Pub Date : 2025-10-14DOI: 10.1016/j.mri.2025.110536
Junde Zhou , Qin Wang , Yanting Liu , Lu Zhang , Jiao Li , Shuo Li , Dong Liu , Jinxia Zhu , Robert Grimm , Alto Stemmer , Shuang Xia , Wenyang Huang , Sheng Xie , Haibo Zhang , Jian Li , Huadan Xue , Zhengyu Jin
Purpose
This study aimed to explore the utility of total tumor apparent diffusion coefficient (ttADC) histogram analysis based on whole-body diffusion-weighted magnetic resonance imaging (DWI-MRI) for prognostic stratification in patients with Revised International Staging System stage II (R-ISS II) multiple myeloma (MM).
Methods
Patients with R-ISS II MM who underwent baseline whole-body MRI prior to treatment were retrospectively enrolled. The ttADC histogram parameters of the whole-body DWI-MRI were obtained using MR Total Tumor Load software (Siemens Healthcare, Erlangen, Germany). The overall survival (OS) and the progression-free survival (PFS) of the cohort was recorded. Cox regression analyses were used to evaluate the clinical features and ttADC histogram parameters for their association with OS and PFS.
Results
A total of 61 R-ISS II MM patients were retrospectively included. During a mean follow-up period of 80 months, 27 patients died, 4 patients lost follow-up for OS, and 4 patients lost follow-up for PFS. Multivariate analysis revealed that increased median ttADC (≥0.620 × 10−3 mm2/s) [hazard ratio (HR) = 2.291, P = 0.046, 95 % confidence interval (CI): 1.014–5.175] was independently associated with OS in R-ISS II MM patients. No significant association was observed between the ttADC histogram parameters and PFS.
Conclusion
The median ttADC of whole-body DWI-MRI is an independent predictor of OS in R-ISS II MM patients, suggesting its potential role in further prognostic stratification in this patients' subgroup.
{"title":"Total tumor apparent diffusion coefficient histogram analysis of whole-body DWI-MRI for prognostic stratification in patients with R-ISS stage II multiple myeloma","authors":"Junde Zhou , Qin Wang , Yanting Liu , Lu Zhang , Jiao Li , Shuo Li , Dong Liu , Jinxia Zhu , Robert Grimm , Alto Stemmer , Shuang Xia , Wenyang Huang , Sheng Xie , Haibo Zhang , Jian Li , Huadan Xue , Zhengyu Jin","doi":"10.1016/j.mri.2025.110536","DOIUrl":"10.1016/j.mri.2025.110536","url":null,"abstract":"<div><h3>Purpose</h3><div>This study aimed to explore the utility of total tumor apparent diffusion coefficient (ttADC) histogram analysis based on whole-body diffusion-weighted magnetic resonance imaging (DWI-MRI) for prognostic stratification in patients with Revised International Staging System stage II (R-ISS II) multiple myeloma (MM).</div></div><div><h3>Methods</h3><div>Patients with R-ISS II MM who underwent baseline whole-body MRI prior to treatment were retrospectively enrolled. The ttADC histogram parameters of the whole-body DWI-MRI were obtained using MR Total Tumor Load software (Siemens Healthcare, Erlangen, Germany). The overall survival (OS) and the progression-free survival (PFS) of the cohort was recorded. Cox regression analyses were used to evaluate the clinical features and ttADC histogram parameters for their association with OS and PFS.</div></div><div><h3>Results</h3><div>A total of 61 R-ISS II MM patients were retrospectively included. During a mean follow-up period of 80 months, 27 patients died, 4 patients lost follow-up for OS, and 4 patients lost follow-up for PFS. Multivariate analysis revealed that increased median ttADC (≥0.620 × 10<sup>−3</sup> mm<sup>2</sup>/s) [hazard ratio (HR) = 2.291, <em>P</em> = 0.046, 95 % confidence interval (CI): 1.014–5.175] was independently associated with OS in R-ISS II MM patients. No significant association was observed between the ttADC histogram parameters and PFS.</div></div><div><h3>Conclusion</h3><div>The median ttADC of whole-body DWI-MRI is an independent predictor of OS in R-ISS II MM patients, suggesting its potential role in further prognostic stratification in this patients' subgroup.</div></div>","PeriodicalId":18165,"journal":{"name":"Magnetic resonance imaging","volume":"125 ","pages":"Article 110536"},"PeriodicalIF":2.0,"publicationDate":"2025-10-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145308715","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-10-10DOI: 10.1016/j.mri.2025.110534
Ruchuan Chen , Guoqing Hu , Bingni Zhou , Hualei Gan , Xiaofeng Liu , Lin Deng , Liangping Zhou , Yajia Gu , Xiaohang Liu
Purpose
To develop Homologous Recombination Repair (HRR) Genes mutations prediction models for prostate cancer using MRI radiomics and clinicopathological features.
