Meta-Radiology最新文献

S-RRG-Bench: Structured radiology report generation with fine-grained evaluation framework S-RRG-Bench：具有细粒度评估框架的结构化放射学报告生成

Meta-Radiology

Pub Date : 2025-12-01 DOI: 10.1016/j.metrad.2025.100171

Yingshu Li , Yunyi Liu , Zhanyu Wang , Xinyu Liang , Lingqiao Liu , Lei Wang , Luping Zhou

Radiology report generation (RRG) for diagnostic images, such as chest X-rays, plays a pivotal role in both clinical practice and AI. Traditional free-text reports suffer from redundancy and inconsistent language, complicating the extraction of critical clinical details. Structured radiology report generation (S-RRG) offers a promising solution by organizing information into standardized, concise formats. However, existing approaches often rely on classification or visual question answering (VQA) pipelines that require predefined label sets and produce only fragmented outputs. Template-based approaches, which generate reports by replacing keywords within fixed sentence patterns, further compromise expressiveness and often omit clinically important details. In this work, we present a novel approach to S-RRG that includes dataset construction, model training, and the introduction of a new evaluation framework. We first create a robust chest X-ray dataset (MIMIC-STRUC) that includes disease names, severity levels, probabilities, and anatomical locations, ensuring that the dataset is both clinically relevant and well-structured. We train an LLM-based model to generate standardized, high-quality reports. To assess the generated reports, we propose a specialized evaluation metric (S-Score) that not only measures disease prediction accuracy but also evaluates the precision of disease-specific details, thus offering a clinically meaningful metric for report quality that focuses on elements critical to clinical decision-making and demonstrates a stronger alignment with human assessments. Our approach highlights the effectiveness of structured reports and the importance of a tailored evaluation metric for S-RRG, providing a more clinically relevant measure of report quality.

用于诊断图像（如胸部x光片）的放射学报告生成（RRG）在临床实践和人工智能中都起着关键作用。传统的自由文本报告存在冗余和不一致的语言，使关键临床细节的提取变得复杂。结构化放射学报告生成（S-RRG）通过将信息组织成标准化、简洁的格式，提供了一个很有前途的解决方案。然而，现有的方法通常依赖于分类或可视化问答（VQA）管道，这些管道需要预定义的标签集，并且只产生碎片化的输出。基于模板的方法通过替换固定句型中的关键字来生成报告，这进一步损害了表达能力，并且经常忽略临床重要的细节。在这项工作中，我们提出了一种新的S-RRG方法，包括数据集构建、模型训练和引入新的评估框架。我们首先创建了一个健壮的胸部x射线数据集（MIMIC-STRUC），其中包括疾病名称、严重程度、概率和解剖位置，确保数据集既与临床相关又结构良好。我们训练一个基于法学硕士的模型来生成标准化的、高质量的报告。为了评估生成的报告，我们提出了一个专门的评估指标（S-Score），该指标不仅衡量疾病预测的准确性，还评估疾病特定细节的准确性，从而为报告质量提供了一个具有临床意义的指标，该指标关注对临床决策至关重要的要素，并展示了与人类评估更强的一致性。我们的方法强调了结构化报告的有效性和为S-RRG量身定制的评估指标的重要性，为报告质量提供了更具临床相关性的衡量标准。

{"title":"S-RRG-Bench: Structured radiology report generation with fine-grained evaluation framework","authors":"Yingshu Li , Yunyi Liu , Zhanyu Wang , Xinyu Liang , Lingqiao Liu , Lei Wang , Luping Zhou","doi":"10.1016/j.metrad.2025.100171","DOIUrl":"10.1016/j.metrad.2025.100171","url":null,"abstract":"<div><div>Radiology report generation (RRG) for diagnostic images, such as chest X-rays, plays a pivotal role in both clinical practice and AI. Traditional free-text reports suffer from redundancy and inconsistent language, complicating the extraction of critical clinical details. Structured radiology report generation (S-RRG) offers a promising solution by organizing information into standardized, concise formats. However, existing approaches often rely on classification or visual question answering (VQA) pipelines that require predefined label sets and produce only fragmented outputs. Template-based approaches, which generate reports by replacing keywords within fixed sentence patterns, further compromise expressiveness and often omit clinically important details. In this work, we present a novel approach to S-RRG that includes dataset construction, model training, and the introduction of a new evaluation framework. We first create a robust chest X-ray dataset (MIMIC-STRUC) that includes disease names, severity levels, probabilities, and anatomical locations, ensuring that the dataset is both clinically relevant and well-structured. We train an LLM-based model to generate standardized, high-quality reports. To assess the generated reports, we propose a specialized evaluation metric (S-Score) that not only measures disease prediction accuracy but also evaluates the precision of disease-specific details, thus offering a clinically meaningful metric for report quality that focuses on elements critical to clinical decision-making and demonstrates a stronger alignment with human assessments. Our approach highlights the effectiveness of structured reports and the importance of a tailored evaluation metric for S-RRG, providing a more clinically relevant measure of report quality.</div></div>","PeriodicalId":100921,"journal":{"name":"Meta-Radiology","volume":"3 4","pages":"Article 100171"},"PeriodicalIF":0.0,"publicationDate":"2025-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145623611","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}

