Physics in medicine and biology最新文献

{"title":"Automated quantification of cervical spine degeneration with development of a segmentation framework based on probabilistic anatomical cognition.","authors":"Jinge Wang, Siyuan Qin, Ruomu Qu, Feifei Zhou, Ning Lang, Xuefeng Wang","doi":"10.1088/1361-6560/ae023a","DOIUrl":"10.1088/1361-6560/ae023a","url":null,"abstract":"<p><p><i>Objective.</i>Ossification of the posterior longitudinal ligament (OPLL) is a prevalent cervical spine degeneration disease leading to significant spinal cord dysfunctions. Due to morphological diversity and data scarcity, traditional OPLL assessment relies on manual measurements, which suffer from low consistency and high cost. To implement automated quantification of the OPLL, a cognition-inspired segmentation framework, named the probabilistic anatomical cognition (PAC) framework, is proposed to encode physicians' anatomical knowledge of the OPLL and mimic their hierarchical logic of inferring lesions.<i>Approach.</i>The OPLL anatomical structure is firstly modeled by a multi-level probabilistic representation from the stochastic global shape of the spinal canal (SC) to the local feature distributions of the lesions. Based on the anatomical prior model, the OPLL segmentation is implemented by the deep-logic shape inference. The logic extracts high-confidence global feature observations of the SC, following with the inference to the local lesions by morphological correlations. The fusion of the anatomical prior and multi-level observations enhances both interpretability and generalization of lesion segmentation and reduces reliance on large datasets.<i>Main results.</i>Tested on a clinical dataset of 439 patients, the PAC framework improves dice similarity coefficient by 10% over the lightweight baseline and achieves high consistency with expert assessments on clinical lesion metrics.<i>Significance.</i>A general automated segmentation pipeline and three-dimensional metrics are provided for the first time by the framework to quantify the OPLL degeneration, which offers valuable insights to support surgical decision-making.</p>","PeriodicalId":20185,"journal":{"name":"Physics in medicine and biology","volume":" ","pages":""},"PeriodicalIF":3.4,"publicationDate":"2025-09-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144965091","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Adaptive k-sparse constrained dictionary learning strategy for bioluminescence tomography reconstruction.","authors":"Bianbian Yang, Yiting He, Nannan Cai, Yi Chen, Huangjian Yi, Xingxing Hao, Chengyi Gao, Xin Cao","doi":"10.1088/1361-6560/ae0c51","DOIUrl":"https://doi.org/10.1088/1361-6560/ae0c51","url":null,"abstract":"<p><strong>Objective: </strong>Bioluminescence tomography (BLT) is a significant molecular imaging modality with promising potential in biomedical research. However, the reconstruction results of BLT are frequently sensitive and imprecise due to the light scattering effect and ill-posed inverse problem.</p><p><strong>Approach: </strong>We propose an accelerated forward-backward splitting and the difference of convex functions algorithm (AFBS-DCA) based on a dictionary learning framework. In the sparse coding phase, a k-sparsity strategy enables adaptive adjustment of the regularization parameter, improving the overall efficiency. The non-convex generalized minimax-concave (GMC) regularization is employed to enhance sparsity, while Nesterov's acceleration strategy improves convergence speed. During dictionary updating, DCA is utilized to efficiently solve a non-convex optimization problem modelled as a difference of two convex functions, effectively reducing computational complexity.</p><p><strong>Main results: </strong>The effectiveness of the AFBS-DCA method was evaluated through numerical simulations and light source implantation experiments. It achieved the highest reconstruction accuracy with an average localization error (LE) of 0.391 mm, an average Dice coefficient (DICE) of 0.774, and a contrast-to-noise ratio (CNR) of 0.872. Compared with three baseline methods, the AFBS-DCA reduced reconstruction errors by 62.8%, 52.5%, and 37.8%, respectively.</p><p><strong>Significance: </strong>The proposed AFBS-DCA method demonstrates superior performance in terms of localization accuracy, morphological recovery, and robustness, indicating its potential to advance the practical application of BLT in biomedical research and molecular imaging.</p>","PeriodicalId":20185,"journal":{"name":"Physics in medicine and biology","volume":" ","pages":""},"PeriodicalIF":3.4,"publicationDate":"2025-09-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145177658","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Microdosimetric characterization of a clinical helium beam using a MicroPlus-Bridge detector and a diamond detector for RBE assessment.","authors":"G Petringa, C Verona, A Attili, L Brighel, R Catalano, G A P Cirrone, V C Elia, F Fede, E Formicola, M Guarrera, Y Hamad, A Kurmanova, A Mairani, T Tessonnier, L Manti","doi":"10.1088/1361-6560/ae0862","DOIUrl":"10.1088/1361-6560/ae0862","url":null,"abstract":"<p><p><i>Objective</i>. This study aims to perform a comprehensive microdosimetric characterization of a clinical helium ion beam at the Heidelberg ion-beam Therapy center, assessing radiation quality along a spread-out Bragg peak (SOBP) and providing experimental data to support the biological effectiveness of helium ions in clinical applications.