt;semantics&gt;&lt;mo&gt;×&lt;/mo&gt; &lt;annotation&gt;$times$&lt;/annotation&gt;&lt;/semantics&gt; &lt;/math&gt; 3 &lt;math&gt;&lt;semantics&gt;&lt;mo&gt;×&lt;/mo&gt; &lt;annotation&gt;$times$&lt;/annotation&gt;&lt;/semantics&gt; &lt;/math&gt; 20 mm LSO crystals. Simulations were run on large high-performance computing clusters (122,000 CPU cores). Open-source CASToR software was used for (TOF MLEM) image reconstruction. The image quality of the scanners was assessed using a range of phantoms (NEMA, Derenzo, XCAT, and a high-resolution brain phantom). The Siemens Biograph Vision PET/CT scanner served as the reference model. The performance of larger 120 &lt;math&gt;&lt;semantics&gt;&lt;mo&gt;×&lt;/mo&gt; &lt;annotation&gt;$times$&lt;/annotation&gt;&lt;/semantics&gt; &lt;/math&gt; 60 cm panels was also evaluated.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Results: &lt;/strong&gt;Sensitivity increases over threefold when panel-panel distance is reduced from 80 to 40 cm. The noise equivalent count rate, unmodified by TOF gain, of the panel detectors matches that of the reference clinical scanner at a distance of approximately 50 cm between the panels. Spatial resolution perpendicular to the panels improves from 8.7 to 1.6 mm when the panel-panel distance is reduced, and 70 ps + DOI detectors are used instead of 200 ps, no-DOI detectors. With enhanced TOF and DOI capabilities, panel detectors achieve image quality that matches or surpasses the reference scanner while using about four times less detector material. These detectors can be extended for LAFOV imaging without distortions or artifacts. Additionally, improving TOF and DOI performance enhances contrast-to-noise ratios, thereby improving lesion detection.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Conclusions: &lt;/strong&gt;A compact 2-panel PET scanner can match the performance of conventional scanners, producing high-quality, distortion-free images. Its mobility and flexibility enable no","PeriodicalId":94136,"journal":{"name":"Medical physics","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2025-03-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143639919","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":"Quantitative susceptibility mapping via deep neural networks with iterative reverse concatenations and recurrent modules.","authors":"Min Li, Chen Chen, Zhuang Xiong, Yin Liu, Pengfei Rong, Shanshan Shan, Feng Liu, Hongfu Sun, Yang Gao","doi":"10.1002/mp.17747","DOIUrl":"https://doi.org/10.1002/mp.17747","url":null,"abstract":"<p><strong>Background: </strong>Quantitative susceptibility mapping (QSM) is a post-processing magnetic resonance imaging (MRI) technique that extracts the distribution of tissue susceptibilities and holds significant promise in the study of neurological diseases. However, the ill-conditioned nature of dipole inversion often results in noise and artifacts during QSM reconstruction from the tissue field. Deep learning methods have shown great potential in addressing these issues; however, most existing approaches rely on basic U-net structures, leading to limited performances and reconstruction artifacts sometimes.</p><p><strong>Purpose: </strong>This study aims to develop a novel deep learning-based method, IR²QSM, for improving QSM reconstruction accuracy while mitigating noise and artifacts by leveraging a unique network architecture that enhances latent feature utilization.</p><p><strong>Methods: </strong>IR²QSM, an advanced U-net architecture featuring four iterations of reverse concatenations and middle recurrent modules, was proposed to optimize feature fusion and improve QSM accuracy, and comparative experiments based on both simulated and in vivo datasets were carried out to compare IR²QSM with two traditional iterative methods (iLSQR, MEDI) and four recently proposed deep learning methods (U-net, xQSM, LPCNN, and MoDL-QSM).