引用次数: 0

Predicting efficacy of electroconvulsive therapy for adolescent major depressive disorder using a dual-branch graph attention network fusing multi-modal MRI 利用双分支图注意网络融合多模态MRI预测电惊厥治疗青少年重度抑郁症的疗效

Meta-Radiology

Pub Date : 2025-10-13 DOI: 10.1016/j.metrad.2025.100184

Jingyu Zhang , Ruiyun Zhu , Jing Min , Tong Zhu , Mengqi Liu , Renqiang Yu , Yang Huang , Chao Li , Sizhu Wu , Du Lei

Purpose

Major Depressive Disorder (MDD) significantly contributes to global disease burden, and Electroconvulsive Therapy (ECT) is an effective yet variable treatment. This study aims to develop an individualized prediction framework for ECT treatment response in adolescent MDD patients using multi-modal magnetic resonance imaging (MRI) and advanced deep learning.

Methods

We recruited 27 adolescent MDD patients undergoing ECT, acquiring structural MRI (sMRI) and functional MRI (fMRI) before and after treatment. Individual morphological similarity networks and functional connectivity networks were created by utilizing sMRI and fMRI data, respectively. We introduced a novel Dual-Branch Graph Attention Network (DBGAN) which integrates two parallel graph attention networks for utilizing similarity and connectivity networks. The proposed deep model dynamically fuses information from sMRI and fMRI via a cross-attention mechanism, improving the prediction performance on ECT treatment response.

Results

Among 27 participants, 21 responded positively to ECT. According to experimental results, our DBGAN outperformed traditional machine learning models and deep learning models, achieving a mean accuracy of 0.853, precision of 0.920, recall of 0.910, and an F1-score of 0.905. Interpretability analyses indicated that predictive decisions were influenced by fMRI signals primarily in the right posterior insula and right dorsal cingulate gyrus, and sMRI signals predominantly from limbic areas, including the left amygdala and right hippocampus.

Conclusions

Our DBGAN model effectively predicts ECT responses in adolescent MDD patients using multi-modal MRI. Our method provides a potential application for personalized treatment of adolescent MDD.

目的重度抑郁症（MDD）是全球疾病负担的重要组成部分，电休克疗法（ECT）是一种有效但多变的治疗方法。本研究旨在利用多模态磁共振成像（MRI）和先进的深度学习技术，建立青少年MDD患者ECT治疗反应的个性化预测框架。方法选取27例青少年MDD患者，在治疗前后分别进行结构MRI （sMRI）和功能MRI （fMRI）检查。利用sMRI和fMRI数据分别构建了个体形态相似网络和功能连接网络。本文提出了一种新的双分支图注意网络（DBGAN），它将两个并行图注意网络集成在一起，以利用相似网络和连通性网络。该深度模型通过交叉注意机制动态融合sMRI和fMRI信息，提高了对ECT治疗反应的预测性能。结果27例患者中，21例对电痉挛治疗有积极反应。实验结果表明，我们的DBGAN优于传统机器学习模型和深度学习模型，平均准确率为0.853，精密度为0.920，召回率为0.910，f1得分为0.905。可解释性分析表明，预测决策主要受到右侧后岛和右侧扣带回背区的fMRI信号的影响，而sMRI信号主要来自边缘区域，包括左侧杏仁核和右侧海马。结论利用多模态MRI， sour DBGAN模型能有效预测青少年MDD患者的ECT反应。我们的方法为青少年重度抑郁症的个性化治疗提供了潜在的应用。