<i>Approach</i>. Microdosimetric spectra were measured at several water depths within the SOBP using two solid-state detectors: a silicon and a synthetic single-crystal diamond detector. Microdosimetric quantities, including dose-mean lineal energy (yD) and frequency-mean lineal energy (yF), were derived. Monte Carlo simulations with the Geant4 toolkit replicated the experimental setup to validate the measured quantities. The modified microdosimetric kinetic model was also applied to estimate the relative biological effectiveness at 10% survival (RBE₁₀). SAOS-2 and U2OS clonogenic assays were performed under identical irradiation conditions at the same facility to benchmark the RBE predictions.<i>Main results</i>. The experimental results demonstrated a progressive increase inyDvalues along the beam path, peaking at the distal edge of the SOBP, consistent with the simulated dose-averaged linear energy transfer distribution. The RBE₁₀values estimated by the two solid-state detectors showed good agreement with the experimental data from clonogenic assays, within the associated uncertainties.<i>Significance</i>. These findings underscore the utility of advanced solid-state detectors for helium beam characterization and highlight the importance of microdosimetry in improving radiobiological modeling. The results provide further evidence supporting the clinical potential of helium ion therapy, particularly for tumors requiring high precision and maximal healthy tissue sparing.</p>","PeriodicalId":20185,"journal":{"name":"Physics in medicine and biology","volume":" ","pages":""},"PeriodicalIF":3.4,"publicationDate":"2025-09-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145080359","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Enhanced SVD filter based on singular vector subspace denoising improves ultrafast ultrasound microvascular imaging performance.","authors":"Yu Xia, Jiabin Zhang, Daichao Chen, Jingyi Yin, Hao Yu, Jue Zhang","doi":"10.1088/1361-6560/ae085f","DOIUrl":"10.1088/1361-6560/ae085f","url":null,"abstract":"<p><p><i>Objective</i>. Ultrafast ultrasound imaging can significantly improve the ability of ultrasound for microvascular visualization. Clutter filtering through singular value decomposition (SVD)-based filter remains a pivotal step in Ultrafast ultrasound imaging. However, the current hard threshold-based SVD filter cannot completely separate blood flow from noise on the basis of filtering tissue clutter, resulting in low contrast in microvascular imaging. This paper proposes a novel enhanced SVD (eSVD) filter to enhance blood flow signals and suppress noise while filtering clutter.<i>Approach</i>. The proposed method innovatively partitions spatial singular vectors into multiple blood flow subspaces followed by subspace-specific weighted reconstruction to amplify blood signatures.<i>Main results</i>. We validate the effectiveness of the eSVD filter in contrast-free ultrafast power Doppler imaging (uPDI), contrast-enhanced uPDI, and ultrasound localization microscopy (ULM) imaging experiments. Qualitative and quantitative experimental results show that compared with the hard threshold-based SVD filter, our method can significantly improve the contrast between vessels and background, and highlight the details of microvessels. Compared with the adaptive SVD filter based on the spatial similarity matrix, our eSVD filter improves contrast-to-noise ratio by 8.36 dB, signal-to-noise ratio by 7.92 dB, and blood-to-clutter ratio by 15.47 dB in the uPDI of mouse contrast-free brain. In the ULM of mouse tumor, our eSVD filter improves the global spatial resolution by about 6 <i>µ</i>m, from 34.49 <i>µ</i>m to 28.15 <i>µ</i>m.<i>Significance</i>. The proposed eSVD filter essentially improves the performance of ultrafast ultrasound microvascular imaging and has the potential for the diagnosis of many diseases related to microvessel change.</p>","PeriodicalId":20185,"journal":{"name":"Physics in medicine and biology","volume":" ","pages":""},"PeriodicalIF":3.4,"publicationDate":"2025-09-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145081386","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Deep learning-driven contactless ECG in MRI via beat pilot tone for motion-resolved image reconstruction and heart rate monitoring.","authors":"Haoyu Sun, Qichen Ding, Sijie Zhong, Zhiyong Zhang","doi":"10.1088/1361-6560/ae0c52","DOIUrl":"https://doi.org/10.1088/1361-6560/ae0c52","url":null,"abstract":"<p><strong>Objective: </strong>Electrocardiogram (ECG) is crucial for synchronizing cardiovascular magnetic resonance imaging (CMRI) acquisition with the cardiac cycle and for continuous heart rate monitoring during prolonged scans. However, conventional electrode-based ECG systems in clinical MRI environments suffer from tedious setup, magnetohydrodynamic (MHD) waveform distortion, skin burn risks, and patient discomfort. This study proposes a contactless ECG measurement method in MRI to address these challenges.