</p><p><strong>Results: </strong>In this work, IR²QSM outperformed all other methods in reducing artifacts and noise in QSM images. It achieved on average the lowest XSIM (84.81%) in simulations, showing improvements of 12.80%, 12.68%, 18.66%, 10.49%, 25.57%, and 19.78% over iLSQR, MEDI, U-net, xQSM, LPCNN, and MoDL-QSM, respectively, and yielded results with the least artifacts on the in vivo data and present the most visually appealing results. In the meantime, it successfully alleviated the over-smoothing and susceptibility underestimation in LPCNN results.</p><p><strong>Conclusion: </strong>Overall, the proposed IR²QSM showed superior QSM results compared to iterative and deep learning-based methods, offering a more accurate QSM solution for clinical applications.</p>","PeriodicalId":94136,"journal":{"name":"Medical physics","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2025-03-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143639932","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":"An MR-only deep learning inference model-based dose estimation algorithm for MR-guided adaptive radiation therapy.","authors":"Zhiqiang Liu, Kuo Men, Weigang Hu, Jianrong Dai, Jiawei Fan","doi":"10.1002/mp.17759","DOIUrl":"https://doi.org/10.1002/mp.17759","url":null,"abstract":"<p><strong>Background: </strong>Magnetic resonance-guided adaptive radiation therapy (MRgART) systems combine Magnetic resonance imaging (MRI) technology with linear accelerators (LINAC) to enhance the precision and efficacy of cancer treatment. These systems enable real-time adjustments of treatment plans based on the latest patient anatomy, creating an urgent need for accurate and rapid dose calculation algorithms. Traditional CT-based dose calculations and ray-tracing (RT) processes are time-consuming and may not be feasible for the online adaptive workflow required in MRgART. Recent advancements in deep learning (DL) offer promising solutions to overcome these limitations.</p><p><strong>Purpose: </strong>This study aims to develop a DL-based dose calculation engine for MRgART that relies solely on MR images. This approach addresses the critical need for accurate and rapid dose calculations in the MRgART workflow without relying on CT images or time-consuming RT processes.</p><p><strong>Methods: </strong>We used a deep residual network inspired by U-Net to establish a direct connection between distance-corrected conical (DCC) fluence maps and dose distributions in the image domain. The study utilized data from 30 prostate cancer patients treated with fixed-beam Intensity-Modulated Radiation Therapy (IMRT) on an MR-guided LINAC system. We trained, validated, and tested the model using a total of 120 online treatment plans, which encompassed 1080 individual beams. We extensively evaluated the network's performance by comparing its dose calculation accuracy against Monte Carlo (MC)-based methods using metrics such as mean absolute error (MAE) of pixel-wise dose differences, 3D gamma analysis, dose-volume histograms (DVHs), dosimetric indices, and isodose line similarity.</p><p><strong>Results: </strong>The proposed DL model demonstrated high accuracy in dose calculations. The median MAE of pixel-wise dose differences was 1.2% for the whole body, 1.9% for targets, and 1.1% for organs at risk (OARs). The median 3D gamma passing rates for the 3%/3  mm criterion were 94.8% for the whole body, 95.7% for targets, and 98.7% for OARs. Additionally, the Dice similarity coefficient (DSC) of isodose lines between the DL-based and MC-based dose calculations averaged 0.94 ± 0.01. There were no big differences between the DL-based and MC-based calculations in the DVH curves and clinical dosimetric indices. This proved that the two methods were clinically equivalent.