Methods
Totally 353 prostate cancer patients (102 with HRR genes mutations) from three centers (center 1: training and internal test cohorts, center 2 and 3: external test cohorts) underwent multiparametric MRI. Each patient's index tumor lesion was delineated on T2-weighted imaging (T2WI), dynamic contrast enhancement (DCE) MRI, diffusion-weighted imaging (DWI), and apparent diffusion coefficient (ADC) images to obtain 428 radiomics features. Features associated with mutations were selected from clinicopathological features using Mann-Whitney U and Logistic regression (LR) test, radiomics features using Least Absolute Shrinkage and Selection Operator. Clinicopathological model was constructed with selected clinicopathological features. Logistic regression (LR), support vector machine (SVM) and linear discriminant analysis (LDA) classifiers were used to construct Radiomics and combined clinicopathological-radiomics models. Predictive efficiencies of models were compared using areas under the receiver operating characteristic curve (AUC).
Results
One clinicopathological and six radiomics features were selected. Radiomics with SVM, LR, LDA and Clinicopathological models achieved AUCs of 0.76, 0.76, 0.76, 0.68 and 0.75, 0.76, 0.67, 0.73 in internal and external test cohort. AUCs of combined clinicopathological-radiomics models with LDA in internal and external test cohort (0.83 and 0.82) were slightly higher than combined models with LR (0.81 and 0.79) and SVM (both 0.80) (P > 0.05), but were significantly higher than radiomics and clinicopathological models in both cohorts (P < 0.05).
Conclusion
LDA classifier incorporating radiomics and clinicopathological features predicting could effectively predict HRR genes mutations in prostate cancer.
{"title":"Combination of clinicopathological and MRI based radiomics features in predicting homologous recombination repair genes mutations in prostate cancer","authors":"Ruchuan Chen , Guoqing Hu , Bingni Zhou , Hualei Gan , Xiaofeng Liu , Lin Deng , Liangping Zhou , Yajia Gu , Xiaohang Liu","doi":"10.1016/j.mri.2025.110534","DOIUrl":"10.1016/j.mri.2025.110534","url":null,"abstract":"<div><h3>Purpose</h3><div>To develop Homologous Recombination Repair (HRR) Genes mutations prediction models for prostate cancer using MRI radiomics and clinicopathological features.</div></div><div><h3>Methods</h3><div>Totally 353 prostate cancer patients (102 with HRR genes mutations) from three centers (center 1: training and internal test cohorts, center 2 and 3: external test cohorts) underwent multiparametric MRI. Each patient's index tumor lesion was delineated on T2-weighted imaging (T2WI), dynamic contrast enhancement (DCE) MRI, diffusion-weighted imaging (DWI), and apparent diffusion coefficient (ADC) images to obtain 428 radiomics features. Features associated with mutations were selected from clinicopathological features using Mann-Whitney U and Logistic regression (LR) test, radiomics features using Least Absolute Shrinkage and Selection Operator. Clinicopathological model was constructed with selected clinicopathological features. Logistic regression (LR), support vector machine (SVM) and linear discriminant analysis (LDA) classifiers were used to construct Radiomics and combined clinicopathological-radiomics models. Predictive efficiencies of models were compared using areas under the receiver operating characteristic curve (AUC).</div></div><div><h3>Results</h3><div>One clinicopathological and six radiomics features were selected. Radiomics with SVM, LR, LDA and Clinicopathological models achieved AUCs of 0.76, 0.76, 0.76, 0.68 and 0.75, 0.76, 0.67, 0.73 in internal and external test cohort. AUCs of combined clinicopathological-radiomics models with LDA in internal and external test cohort (0.83 and 0.82) were slightly higher than combined models with LR (0.81 and 0.79) and SVM (both 0.80) (<em>P</em> > 0.05), but were significantly higher than radiomics and clinicopathological models in both cohorts (<em>P</em> < 0.05).</div></div><div><h3>Conclusion</h3><div>LDA classifier incorporating radiomics and clinicopathological features predicting could effectively predict HRR genes mutations in prostate cancer.</div></div>","PeriodicalId":18165,"journal":{"name":"Magnetic resonance imaging","volume":"124 ","pages":"Article 110534"},"PeriodicalIF":2.0,"publicationDate":"2025-10-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145275099","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-10-09DOI: 10.1016/j.mri.2025.110540
Ke Xu, Daniel B. Rowe
FMRI has been a safe medical imaging tool to study brain function by demonstrating the spatial and temporal changes in brain metabolism in recent decades. To capture brain functionality more efficiently, efforts focus on accelerating image acquisition acquired per unit of time that create each volume image without losing full anatomical structure. The Simultaneous Multi-Slice (SMS) technique provides a reconstruction method where multiple slices are acquired and aliased concurrently. Traditional imaging techniques such as SENSE and GRAPPA can reconstruct an image from less measured data but have their drawbacks. The Controlled Aliasing in Parallel Imaging (CAIPI) and view angle tilting (VAT) techniques achieve slice-wise image shift to decrease the influence of the geometry factor (g-factor) of coil sensitivities and prevent the singular problem of the design matrix. In this paper, a Bayesian CAIPIVAT approach for multi-coil separation of parallel encoded complex-valued slices (mSPECS-CAIPIVAT) with a novel SMS approach is presented and combined with the Hadamard phase-encoding method for image separation. Our proposed approach was applied to simulation and experimental studies showing a decrease in the influence of the g-factor while increasing the brain activation detection rate. The signal-to-noise ratio and the contrast-to-noise ratio are also improved by our approach.