</p><p><strong>Approach: </strong>We integrated Beat Pilot Tone (BPT)-a contactless, high motion sensitivity, and easily integrable RF motion sensing modality-into CMRI to capture cardiac motion without direct patient contact. A deep neural network was trained to map the BPT-derived cardiac mechanical motion signals to corresponding ECG waveforms. The reconstructed ECG was evaluated against simultaneously acquired ground truth ECG through multiple metrics: Pearson correlation coefficient, relative root mean square error (RRMSE), cardiac trigger timing accuracy, and heart rate estimation error. Additionally, we performed MRI retrospective binning reconstruction using reconstructed ECG reference and evaluated image quality under both standard clinical conditions and challenging scenarios involving arrhythmias and subject motion. To examine scalability of our approach across field strength, the model pretrained on 1.5T data was applied to 3T BPT cardiac acquisitions.</p><p><strong>Main results: </strong>In optimal acquisition scenarios, the reconstructed ECG achieved a median Pearson correlation of 89% relative to the ground truth, while cardiac triggering accuracy reached 94%, and heart rate estimation error remained below 1 bpm. The quality of the reconstructed images was comparable to that of ground truth synchronization. The method exhibited a degree of adaptability to irregular heart rate patterns and subject motion, and scaled effectively across MRI systems operating at different field strengths.</p><p><strong>Significance: </strong>The proposed contactless ECG measurement method has the potential to streamline CMRI workflows, improve patient safety and comfort, mitigate MHD distortion challenges and find a robust clinical application.</p>","PeriodicalId":20185,"journal":{"name":"Physics in medicine and biology","volume":" ","pages":""},"PeriodicalIF":3.4,"publicationDate":"2025-09-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145177663","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Fixed point method for PET reconstruction with learned plug-and-play regularization.","authors":"Marion Savanier, Claude Comtat, Florent Sureau","doi":"10.1088/1361-6560/ae05ac","DOIUrl":"10.1088/1361-6560/ae05ac","url":null,"abstract":"<p><p><i>Objective.</i>Deep learning has shown great promise for improving medical image reconstruction, including positron emission tomography (PET). However, concerns remain about the stability and robustness of these methods, especially when trained on limited data. This work aims to explore the use of the Plug-and-Play (PnP) framework in PET reconstruction to address these concerns.<i>Approach.</i>We propose a convergent PnP algorithm for low-count PET reconstruction based on the Douglas-Rachford splitting method. We consider several denoisers trained to satisfy fixed-point conditions, with convergence properties ensured either during training or by design, including a spectrally normalized network and a deep equilibrium model. We evaluate the bias-standard deviation tradeoff across clinically relevant regions and an unseen pathological case in a synthetic experiment and a real study. Comparisons are made with model-based iterative reconstruction, post-reconstruction denoising, a deep end-to-end unfolded network and PnP with a Gaussian denoiser.<i>Main results.</i>Our method achieves lower bias than post-reconstruction processing and reduced standard deviation at matched bias compared to model-based iterative reconstruction. While spectral normalization underperforms in generalization, the deep equilibrium model remains competitive with convolutional networks for PnP reconstruction and generalizes better to the unseen pathology. Compared to the end-to-end unfolded network, it also generalizes more consistently.<i>Significance.</i>This study demonstrates the potential of the PnP framework to improve image quality and quantification accuracy in PET reconstruction. It also highlights the importance of how convergence conditions are imposed on the denoising network to ensure robust and generalizable performance.</p>","PeriodicalId":20185,"journal":{"name":"Physics in medicine and biology","volume":" ","pages":""},"PeriodicalIF":3.4,"publicationDate":"2025-09-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145034088","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Derivation and properties of the convolution model for MRI gradient-induced cardiac stimulation.","authors":"Seung-Kyun Lee, Timothy P Eagan, Desmond Teck Beng Yeo","doi":"10.1088/1361-6560/ae06ec","DOIUrl":"10.1088/1361-6560/ae06ec","url":null,"abstract":"<p><p><i>Objective.</i>Reliable prediction of gradient-induced peripheral nerve stimulation (PNS) and cardiac stimulation (CS) is important to ensure patient safety and maximize imaging performance in modern MRI scanners. Here we extend the dynamic convolution-based PNS prediction model to CS, and present theoretical analysis and numerical survey of general properties of the convolution model.