</p><p><strong>Conclusion: </strong>This study presents a novel MR-only dose calculation engine that eliminates the need for CT images and complex RT processes. By leveraging DL, the proposed method significantly enhances the efficiency and accuracy of the MRgART workflow, particularly for prostate cancer treatment. This approach holds potential for broader applications across different cancer types and MR-linac systems, paving the way for more streamlined and precise radiation therapy planning.</p>","PeriodicalId":94136,"journal":{"name":"Medical physics","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2025-03-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143639912","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":"Decentralized learning for medical image classification with prototypical contrastive network.","authors":"Zhantao Cao, Yuanbing Shi, Shuli Zhang, Huanan Chen, Weide Liu, Guanghui Yue, Huazhen Lin","doi":"10.1002/mp.17753","DOIUrl":"https://doi.org/10.1002/mp.17753","url":null,"abstract":"<p><strong>Background: </strong>Recently, deep convolutional neural networks (CNNs) have shown great potential in medical image classification tasks. However, the practical usage of the methods is constrained by two challenges: 1) the challenge of using nonindependent and identically distributed (non-IID) datasets from various medical institutions while ensuring privacy, and 2) the data imbalance problem due to the frequency of different diseases.</p><p><strong>Purpose: </strong>The objective of this paper is to present a novel approach for addressing these challenges through a decentralized learning method using a prototypical contrastive network to achieve precise medical image classification while mitigating the non-IID problem across different clients.</p><p><strong>Methods: </strong>We propose a prototype contrastive network that minimizes disparities among heterogeneous clients. This network utilizes an approximate global prototype to alleviate the non-IID dataset problem for each local client by projecting data onto a balanced prototype space. To validate the effectiveness of our algorithm, we employed three distinct datasets of color fundus photographs for diabetic retinopathy: the EyePACS, APTOS, and IDRiD datasets. During training, we incorporated 35k images from EyePACS, 3662 from APTOS, and 516 from IDRiD. For testing, we used 53k images from EyePACS. Additionally, we included the COVIDx dataset of chest X-rays for comparative analysis, comprising 29 986 training images and 400 test samples.</p><p><strong>Results: </strong>In this study, we conducted comprehensive comparisons with existing works using four medical image datasets. Specifically, on the EyePACS dataset under the balanced IID setting, our method outperformed the FedAvg baseline by 3.7% in accuracy. In the Dirichlet non-IID setting, which presents an extremely unbalanced distribution, our method showed a notable 6.6% enhancement in accuracy over FedAvg. Similarly, on the APTOS dataset, our method achieved a 3.7% improvement in accuracy over FedAvg under the balanced IID setting and a 5.0% improvement under the Dirichlet non-IID setting. Notably, on the DCC non-IID and COVID-19 datasets, our method established a new state-of-the-art across all evaluation metrics, including WAccuracy, WPrecision, WRecall, and WF-score.</p><p><strong>Conclusions: </strong>Our proposed prototypical contrastive loss guides the local client's data distribution to align with the global distribution. Additionally, our method uses an approximate global prototype to address unbalanced dataset distribution across local clients by projecting all data onto a new balanced prototype space. Our model achieves state-of-the-art performance on the EyePACS, APTOS, IDRiD, and COVIDx datasets.