{"title":"Predicting efficacy of electroconvulsive therapy for adolescent major depressive disorder using a dual-branch graph attention network fusing multi-modal MRI","authors":"Jingyu Zhang , Ruiyun Zhu , Jing Min , Tong Zhu , Mengqi Liu , Renqiang Yu , Yang Huang , Chao Li , Sizhu Wu , Du Lei","doi":"10.1016/j.metrad.2025.100184","DOIUrl":"10.1016/j.metrad.2025.100184","url":null,"abstract":"<div><h3>Purpose</h3><div>Major Depressive Disorder (MDD) significantly contributes to global disease burden, and Electroconvulsive Therapy (ECT) is an effective yet variable treatment. This study aims to develop an individualized prediction framework for ECT treatment response in adolescent MDD patients using multi-modal magnetic resonance imaging (MRI) and advanced deep learning.</div></div><div><h3>Methods</h3><div>We recruited 27 adolescent MDD patients undergoing ECT, acquiring structural MRI (sMRI) and functional MRI (fMRI) before and after treatment. Individual morphological similarity networks and functional connectivity networks were created by utilizing sMRI and fMRI data, respectively. We introduced a novel Dual-Branch Graph Attention Network (DBGAN) which integrates two parallel graph attention networks for utilizing similarity and connectivity networks. The proposed deep model dynamically fuses information from sMRI and fMRI via a cross-attention mechanism, improving the prediction performance on ECT treatment response.</div></div><div><h3>Results</h3><div>Among 27 participants, 21 responded positively to ECT. According to experimental results, our DBGAN outperformed traditional machine learning models and deep learning models, achieving a mean accuracy of 0.853, precision of 0.920, recall of 0.910, and an F1-score of 0.905. Interpretability analyses indicated that predictive decisions were influenced by fMRI signals primarily in the right posterior insula and right dorsal cingulate gyrus, and sMRI signals predominantly from limbic areas, including the left amygdala and right hippocampus.</div></div><div><h3>Conclusions</h3><div>Our DBGAN model effectively predicts ECT responses in adolescent MDD patients using multi-modal MRI. Our method provides a potential application for personalized treatment of adolescent MDD.</div></div>","PeriodicalId":100921,"journal":{"name":"Meta-Radiology","volume":"3 4","pages":"Article 100184"},"PeriodicalIF":0.0,"publicationDate":"2025-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145579423","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}

引用次数: 0

Consistency-driven state-space model for incomplete multimodal MRI brain tumor segmentation 不完全多模态MRI脑肿瘤分割的一致性驱动状态空间模型

Meta-Radiology

Pub Date : 2025-10-09 DOI: 10.1016/j.metrad.2025.100185

Debao Liu , Xiaozhi Zhang , Hong Zhou , Kok Lay Teo

Brain tumor segmentation based on multimodal magnetic resonance imaging (MRI) plays a crucial role in clinical diagnosis and treatment planning. However, the absence or unavailability of certain modalities often hampers segmentation performance in real-world clinical settings. In this study, we propose a novel consistency-driven state-space model (SSM), which incorporates learnable consistency features into the SSM architecture to establish a robust framework for brain tumor segmentation under incomplete modality conditions. This approach introduces an innovative strategy for explicitly capturing cross-modal consistency within the context of efficient long-range dependency modeling, specifically designed for segmentation tasks involving missing modalities. Specifically, we design a scale-aware fusion block that integrates learnable consistency features to aggregate modality-specific information at an early stage. Subsequently, a Mamba-based multimodal consistency fusion technique is employed, enabling efficient long-range dependency modeling with linear computational complexity. To prevent modality bias, we further introduce a progressive attention weighting module that dynamically balances modality-specific features. Additionally, an adaptive feature correction mechanism is incorporated to refine both modality-specific and consistency features along both spatial and channel dimensions. The proposed method effectively facilitates modality integration while minimizing conflicts that may arise from directly fusing potentially inconsistent modalities. Comprehensive experiments conducted on the BraTS2018 and BraTS2020 datasets demonstrate that our model surpasses existing state-of-the-art approaches under various incomplete modality scenarios.

基于多模态磁共振成像（MRI）的脑肿瘤分割在临床诊断和治疗规划中起着至关重要的作用。然而，缺乏或不可用的某些模式往往阻碍分割性能在现实世界的临床设置。在这项研究中，我们提出了一种新的一致性驱动状态空间模型（SSM），该模型将可学习的一致性特征融入到SSM架构中，以建立一个不完全模态条件下脑肿瘤分割的鲁棒框架。该方法引入了一种创新策略，用于在高效的远程依赖建模的上下文中显式捕获跨模态一致性，该策略专门为涉及缺失模态的分割任务而设计。具体来说，我们设计了一个规模感知融合块，它集成了可学习的一致性特征，以便在早期阶段聚合特定于模态的信息。随后，采用基于mamba的多模态一致性融合技术，实现了具有线性计算复杂度的高效远程依赖关系建模。为了防止模态偏差，我们进一步引入了一个渐进的注意力加权模块，该模块可以动态平衡模态特定的功能。此外，还采用了自适应特征校正机制，以沿着空间和通道维度改进特定于模态和一致性的特征。提出的方法有效地促进了模态集成，同时最大限度地减少了直接融合潜在不一致模态可能产生的冲突。在BraTS2018和BraTS2020数据集上进行的综合实验表明，我们的模型在各种不完全模态场景下优于现有的最先进方法。