<i>Approach.</i>CS convolution kernel was derived from the exponential model of the strength-duration curve of excitable tissue stimulation with representative stimulation parameters for a whole-body gradient coil. Self-consistency of the convolution method and the properties of the convolution output (response function) for a periodic trapezoidal wave were theoretically analyzed. PNS and CS response functions were computed for clinical 3T brain and pelvic imaging sequences for comparison.<i>Main results.</i>CS convolution kernel takes the form of a simple, decaying exponential function. For both PNS and CS kernels, the convolution model is consistent with the strength-duration curve when applied to a rectangular d<i>G</i>/d<i>t</i>pulse. The long time constant of a CS kernel tends to suppress stimulation by short d<i>G</i>/d<i>t</i>pulses, and makes dynamic CS response correlate more with gradient amplitude than slew rate. On a trapezoidal gradient pulse train, the maximum PNS or CS occurs at the end of the first full slope of the waveform, independent of the number of cycles. In light of the available evidence to the contrary, such independence indicates limitation of the convolution model which is strictly linear.<i>Significance.</i>The proposed CS convolution model can supplement existing PNS models to better assess patient safety of arbitrary gradient waveforms. General theoretical properties of the convolution model can help guide waveform design to minimize risks. While our method was demonstrated primarily on whole-body gradient systems, it can also inform PNS and CS prediction for anatomy-specific scanners employing fast and strong gradient fields.</p>","PeriodicalId":20185,"journal":{"name":"Physics in medicine and biology","volume":" ","pages":""},"PeriodicalIF":3.4,"publicationDate":"2025-09-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145070238","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Evaluation of nanodosimetric quantities for ion radiotherapy treatment planning based on the degree of association of survival with cluster dose.","authors":"Ramon Ortiz, José Ramos-Méndez, Jian-Hua Mao, Reinhard Schulte, Bruce Faddegon","doi":"10.1088/1361-6560/ae07a3","DOIUrl":"10.1088/1361-6560/ae07a3","url":null,"abstract":"<p><p><i>Objective.</i>To demonstrate potential for a close association between cell survival and cluster dose for ionization parameters (<i>I</i>_p), and to investigate the means to quantify the degree of this association when calculating cluster dose using these nanodosimetric quantities.<i>Approach.</i>The definitions of<i>I</i>_pconsidered were the number of clusters of<i>k</i>or more ionizations per unit track length {<i>C_k, k</i>= 1,…10}. For this<i>I</i>_pdefinition, cluster dose is the number of clusters of<i>k</i>or more ionizations per unit mass. Three sets of published cell survival data, covering a range of clinically relevant particle types and energies, normal and tumor human cells, and aerobic and hypoxic conditions, were used to assess these<i>I</i>_p. Values of<i>C_k</i>were previously calculated for this survival data and evaluated for their application in treatment planning. New to this study, the dependence of cell survival on cluster dose, calculated as local fluence times the mean mass<i>I</i>_p, was used. The degree of association of cell survival with cluster dose was quantified using three statistical methods: the moving window method, the residuals of linear quadratic fit, and the Bayesian information criteria.<i>Results.</i>all three methods identified<i>C</i>₅as the most closely associated with cell survival under aerobic conditions, and<i>C</i>₇under hypoxic conditions, consistent with visual observations. Remarkably,<i>C_k</i>preferred for their close association with cell survival for different particle types having the same fluence, compared to alternative definitions, resulted in a statistically significant closer association of cell survival with cluster dose, regardless of particle fluence.<i>Significance.</i>fluence is a critical property Cluster dose has the potential of supplementing or even replacing RBE-weighted dose in optimization of ion therapy treatment plans. The proposed methodology lays the groundwork for rigorous identification of<i>I</i>_pthat exhibit the highest degree of association of cell survival with cluster dose, a trait that greatly enhances the potential clinical impact of cluster dose.</p>","PeriodicalId":20185,"journal":{"name":"Physics in medicine and biology","volume":" ","pages":""},"PeriodicalIF":3.4,"publicationDate":"2025-09-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12461466/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145075960","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Experimental characterization of a diamond detection system for combined dose, LET and RBE assessment in clinical proton beams.","authors":"Claudio Verona, Andrea Fabbri, Alberto Fazzi, Lucrezia Bianchi, Anna Bianchi, Pablo Cirrone, Valeria Conte, Giada Petringa, Angelo Maria Raso, Emanuele Scifoni, Anna Selva, Francesco Tommasino, Gianluca Verona Rinati, Enrico Verroi","doi":"10.1088/1361-6560/ae0be8","DOIUrl":"https://doi.org/10.1088/1361-6560/ae0be8","url":null,"abstract":"<p><p><i>Objective:</i>This work presents the first experimental characterization of the DIODE detector, a novel detection system based on single-crystal diamond, designed for simultaneous dosimetric and microdosimetric measurements in clinical proton therapy. The aim is to evaluate its capability to measure absorbed dose, and estimate LET and RBE variations along a clinical proton beam.