{"title":"A Bayesian CAIPIVAT approach with through-plane acceleration to enhance efficiency of simultaneously encoded slice acquisition in FMRI","authors":"Ke Xu, Daniel B. Rowe","doi":"10.1016/j.mri.2025.110540","DOIUrl":"10.1016/j.mri.2025.110540","url":null,"abstract":"<div><div>FMRI has been a safe medical imaging tool to study brain function by demonstrating the spatial and temporal changes in brain metabolism in recent decades. To capture brain functionality more efficiently, efforts focus on accelerating image acquisition acquired per unit of time that create each volume image without losing full anatomical structure. The Simultaneous Multi-Slice (SMS) technique provides a reconstruction method where multiple slices are acquired and aliased concurrently. Traditional imaging techniques such as SENSE and GRAPPA can reconstruct an image from less measured data but have their drawbacks. The Controlled Aliasing in Parallel Imaging (CAIPI) and view angle tilting (VAT) techniques achieve slice-wise image shift to decrease the influence of the geometry factor (<em>g</em>-factor) of coil sensitivities and prevent the singular problem of the design matrix. In this paper, a Bayesian CAIPIVAT approach for multi-coil separation of parallel encoded complex-valued slices (mSPECS-CAIPIVAT) with a novel SMS approach is presented and combined with the Hadamard phase-encoding method for image separation. Our proposed approach was applied to simulation and experimental studies showing a decrease in the influence of the <em>g</em>-factor while increasing the brain activation detection rate. The signal-to-noise ratio and the contrast-to-noise ratio are also improved by our approach.</div></div>","PeriodicalId":18165,"journal":{"name":"Magnetic resonance imaging","volume":"124 ","pages":"Article 110540"},"PeriodicalIF":2.0,"publicationDate":"2025-10-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145258594","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-10-08DOI: 10.1016/j.mri.2025.110535
Qinfeng Zhu , Ruicheng Ba , Zuozhen Cao , Yao Shen , Haotian Li , Yi-Cheng Hsu , Xu Yan , Dan Wu
Purpose
This study proposes an optimized acquisition protocol to minimize the unwanted diffusion weighting in DW-STEAM and facilitate time-dependent diffusion kurtosis imaging (tDKI).
Methods
We first corrected the diffusion-direction-dependent shift by optimizing the diffusion gradient amplitude. We then proposed to use low-b-value in the reference image, instead of the conventional non-diffusion-weighted (b0) acquisition, and removed the crusher gradients. The tDKI measurements from the proposed strategy were compared with conventional DWIs that included crushers in the b0 image, in a water phantom and in healthy adults (n = 8) on 3 T, and the water exchange time (τex) was calculated from the tDKI measurements. The optimal strategy was tested in five ex vivo human brains, and the results were compared with in vivo data.
Results
Neglecting the diffusion weighting in b0 images introduced an artificial time dependence in the apparent diffusivity of the water phantom, and resulted in elevated estimates of water exchange time (τex) in in the in vivo data. Additionally, the AIC indicated that, even when diffusion weighting in the b0 images was accounted for, the kurtosis estimation remained less stable in the in vivo than with the crusher-free approach. In contrast, using low-b-value images as reference measurements yielded reasonable tDKI estimates, with τex ranging from 15 to 40 ms for gray matter in vivo and from 20 to 60 ms for gray matter ex vivo.
Conclusions
The optimized DW-STEAM acquisition eliminated artificial diffusion-time dependence, enabling the acquisition of accurate tDKI data for mapping structural morphology and transmembrane permeability.