{"title":"Consistency-driven state-space model for incomplete multimodal MRI brain tumor segmentation","authors":"Debao Liu , Xiaozhi Zhang , Hong Zhou , Kok Lay Teo","doi":"10.1016/j.metrad.2025.100185","DOIUrl":"10.1016/j.metrad.2025.100185","url":null,"abstract":"<div><div>Brain tumor segmentation based on multimodal magnetic resonance imaging (MRI) plays a crucial role in clinical diagnosis and treatment planning. However, the absence or unavailability of certain modalities often hampers segmentation performance in real-world clinical settings. In this study, we propose a novel consistency-driven state-space model (SSM), which incorporates learnable consistency features into the SSM architecture to establish a robust framework for brain tumor segmentation under incomplete modality conditions. This approach introduces an innovative strategy for explicitly capturing cross-modal consistency within the context of efficient long-range dependency modeling, specifically designed for segmentation tasks involving missing modalities. Specifically, we design a scale-aware fusion block that integrates learnable consistency features to aggregate modality-specific information at an early stage. Subsequently, a Mamba-based multimodal consistency fusion technique is employed, enabling efficient long-range dependency modeling with linear computational complexity. To prevent modality bias, we further introduce a progressive attention weighting module that dynamically balances modality-specific features. Additionally, an adaptive feature correction mechanism is incorporated to refine both modality-specific and consistency features along both spatial and channel dimensions. The proposed method effectively facilitates modality integration while minimizing conflicts that may arise from directly fusing potentially inconsistent modalities. Comprehensive experiments conducted on the BraTS2018 and BraTS2020 datasets demonstrate that our model surpasses existing state-of-the-art approaches under various incomplete modality scenarios.</div></div>","PeriodicalId":100921,"journal":{"name":"Meta-Radiology","volume":"3 4","pages":"Article 100185"},"PeriodicalIF":0.0,"publicationDate":"2025-10-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145529067","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}

引用次数: 0

Transformer-based integration of radiomics and deep learning for differentiating lipid-poor adrenal adenomas from malignant tumors 基于转换器的放射组学和深度学习的整合用于区分低脂肾上腺腺瘤和恶性肿瘤

Meta-Radiology

Pub Date : 2025-10-01 DOI: 10.1016/j.metrad.2025.100183

Kai Zhao , Zhongqi Sun , Hao Jiang , Zhixuan Zou , Qiong Wu , Yanjie Xin , Xiangru Liu , Huijie Jiang

Purpose

To evaluate the effectiveness of a Transformer model based on contrast-enhanced computed tomography (CECT) that integrates radiomics and deep learning features in differentiating adrenal lipid-poor adenomas (LPA) and malignant tumors (MT).

Methods

This retrospective study included 282 patients with adrenal tumors from two medical centers between October 2018 and October 2024. The patients were classified into adrenal (LPA) and adrenal (MT) groups. Radiomics and deep learning features were extracted from CECT images. A total of 240 patients from the first center were randomly divided into Training Set and Test Set at a 7:3 ratio, while 42 patients from the second center served as an External Validation Set. A Transformer algorithm was employed to integrate radiomics and deep learning features for building predictive models. Its self-attention mechanism was utilized to capture intrinsic associations within each feature type and to uncover hidden information related to clinical outcomes. Additionally, a Radiomics model, a Deep Learning model (DL_model), and a Traditional Combined model integrating radiomics and deep learning features were constructed. Model performance was assessed using the area under the receiver operating characteristic (ROC) curve (AUC) and radar chart. Calibration curves and decision curve analysis (DCA) were employed to assess the predictive accuracy and clinical net benefit of the models. Furthermore, radiomics feature activation maps and gradient-weighted class activation mapping (Grad-CAM) were utilized to visualize radiomics and deep learning features, respectively.

Results

The Transformer model achieved the best predictive performance in the training, test, and external validation sets, with AUCs of 0.949, 0.917, and 0.852, respectively. The DeLong test indicated that the performance differences between this model and the other models were statistically significant. Furthermore, the radar chart illustrated that the Transformer model achieved superior overall performance, and DCA confirmed its higher clinical net benefit compared with the other models.