&#xD;<i>Approach:</i>The detector was tested under a 70 MeV monoenergetic proton beam at the Trento Proton Therapy Centre. Depth-dose and lateral profiles were compared with EBT3 radiochromic films. Microdosimetric spectra were simultaneously acquired and benchmarked against a mini tissue-equivalent proportional counter (TEPC). Monte Carlo (MC) simulations were also performed to model the experimental setup and the DIODE geometry.&#xD;<i>Main results:</i>The detector showed linearity with dose with a sensitivity of 0.60 ± 0.01 nC/Gy, with a dark current below 0.1 pA, ensuring a good signal-to-noise ratio. Depth-dose profiles matched EBT3 film data and MC simulations, with differences below 2% in peak-to-plateau ratios, indicating limited LET dependence.&#xD;The dose-mean lineal energy y_Dclosely agrees with simulated dose-averaged LET values, except in the beam entrance region. In this area, y_Dvalues were lower than those obtained with the mini-TEPC due to electronic saturation limiting the detection of rare high-LET secondary fragments.&#xD;Variation in RBE was assessed from the microdosimetric data, based on Loncol's weighting function, which refers to clonogenic cell survival. RBE values ranged from ~1.1 at the entrance to ~1.8 in the distal region, consistent with mini-TEPC data and literature RBE₁₀ values for V79 and U89 cells.&#xD;Lateral dose and microdosimetric profiles confirmed high spatial resolution and revealed proton energy variations near the Bragg peak.&#xD;<i>Significance:</i>The DIODE detector demonstrated reliable performance for simultaneous dosimetric and microdosimetric characterization of clinical proton beams. Its ability to measure dose, and to support LET and RBE assessment through microdosimetric modelling, in a single device, highlights its potential as an advanced tool for beam quality assessment and biologically optimized treatment planning in proton therapy.</p>","PeriodicalId":20185,"journal":{"name":"Physics in medicine and biology","volume":" ","pages":""},"PeriodicalIF":3.4,"publicationDate":"2025-09-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145150455","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"MedFormer: hierarchical medical vision transformer with content-aware dual sparse selection attention.","authors":"Zunhui Xia, Hongxing Li, Libin Lan","doi":"10.1088/1361-6560/ae07a1","DOIUrl":"10.1088/1361-6560/ae07a1","url":null,"abstract":"<p><p><i>Objective</i>. Medical image recognition serves as a key way to aid in clinical diagnosis, enabling more accurate and timely identification of diseases and abnormalities. Vision transformer-based approaches have proven effective in handling various medical recognition tasks. However, these methods encounter two primary challenges. First, they are often task-specific and architecture-tailored, limiting their general applicability. Second, they usually either adopt full attention to model long-range dependencies, resulting in high computational costs, or rely on handcrafted sparse attention, potentially leading to suboptimal performance. To tackle these issues, we present MedFormer, an efficient medical vision transformer with two key ideas.<i>Approach</i>. First, it employs a pyramid scaling structure as a versatile backbone for various medical image recognition tasks, including image classification and dense prediction tasks such as semantic segmentation and lesion detection. This structure facilitates hierarchical feature representation while reducing the computation load of feature maps, highly beneficial for boosting performance. Second, it introduces a novel Dual Sparse Selection Attention (DSSA) with content awareness to improve computational efficiency and robustness against noise while maintaining high performance. As the core building technique of MedFormer, DSSA is designed to explicitly attend to the most relevant content. Theoretical analysis demonstrates that MedFormer outperforms existing medical vision transformers in terms of generality and efficiency.<i>Main results</i>. Extensive experiments across various imaging modality datasets show that MedFormer consistently enhances performance in all three medical image recognition tasks mentioned above.<i>Significance</i>. MedFormer provides an efficient and versatile solution for medical image recognition, with strong potential for clinical application. The code is available onGitHub.</p>","PeriodicalId":20185,"journal":{"name":"Physics in medicine and biology","volume":" ","pages":""},"PeriodicalIF":3.4,"publicationDate":"2025-09-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145075496","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}