{"title":"Correcting unwanted diffusion weighting in diffusion-weighted-STEAM sequence for time-dependent diffusion kurtosis imaging","authors":"Qinfeng Zhu , Ruicheng Ba , Zuozhen Cao , Yao Shen , Haotian Li , Yi-Cheng Hsu , Xu Yan , Dan Wu","doi":"10.1016/j.mri.2025.110535","DOIUrl":"10.1016/j.mri.2025.110535","url":null,"abstract":"<div><h3>Purpose</h3><div>This study proposes an optimized acquisition protocol to minimize the unwanted diffusion weighting in DW-STEAM and facilitate time-dependent diffusion kurtosis imaging (tDKI).</div></div><div><h3>Methods</h3><div>We first corrected the diffusion-direction-dependent shift by optimizing the diffusion gradient amplitude. We then proposed to use low-b-value in the reference image, instead of the conventional non-diffusion-weighted (b0) acquisition, and removed the crusher gradients. The tDKI measurements from the proposed strategy were compared with conventional DWIs that included crushers in the b0 image, in a water phantom and in healthy adults (<em>n</em> = 8) on 3 T, and the water exchange time (<em>τ</em><sub><em>ex</em></sub>) was calculated from the tDKI measurements. The optimal strategy was tested in five <em>ex vivo</em> human brains, and the results were compared with <em>in vivo</em> data.</div></div><div><h3>Results</h3><div>Neglecting the diffusion weighting in b0 images introduced an artificial time dependence in the apparent diffusivity of the water phantom, and resulted in elevated estimates of water exchange time (<em>τ</em><sub><em>ex</em></sub>) in in the <em>in vivo</em> data. Additionally, the AIC indicated that, even when diffusion weighting in the b0 images was accounted for, the kurtosis estimation remained less stable in the <em>in vivo</em> than with the crusher-free approach. In contrast, using low-b-value images as reference measurements yielded reasonable tDKI estimates, with <em>τ</em><sub><em>ex</em></sub> ranging from 15 to 40 ms for gray matter <em>in vivo</em> and from 20 to 60 ms for gray matter <em>ex vivo</em>.</div></div><div><h3>Conclusions</h3><div>The optimized DW-STEAM acquisition eliminated artificial diffusion-time dependence, enabling the acquisition of accurate tDKI data for mapping structural morphology and transmembrane permeability.</div></div>","PeriodicalId":18165,"journal":{"name":"Magnetic resonance imaging","volume":"124 ","pages":"Article 110535"},"PeriodicalIF":2.0,"publicationDate":"2025-10-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145267178","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-10-06DOI: 10.1016/j.mri.2025.110539
Elyssa M. McMaster , Nancy R. Newlin , Chloe Cho , Gaurav Rudravaram , Adam M. Saunders , Aravind R. Krishnan , Lucas W. Remedios , Michael E. Kim , Hanliang Xu , Kurt G. Schilling , François Rheault , Laurie E. Cutting , Bennett A. Landman
Purpose
Diffusion weighted MRI (dMRI) and its models of neural structure provide insight into human brain organization and variations in white matter. A recent study by McMaster, et al. showed that complex graph measures of the connectome, the graphical representation of a tractogram, vary with spatial sampling changes, but biases introduced by anisotropic voxels in the process have not been well characterized. This study uses microstructural measures (fractional anisotropy and mean diffusivity) and white matter bundle properties (bundle volume, length, and surface area) to further understand the effect of anisotropic voxels on microstructure and tractography.
Methods
The statistical significance of the selected measures derived from dMRI data were assessed by comparing three white matter bundles at different spatial resolutions with 44 subjects from the Human Connectome Project – Young Adult dataset scan/rescan data using the Wilcoxon Signed-Rank test. The original isotropic resolution (1.25 mm isotropic) was explored with 6 anisotropic resolutions with 0.25 mm incremental steps in the z dimension. Then, all generated resolutions were upsampled to 1.25 mm isotropic and 1 mm isotropic.
Results
There were statistically significant differences between at least one microstructural and one bundle measure at every resolution (, corrected for multiple comparisons). Cohen's coefficient evaluated the effect size of anisotropic voxels on microstructure and tractography.
Conclusion
Fractional anisotropy and mean diffusivity cannot be recovered with basic up-sampling from low quality data with gold-standard data with the methods selected for this study. However, the bundle measures across our selected regions of interest become more repeatable when voxels are resampled to 1 mm isotropic.
{"title":"Sensitivity of quantitative diffusion MRI tractography and microstructure to anisotropic spatial sampling","authors":"Elyssa M. McMaster , Nancy R. Newlin , Chloe Cho , Gaurav Rudravaram , Adam M. Saunders , Aravind R. Krishnan , Lucas W. Remedios , Michael E. Kim , Hanliang Xu , Kurt G. Schilling , François Rheault , Laurie E. Cutting , Bennett A. Landman","doi":"10.1016/j.mri.2025.110539","DOIUrl":"10.1016/j.mri.2025.110539","url":null,"abstract":"<div><h3>Purpose</h3><div>Diffusion weighted MRI (dMRI) and its models of neural structure provide insight into human brain organization and variations in white matter. A recent study by McMaster, et al. showed that complex graph measures of the connectome, the graphical representation of a tractogram, vary with spatial sampling changes, but biases introduced by anisotropic voxels in the process have not been well characterized. This study uses microstructural measures (fractional anisotropy and mean diffusivity) and white matter bundle properties (bundle volume, length, and surface area) to further understand the effect of anisotropic voxels on microstructure and tractography.</div></div><div><h3>Methods</h3><div>The statistical significance of the selected measures derived from dMRI data were assessed by comparing three white matter bundles at different spatial resolutions with 44 subjects from the Human Connectome Project – Young Adult dataset scan/rescan data using the Wilcoxon Signed-Rank test. The original isotropic resolution (1.25 mm isotropic) was explored with 6 anisotropic resolutions with 0.25 mm incremental steps in the <em>z</em> dimension. Then, all generated resolutions were upsampled to 1.25 mm isotropic and 1 mm isotropic.</div></div><div><h3>Results</h3><div>There were statistically significant differences between at least one microstructural and one bundle measure at every resolution (<span><math><mi>p</mi><mo>≤</mo><mn>0.05</mn></math></span>, corrected for multiple comparisons). Cohen's <span><math><mi>d</mi></math></span> coefficient evaluated the effect size of anisotropic voxels on microstructure and tractography.</div></div><div><h3>Conclusion</h3><div>Fractional anisotropy and mean diffusivity cannot be recovered with basic up-sampling from low quality data with gold-standard data with the methods selected for this study. However, the bundle measures across our selected regions of interest become more repeatable when voxels are resampled to 1 mm isotropic.</div></div>","PeriodicalId":18165,"journal":{"name":"Magnetic resonance imaging","volume":"124 ","pages":"Article 110539"},"PeriodicalIF":2.0,"publicationDate":"2025-10-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145244670","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-10-06DOI: 10.1016/j.mri.2025.110538
Jorge Escartín , Pilar López-Úbeda , Teodoro Martín-Noguerol , Antonio Luna
Purpose
Ischemic stroke, a leading cause of global disability and mortality, demands precise etiological classification for effective management. The variability in the use of existing stroke classification systems, along with the challenges in manual etiological labeling from brain MRI radiological reports, calls for an innovative approach. This study aims to develop and evaluate a Natural Language Processing (NLP) algorithm using transformer-based models for the extraction and classification of ischemic stroke types from MRI reports, enhancing diagnostic efficiency and stroke management.