</p>","PeriodicalId":94136,"journal":{"name":"Medical physics","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2025-03-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143639916","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":"S²Net: Self-adaptive weighted fusion and self-adaptive aligned network for multi-modal MRI segmentation.","authors":"Chengzhi Gui, Xingwei An, Shuang Liu, Dong Ming","doi":"10.1002/mp.17742","DOIUrl":"https://doi.org/10.1002/mp.17742","url":null,"abstract":"<p><strong>Background: </strong>Accurate segmentation of lesions is beneficial for quantitative analysis and precision medicine in multimodal magnetic resonance imaging (MRI).</p><p><strong>Purpose: </strong>Currently, multimodal MRI fusion segmentation networks still face two main issues. On one hand, simple feature concatenation fails to fully capture the complex relationships between different modalities, as it overlooks the importance of dynamically changing feature weights across modalities. On the other hand, the unlearnable nature of upsampling in segmentation networks leads to feature misalignment issues during feature aggregation with the decoder, resulting in spatial misalignments between feature maps of different levels and ultimately pixel-level classification errors in predictions.</p><p><strong>Methods: </strong>This paper introduces the Self-adaptive weighted fusion and Self-adaptive aligned Network (S²Net), which comprises two key modules: the Self-Adaptive Weighted Fusion Module (SWFM) and the Self-Adaptive Aligned Module (SAM). S²Net can adaptively assign fusion weights based on the importance of different modalities and adaptively learn feature deformation fields to generate dynamic and flexible variability grids for feature alignment. This approach results in the generation of upsampled late-stage features with correct spatial locations and precise lesion boundaries.</p><p><strong>Results: </strong>This paper conducts experiments on two MRI datasets: ISLES 2022 and BraTS 2020. In the ISLES 2022 dataset, compared to the sub-optimal network MedNeXt, the proposed S²Net showed improvements of 3.52% in Dice Similarity Coefficient (DSC), 1.67% in Intersection over Union (IoU), and 4.7% in sensitivity, with a decrease of 0.33 mm in Hausdorff Distance 95 (HD95). In the BraTS 2020 dataset, compared to the sub-optimal network MedNeXt, the proposed S²Net achieved increases of 1.32% in mean DSC, 2.07% in mean IoU, and 2.17% in mean sensitivity, with a decrease of 0.10 mm in mean HD95. The code is open-sourced and available at: https://github.com/Cooper-Gu/S2Net.</p><p><strong>Conclusions: </strong>Experimental results demonstrate that S²Net exhibits superior segmentation performance in multimodal MRI segmentation compared to MedNeXt, FFNet, and ACMINet.</p>","PeriodicalId":94136,"journal":{"name":"Medical physics","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2025-03-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143639930","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 robotic treatment delivery system to facilitate dynamic conformal synchrotron radiotherapy.","authors":"Micah J Barnes, Nader Afshar, Taran Batty, Tom Fiala, Matthew Cameron, Daniel Hausermann, Nicholas Hardcastle, Michael Lerch","doi":"10.1002/mp.17750","DOIUrl":"https://doi.org/10.1002/mp.17750","url":null,"abstract":"&lt;p&gt;&lt;strong&gt;Background: &lt;/strong&gt;In clinical radiotherapy, the patient remains static during treatment and only the source is dynamically manipulated. In synchrotron radiotherapy, the beam is fixed, and is horizontally wide and vertically small, requiring the patient to be moved through the beam to ensure full target coverage, while shaping the field to conform to the target. No clinical system exists that performs both dynamic motion of the patient and dynamic shaping of the beam.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Purpose: &lt;/strong&gt;We developed and tested a new dynamic treatment delivery system capable of delivering conformal fields with a robotic patient positioning system for use on the Imaging and Medical Beamline (IMBL) at the Australian Nuclear Science and Technology Organisation, Australian Synchrotron.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Methods: &lt;/strong&gt;An industrial robotic manipulator was modified to enable dynamic radiotherapy treatments on IMBL. The robot, combined with a carbon-fiber treatment couch-top and a recently developed dynamic collimator, formed the basis of the new treatment delivery system. To synchronize the motions of the robot and collimator, a real-time, hardware-based event-handling system was utilized. To test the system, a ball bearing in a medical physics phantom was treated with circular fields ranging from 5 to 40 mm in diameter and at treatment speeds from 2 to 50 mm  &lt;math&gt; &lt;semantics&gt;&lt;msup&gt;&lt;mi&gt;s&lt;/mi&gt; &lt;mrow&gt;&lt;mo&gt;-&lt;/mo&gt; &lt;mn&gt;1&lt;/mn&gt;&lt;/mrow&gt; &lt;/msup&gt; &lt;annotation&gt;${rm s}^{-1}$&lt;/annotation&gt;&lt;/semantics&gt; &lt;/math&gt; . The position of the ball bearing was compared to the center of the circular fields and the positional and temporal accuracy of the treatment delivery system was assessed, and appropriate treatment margins for the system were determined.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Results: &lt;/strong&gt;The vertical position of the ball bearing varied with treatment delivery speed ( &lt;math&gt; &lt;semantics&gt;&lt;mrow&gt;&lt;mo&gt;-&lt;/mo&gt; &lt;mn&gt;1.06&lt;/mn&gt; &lt;mspace&gt;&lt;/mspace&gt; &lt;mi&gt;to&lt;/mi&gt; &lt;mspace&gt;&lt;/mspace&gt; &lt;mn&gt;0.93&lt;/mn&gt; &lt;mspace&gt;&lt;/mspace&gt; &lt;mi&gt;mm&lt;/mi&gt;&lt;/mrow&gt; &lt;annotation&gt;$-1.06 , {rm to}, 0.93 ,mathrm{mm}$&lt;/annotation&gt;&lt;/semantics&gt; &lt;/math&gt; ) while the horizontal position remained consistent ( &lt;math&gt; &lt;semantics&gt;&lt;mrow&gt;&lt;mo&gt;-&lt;/mo&gt; &lt;mn&gt;0.05&lt;/mn&gt; &lt;mspace&gt;&lt;/mspace&gt; &lt;mi&gt;to&lt;/mi&gt; &lt;mspace&gt;&lt;/mspace&gt; &lt;mn&gt;0.09&lt;/mn&gt; &lt;mspace&gt;&lt;/mspace&gt; &lt;mi&gt;mm&lt;/mi&gt;&lt;/mrow&gt; &lt;annotation&gt;$-0.05 ,{rm to}, 0.09 ,mathrm{mm}$&lt;/annotation&gt;&lt;/semantics&gt; &lt;/math&gt; ). The time-delay between the robot and the collimator remained consistent ( &lt;math&gt; &lt;semantics&gt;&lt;mrow&gt;&lt;mo&gt;-&lt;/mo&gt; &lt;mn&gt;35.5&lt;/mn&gt; &lt;mspace&gt;&lt;/mspace&gt; &lt;mi&gt;ms&lt;/mi&gt; &lt;mspace&gt;&lt;/mspace&gt; &lt;mi&gt;to&lt;/mi&gt; &lt;mspace&gt;&lt;/mspace&gt; &lt;mn&gt;18.5&lt;/mn&gt; &lt;mspace&gt;&lt;/mspace&gt; &lt;mi&gt;ms&lt;/mi&gt;&lt;/mrow&gt; &lt;annotation&gt;$-35.5 ,mathrm{ms},{rm to}, 18.5 ,mathrm{ms}$&lt;/annotation&gt;&lt;/semantics&gt; &lt;/math&gt; ) at treatment speeds above &lt;math&gt; &lt;semantics&gt;&lt;mrow&gt;&lt;mn&gt;2&lt;/mn&gt; &lt;mspace&gt;&lt;/mspace&gt; &lt;msup&gt;&lt;mi&gt;mms&lt;/mi&gt; &lt;mrow&gt;&lt;mo&gt;-&lt;/mo&gt; &lt;mn&gt;1&lt;/mn&gt;&lt;/mrow&gt; &lt;/msup&gt; &lt;/mrow&gt; &lt;annotation&gt;$2 ,mathrm{mm s}^{-1}$&lt;/annotation&gt;&lt;/semantics&gt; &lt;/math&gt; . Data at &lt;math&gt; &lt;semantics&gt;&lt;mrow&gt;&lt;mn&gt;2&lt;/mn&gt; &lt;mspa","PeriodicalId":94136,"journal":{"name":"Medical physics","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2025-03-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143639911","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":"Stable and discriminating OCT-derived radiomics features for predicting anti-VEGF treatment response in diabetic macular edema.","authors":"Sudeshna Sil Kar, Hasan Cetin, Sunil K Srivastava, Anant Madabhushi, Justis P