Methods
We built a dataset comprising 635 brain MRI reports, annotated for four distinct ischemic stroke types. All were clinically consistent with focal neurologic impairment due to stroke. The study involved evaluating two pre-trained models BERT (Bert clinical and Beto) and two models RoBERTa (Roberta clinical trials and Roberta biomedical), focusing on their ability to accurately classify stroke subtypes.
Results
The Roberta biomedical model emerged as the most effective, demonstrating superior performance with an accuracy of 76.7 % with statistically significant results. This model also achieved the highest precision, recall, and F1 scores across all stroke types, indicating its robustness in stroke subtype classification.
Conclusion
The study highlights the potential of NLP algorithms in automating stroke classification from MRI reports, which could significantly aid in diagnostic processes and streamline stroke management strategies.
{"title":"Role of large language models for etiological classification of brain stroke based on MRI brain reports: a feasibility study","authors":"Jorge Escartín , Pilar López-Úbeda , Teodoro Martín-Noguerol , Antonio Luna","doi":"10.1016/j.mri.2025.110538","DOIUrl":"10.1016/j.mri.2025.110538","url":null,"abstract":"<div><h3>Purpose</h3><div>Ischemic stroke, a leading cause of global disability and mortality, demands precise etiological classification for effective management. The variability in the use of existing stroke classification systems, along with the challenges in manual etiological labeling from brain MRI radiological reports, calls for an innovative approach. This study aims to develop and evaluate a Natural Language Processing (NLP) algorithm using transformer-based models for the extraction and classification of ischemic stroke types from MRI reports, enhancing diagnostic efficiency and stroke management.</div></div><div><h3>Methods</h3><div>We built a dataset comprising 635 brain MRI reports, annotated for four distinct ischemic stroke types. All were clinically consistent with focal neurologic impairment due to stroke. The study involved evaluating two pre-trained models BERT (Bert clinical and Beto) and two models RoBERTa (Roberta clinical trials and Roberta biomedical), focusing on their ability to accurately classify stroke subtypes.</div></div><div><h3>Results</h3><div>The Roberta biomedical model emerged as the most effective, demonstrating superior performance with an accuracy of 76.7 % with statistically significant results. This model also achieved the highest precision, recall, and F1 scores across all stroke types, indicating its robustness in stroke subtype classification.</div></div><div><h3>Conclusion</h3><div>The study highlights the potential of NLP algorithms in automating stroke classification from MRI reports, which could significantly aid in diagnostic processes and streamline stroke management strategies.</div></div>","PeriodicalId":18165,"journal":{"name":"Magnetic resonance imaging","volume":"124 ","pages":"Article 110538"},"PeriodicalIF":2.0,"publicationDate":"2025-10-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145244717","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-10-03DOI: 10.1016/j.mri.2025.110537
Zhiyuan He , Yiwen Wang , Yunxi Li , Laimin Zhu , Yueqin Chen , Ning Mao , Wenwen Zhao , Xiuzheng Yue , Yufei Xue , Shuzhen Wang , Weiwei Wang , Zhanguo Sun
Purpose
To develop and validate a combined model integrating multimodal magnetic resonance imaging (MRI) and clinical-pathological parameters to assess hypoxia status and hypoxia-inducible factor-1α (HIF-1α) expression in mass-like breast cancer.
Methods
This retrospective cohort study included 197 patients with mass-like breast cancer from two medical centers, who were divided into a training set (n = 104), an internal validation set (n = 45), and an external validation set (n = 48). Clinical-pathological and multimodal MRI parameters were analyzed using histopathology as the reference. The combined model was developed through logistic and least absolute shrinkage and selection operator (LASSO) regression analysis to identify features, and visualized using nomograms.
Results
Axillary lymph node (ALN) metastasis, time-intensity curve (TIC) type, mean diffusivity (MD), volumetric transfer constant (Ktrans), and rate constant (Kep) were selected to construct the combined model. The diagnostic performance of the combined model [area under the curve (AUC) = 0.967, 95 % CI: 0.913–0.992] (training), was significantly better than that of other individual models (P < 0.001). This performance was replicated in internal (AUC= 0.907) and external (AUC = 0.928) validation sets. The nomogram of the combined model showed excellent calibration (Hosmer-Lemeshow P = 0.653) and the highest net benefit across threshold probabilities in the decision curve analysis (DCA). Tumors with high HIF-1α expression exhibited higher ALN metastasis, histological grade, unclear margins, Type 3 TIC, and elevated Ktrans, Kep, and MK, but reduced MD values.
Conclusions
The combined model (ALN + TIC + MD + Ktrans + Kep) shows potential as a reliable tool for predicting HIF-1α expression levels in mass-like breast cancer.
{"title":"Combined model based on multimodal imaging and clinical pathology: A new pathway to assess hypoxia status and HIF-1α expression in breast cancer","authors":"Zhiyuan He , Yiwen Wang , Yunxi Li , Laimin Zhu , Yueqin Chen , Ning Mao , Wenwen Zhao , Xiuzheng Yue , Yufei Xue , Shuzhen Wang , Weiwei Wang , Zhanguo Sun","doi":"10.1016/j.mri.2025.110537","DOIUrl":"10.1016/j.mri.2025.110537","url":null,"abstract":"<div><h3>Purpose</h3><div>To develop and validate a combined model integrating multimodal magnetic resonance imaging (MRI) and clinical-pathological parameters to assess hypoxia status and hypoxia-inducible factor-1α (HIF-1α) expression in mass-like breast cancer.</div></div><div><h3>Methods</h3><div>This retrospective cohort study included 197 patients with mass-like breast cancer from two medical centers, who were divided into a training set (<em>n</em> = 104), an internal validation set (<em>n</em> = 45), and an external validation set (<em>n</em> = 48). Clinical-pathological and multimodal MRI parameters were analyzed using histopathology as the reference. The combined model was developed through logistic and least absolute shrinkage and selection operator (LASSO) regression analysis to identify features, and visualized using nomograms.</div></div><div><h3>Results</h3><div>Axillary lymph node (ALN) metastasis, time-intensity curve (TIC) type, mean diffusivity (MD), volumetric transfer constant (K<sup>trans</sup>), and rate constant (Kep) were selected to construct the combined model. The diagnostic performance of the combined model [area under the curve (AUC) = 0.967, 95 % CI: 0.913–0.992] (training), was significantly better than that of other individual models (<em>P</em> < 0.001). This performance was replicated in internal (AUC= 0.907) and external (AUC = 0.928) validation sets. The nomogram of the combined model showed excellent calibration (Hosmer-Lemeshow <em>P</em> = 0.653) and the highest net benefit across threshold probabilities in the decision curve analysis (DCA). Tumors with high HIF-1α expression exhibited higher ALN metastasis, histological grade, unclear margins, Type 3 TIC, and elevated K<sup>trans</sup>, Kep, and MK, but reduced MD values.</div></div><div><h3>Conclusions</h3><div>The combined model (ALN + TIC + MD + K<sup>trans</sup> + Kep) shows potential as a reliable tool for predicting HIF-1α expression levels in mass-like breast cancer.</div></div>","PeriodicalId":18165,"journal":{"name":"Magnetic resonance imaging","volume":"124 ","pages":"Article 110537"},"PeriodicalIF":2.0,"publicationDate":"2025-10-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145232921","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-09-30DOI: 10.1016/j.mri.2025.110531
Seong-Eun Kim , John A. Roberts , J. Rock Hadley , J. Scott McNally , Gerald S. Treiman , Yibin Xie , Debiao Li , Kim-Lien Nguyen , Dimitrios Mitsouras , Jonas Schollenberger , David Saloner , Arunbalaji Pugazhendhi , Kevin J. Johnson , Maria Altbach , Herman Morris , Kevin DeMarco , Vibhas Deshpande , Pedro Itriago-Leon , Dennis L. Parker
Purpose
This study aims to optimize the Magnetization Prepared Rapid Acquisition Gradient Echo (MPRAGE) technique to improve the accuracy of intraplaque hemorrhage (IPH) detection in carotid diseases and enhance reliability for clinical use, addressing the challenge of inconsistent blood suppression observed in commonly used clinical protocols.
Methods
Bloch equation simulations were used to evaluate the effects of inversion time and flip angle on blood suppression and tissue contrast. The optimized parameters were implemented on a clinical 3 T scanner and tested in four subjects. Quantitative measures included blood–muscle contrast and contrast-to-noise ratios (CNR) for lumen–wall and lumen–IPH. Imaging was performed in 21 patients with carotid artery disease, two MPRAGE acquisitions were performed per patient: (1) at the magnet isocenter with the standard IR pulse, and (2) at a 50-mm shifted position toward the heart with the wideband IR pulse. Qualitative image quality was evaluated independently by two neuroradiologists using a predefined 4-point scale (1 = poor, 4 = excellent) for blood suppression, vessel wall clarity, motion artifacts, and IPH visualization, in addition to quantitative SNR and CNR measurements.
Results
The optimized MPRAGE sequence demonstrated significantly improved blood suppression, with blood–muscle contrast increasing from 0.42 ± 0.08 to 0.61 ± 0.07 (p = 0.002). Lumen–wall CNR increased from 15.2 ± 3.4 to 22.8 ± 4.1 (p = 0.01), and lumen–IPH CNR increased from 18.7 ± 5.2 to 26.3 ± 6.0 (p = 0.004). These improvements enhanced vessel wall delineation and IPH conspicuity.