Ehlers","doi":"10.1002/mp.17695","DOIUrl":"10.1002/mp.17695","url":null,"abstract":"&lt;p&gt;&lt;strong&gt;Background: &lt;/strong&gt;Radiomics-based characterization of fluid and retinal tissue compartments of spectral-domain optical coherence tomography (SD-OCT) scans has shown promise to predict anti-VEGF therapy treatment response in diabetic macular edema (DME). Radiomics features are sensitive to different image acquisition parameters of OCT scanners such as axial resolution, A-scan rate, and voxel size; consequently, the predictive capability of the radiomics features might be impacted by inter-site and inter-scanner variations.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Purpose: &lt;/strong&gt;The main objective of this study was (1) to develop a more generalized classifier by identifying the OCT-derived texture-based radiomics features that are both stable (across multiple scanners) as well as discriminative of therapeutic response in DME and (2) to identify the relative stability of individual radiomic features that are associated with specific spatial compartments (e/g. fluid or tissue) within the eye.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Methods: &lt;/strong&gt;A combination of 151 optimal responders and rebounders of anti-VEGF therapy in DME were included from the PERMEATE (imaged using Cirrus HD-OCT scanner) and VISTA clinical trials (imaged using Cirrus HD-OCT and Spectralis scanners). For each patient within the study, a set of 494 texture-based radiomics features were extracted from the fluid and the retinal tissue compartment of OCT images. The training set ( &lt;math&gt; &lt;semantics&gt;&lt;msub&gt;&lt;mi&gt;S&lt;/mi&gt; &lt;mi&gt;t&lt;/mi&gt;&lt;/msub&gt; &lt;annotation&gt;${{S}_t}$&lt;/annotation&gt;&lt;/semantics&gt; &lt;/math&gt; ) included 76 patients and the independent test set &lt;math&gt; &lt;semantics&gt;&lt;mrow&gt;&lt;mo&gt;(&lt;/mo&gt; &lt;msub&gt;&lt;mi&gt;S&lt;/mi&gt; &lt;mi&gt;v&lt;/mi&gt;&lt;/msub&gt; &lt;/mrow&gt; &lt;annotation&gt;$({{S}_v}$&lt;/annotation&gt;&lt;/semantics&gt; &lt;/math&gt; ) comprised of 75 patients. Features were ranked based on (i) only discriminability criteria, that is, maximizing area under the receiver operating characteristic curve (AUC) and (ii) both stability and discriminability criteria. The subset of radiomic features for which the feature expression remained relatively consistent between the two datasets, as assessed by Wilcoxon rank-sum test, were considered to be stable. Different machine learning (ML) classifiers (such as k-nearest neighbors, Random Forest, Linear Discriminant Analysis, Quadratic Discriminant Analysis, Support Vector Machine using linear and radial basis kernel, Naive Bayes) were trained using the features selected based on both the stability and discriminability criteria on &lt;math&gt; &lt;semantics&gt;&lt;msub&gt;&lt;mi&gt;S&lt;/mi&gt; &lt;mi&gt;t&lt;/mi&gt;&lt;/msub&gt; &lt;annotation&gt;${{S}_t}$&lt;/annotation&gt;&lt;/semantics&gt; &lt;/math&gt; and then subsequently validated on &lt;math&gt; &lt;semantics&gt;&lt;msub&gt;&lt;mi&gt;S&lt;/mi&gt; &lt;mi&gt;v&lt;/mi&gt;&lt;/msub&gt; &lt;annotation&gt;${{S}_v}$&lt;/annotation&gt;&lt;/semantics&gt; &lt;/math&gt; . The ML classifier ( &lt;math&gt; &lt;semantics&gt;&lt;msub&gt;&lt;mi&gt;M&lt;/mi&gt; &lt;mi&gt;g&lt;/mi&gt;&lt;/msub&gt; &lt;annotation&gt;${{M}_g}$&lt;/annotation&gt;&lt;/semantics&gt; &lt;/math&gt; ) that yielded maximum AUC on &lt;math&gt; &lt;semantics&gt;&lt;msub&gt;&lt;mi&gt;S&lt;/mi&gt; &lt;mi&gt;v&lt;/mi&gt;&lt;/msub&gt; &lt;annotation&gt;${{S}_v}$&lt;/annotation&gt;&lt;/semantics&gt; &lt;/math&gt; ","PeriodicalId":94136,"journal":{"name":"Medical physics","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2025-03-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143627343","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}