Conclusion
Optimization of MPRAGE parameters enhances blood suppression and contrast, providing improved visualization of the carotid vessel wall and more reliable detection of IPH. This approach may increase the diagnostic accuracy of carotid plaque imaging.
{"title":"Optimizing MPRAGE for enhanced blood suppression and intraplaque hemorrhage detection in carotid artery imaging","authors":"Seong-Eun Kim , John A. Roberts , J. Rock Hadley , J. Scott McNally , Gerald S. Treiman , Yibin Xie , Debiao Li , Kim-Lien Nguyen , Dimitrios Mitsouras , Jonas Schollenberger , David Saloner , Arunbalaji Pugazhendhi , Kevin J. Johnson , Maria Altbach , Herman Morris , Kevin DeMarco , Vibhas Deshpande , Pedro Itriago-Leon , Dennis L. Parker","doi":"10.1016/j.mri.2025.110531","DOIUrl":"10.1016/j.mri.2025.110531","url":null,"abstract":"<div><h3>Purpose</h3><div>This study aims to optimize the Magnetization Prepared Rapid Acquisition Gradient Echo (MPRAGE) technique to improve the accuracy of intraplaque hemorrhage (IPH) detection in carotid diseases and enhance reliability for clinical use, addressing the challenge of inconsistent blood suppression observed in commonly used clinical protocols.</div></div><div><h3>Methods</h3><div>Bloch equation simulations were used to evaluate the effects of inversion time and flip angle on blood suppression and tissue contrast. The optimized parameters were implemented on a clinical 3 T scanner and tested in four subjects. Quantitative measures included blood–muscle contrast and contrast-to-noise ratios (CNR) for lumen–wall and lumen–IPH. Imaging was performed in 21 patients with carotid artery disease, two MPRAGE acquisitions were performed per patient: (1) at the magnet isocenter with the standard IR pulse, and (2) at a 50-mm shifted position toward the heart with the wideband IR pulse. Qualitative image quality was evaluated independently by two neuroradiologists using a predefined 4-point scale (1 = poor, 4 = excellent) for blood suppression, vessel wall clarity, motion artifacts, and IPH visualization, in addition to quantitative SNR and CNR measurements.</div></div><div><h3>Results</h3><div>The optimized MPRAGE sequence demonstrated significantly improved blood suppression, with blood–muscle contrast increasing from 0.42 ± 0.08 to 0.61 ± 0.07 (<em>p</em> = 0.002). Lumen–wall CNR increased from 15.2 ± 3.4 to 22.8 ± 4.1 (<em>p</em> = 0.01), and lumen–IPH CNR increased from 18.7 ± 5.2 to 26.3 ± 6.0 (<em>p</em> = 0.004). These improvements enhanced vessel wall delineation and IPH conspicuity.</div></div><div><h3>Conclusion</h3><div>Optimization of MPRAGE parameters enhances blood suppression and contrast, providing improved visualization of the carotid vessel wall and more reliable detection of IPH. This approach may increase the diagnostic accuracy of carotid plaque imaging.</div></div>","PeriodicalId":18165,"journal":{"name":"Magnetic resonance imaging","volume":"124 ","pages":"Article 110531"},"PeriodicalIF":2.0,"publicationDate":"2025-09-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145213152","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-09-25DOI: 10.1016/j.mri.2025.110532
Zhengyi Lu , Hao Liang , Ming Lu , Dann Martin , Benjamin M. Hardy , Benoit M. Dawant , Xiao Wang , Xinqiang Yan , Yuankai Huo
Accurate and individualized human head models are becoming increasingly important for electromagnetic (EM) simulations. These simulations depend on precise anatomical representations to realistically model electric and magnetic field distributions, particularly when evaluating Specific Absorption Rate (SAR) within safety guidelines. State of the art simulations use the Virtual Population due to limited public resources and the impracticality of manually annotating patient data at scale. This paper introduces Personalized Head-based Automatic Simulation for EM properties (PHASE), an automated open-source toolbox that generates high-resolution, patient-specific head models for EM simulations using paired T1-weighted (T1w) magnetic resonance imaging (MRI) and computed tomography (CT) scans with 14 tissue labels. To evaluate the performance of PHASE models, we conduct semi-automated segmentation and EM simulations on 15 real human patients, serving as the gold standard reference. The PHASE model achieved comparable global SAR and localized SAR averaged over 10 grams of tissue (SAR-10g), demonstrating its potential as a promising tool for generating large-scale human model datasets in the future. The code and models of PHASE toolbox have been made publicly available: https://github.com/hrlblab/PHASE.
{"title":"PHASE: Personalized Head-based Automatic Simulation for Electromagnetic properties in 7T MRI","authors":"Zhengyi Lu , Hao Liang , Ming Lu , Dann Martin , Benjamin M. Hardy , Benoit M. Dawant , Xiao Wang , Xinqiang Yan , Yuankai Huo","doi":"10.1016/j.mri.2025.110532","DOIUrl":"10.1016/j.mri.2025.110532","url":null,"abstract":"<div><div>Accurate and individualized human head models are becoming increasingly important for electromagnetic (EM) simulations. These simulations depend on precise anatomical representations to realistically model electric and magnetic field distributions, particularly when evaluating Specific Absorption Rate (SAR) within safety guidelines. State of the art simulations use the Virtual Population due to limited public resources and the impracticality of manually annotating patient data at scale. This paper introduces Personalized Head-based Automatic Simulation for EM properties (PHASE), an automated open-source toolbox that generates high-resolution, patient-specific head models for EM simulations using paired T1-weighted (T1w) magnetic resonance imaging (MRI) and computed tomography (CT) scans with 14 tissue labels. To evaluate the performance of PHASE models, we conduct semi-automated segmentation and EM simulations on 15 real human patients, serving as the gold standard reference. The PHASE model achieved comparable global SAR and localized SAR averaged over 10 grams of tissue (SAR-10g), demonstrating its potential as a promising tool for generating large-scale human model datasets in the future. The code and models of PHASE toolbox have been made publicly available: <span><span>https://github.com/hrlblab/PHASE</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":18165,"journal":{"name":"Magnetic resonance imaging","volume":"124 ","pages":"Article 110532"},"PeriodicalIF":2.0,"publicationDate":"2025-09-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145182084","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-09-23DOI: 10.1016/j.mri.2025.110533
Kamil Lipiński , Grzegorz Domański , Piotr Bogorodzki
Quantitative, non-invasive measurement of perfusion remains a challenge in MRI. The Intravoxel Incoherent Motion (IVIM) technique, based on Diffusion-Weighted Imaging (DWI), offers a method to estimate perfusion at the sub-voxel level by modeling blood flow as a pseudo-diffusive process within a microvascular network. Key IVIM-derived parameters include the perfusion fraction (f) and the pseudo-diffusion coefficient (D*). While under specific structural assumptions, IVIM can be used to estimate Cerebral Blood Flow (CBF), direct conversions from IVIM metrics to CBF are rarely validated. This study investigates the accuracy of the IVIM-based perfusion measures, exploring coherent flow influence on signal from voxel. To assess the relation between flow (ground truth CBF), IVIM parameters, and calculated CBF, a phantom, composed of agarose gel and tubes mimicking arterioles with regulated flow velocity, was constructed. Thus, a signal fraction from the “fast” compartment, described by D* in the IVIM model, is represented by flow in tubes occupying a fixed fraction of the imaged voxel, simulating desired Cerebral Blood Volume (CBV) and CBF values. The phantom was scanned using IVIM-optimized DWI protocol, and data were processed for the IVIM measures. Besides IVIM, we also propose a model based on extended Bloch-Torrey equations with coherent flow term. Results demonstrate that IVIM can yield reasonable estimates of CBF and CBV within physiologically relevant flow ranges (0.5–2 mm/s) observed in the brain vasculature. However, a consistent overestimation of flow (up to 200 %) was obscerved at higher velocities, especially in arterial or vein-like flow conditions.
{"title":"Coherent flow effects on IVIM-based perfusion measurements: A phantom study in 3 T","authors":"Kamil Lipiński , Grzegorz Domański , Piotr Bogorodzki","doi":"10.1016/j.mri.2025.110533","DOIUrl":"10.1016/j.mri.2025.110533","url":null,"abstract":"<div><div>Quantitative, non-invasive measurement of perfusion remains a challenge in MRI. The Intravoxel Incoherent Motion (IVIM) technique, based on Diffusion-Weighted Imaging (DWI), offers a method to estimate perfusion at the sub-voxel level by modeling blood flow as a pseudo-diffusive process within a microvascular network. Key IVIM-derived parameters include the perfusion fraction (f) and the pseudo-diffusion coefficient (D*). While under specific structural assumptions, IVIM can be used to estimate Cerebral Blood Flow (CBF), direct conversions from IVIM metrics to CBF are rarely validated. This study investigates the accuracy of the IVIM-based perfusion measures, exploring coherent flow influence on signal from voxel. To assess the relation between flow (ground truth CBF), IVIM parameters, and calculated CBF, a phantom, composed of agarose gel and tubes mimicking arterioles with regulated flow velocity, was constructed. Thus, a signal fraction from the “fast” compartment, described by D* in the IVIM model, is represented by flow in tubes occupying a fixed fraction of the imaged voxel, simulating desired Cerebral Blood Volume (CBV) and CBF values. The phantom was scanned using IVIM-optimized DWI protocol, and data were processed for the IVIM measures. Besides IVIM, we also propose a model based on extended Bloch-Torrey equations with coherent flow term. Results demonstrate that IVIM can yield reasonable estimates of CBF and CBV within physiologically relevant flow ranges (0.5–2 mm/s) observed in the brain vasculature. However, a consistent overestimation of flow (up to 200 %) was obscerved at higher velocities, especially in arterial or vein-like flow conditions.</div></div>","PeriodicalId":18165,"journal":{"name":"Magnetic resonance imaging","volume":"124 ","pages":"Article 110533"},"PeriodicalIF":2.0,"publicationDate":"2025-09-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145149806","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
扫码关注我们
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号