Conclusion

The Transformer model that integrates radiomics and deep learning features can accurately distinguish between LPA and MT. Furthermore, the visual analysis of radiomics feature activation maps and Grad-CAM intuitively illustrates the distribution of radiomics and deep learning features, enhancing their potential for clinical application in preoperative assessment of adrenal tumors.

目的评估基于对比增强计算机断层扫描（CECT）的Transformer模型在鉴别肾上腺脂质低下腺瘤（LPA）和恶性肿瘤（MT）中的有效性，该模型结合放射组学和深度学习特征。方法回顾性研究纳入2018年10月至2024年10月来自两家医疗中心的282例肾上腺肿瘤患者。将患者分为肾上腺素（LPA）组和肾上腺素（MT）组。从CECT图像中提取放射组学和深度学习特征。第一中心240例患者按7:3的比例随机分为Training Set和Test Set，第二中心42例患者作为External Validation Set。利用Transformer算法整合放射组学和深度学习特征，构建预测模型。它的自我注意机制被用来捕捉每个特征类型的内在联系，并揭示与临床结果相关的隐藏信息。此外，构建了放射组学模型、深度学习模型（DL_model）和传统放射组学与深度学习相结合的组合模型。利用接收机工作特性曲线（ROC）下面积和雷达图评估模型性能。采用校正曲线和决策曲线分析（DCA）来评估模型的预测准确性和临床净效益。此外，利用放射组学特征激活图和梯度加权类激活图（Grad-CAM）分别可视化放射组学和深度学习特征。结果Transformer模型在训练集、测试集和外部验证集的预测性能最佳，auc分别为0.949、0.917和0.852。DeLong检验表明，该模型与其他模型的性能差异具有统计学意义。此外，雷达图显示Transformer模型具有优越的综合性能，DCA证实其与其他模型相比具有更高的临床净效益。结论整合放射组学和深度学习特征的Transformer模型可以准确区分LPA和MT，并且通过放射组学特征激活图的可视化分析和Grad-CAM可以直观地说明放射组学和深度学习特征的分布，增强其在肾上腺肿瘤术前评估中的临床应用潜力。

{"title":"Transformer-based integration of radiomics and deep learning for differentiating lipid-poor adrenal adenomas from malignant tumors","authors":"Kai Zhao , Zhongqi Sun , Hao Jiang , Zhixuan Zou , Qiong Wu , Yanjie Xin , Xiangru Liu , Huijie Jiang","doi":"10.1016/j.metrad.2025.100183","DOIUrl":"10.1016/j.metrad.2025.100183","url":null,"abstract":"<div><h3>Purpose</h3><div>To evaluate the effectiveness of a Transformer model based on contrast-enhanced computed tomography (CECT) that integrates radiomics and deep learning features in differentiating adrenal lipid-poor adenomas (LPA) and malignant tumors (MT).</div></div><div><h3>Methods</h3><div>This retrospective study included 282 patients with adrenal tumors from two medical centers between October 2018 and October 2024. The patients were classified into adrenal (LPA) and adrenal (MT) groups. Radiomics and deep learning features were extracted from CECT images. A total of 240 patients from the first center were randomly divided into Training Set and Test Set at a 7:3 ratio, while 42 patients from the second center served as an External Validation Set. A Transformer algorithm was employed to integrate radiomics and deep learning features for building predictive models. Its self-attention mechanism was utilized to capture intrinsic associations within each feature type and to uncover hidden information related to clinical outcomes. Additionally, a Radiomics model, a Deep Learning model (DL_model), and a Traditional Combined model integrating radiomics and deep learning features were constructed. Model performance was assessed using the area under the receiver operating characteristic (ROC) curve (AUC) and radar chart. Calibration curves and decision curve analysis (DCA) were employed to assess the predictive accuracy and clinical net benefit of the models. Furthermore, radiomics feature activation maps and gradient-weighted class activation mapping (Grad-CAM) were utilized to visualize radiomics and deep learning features, respectively.</div></div><div><h3>Results</h3><div>The Transformer model achieved the best predictive performance in the training, test, and external validation sets, with AUCs of 0.949, 0.917, and 0.852, respectively. The DeLong test indicated that the performance differences between this model and the other models were statistically significant. Furthermore, the radar chart illustrated that the Transformer model achieved superior overall performance, and DCA confirmed its higher clinical net benefit compared with the other models.</div></div><div><h3>Conclusion</h3><div>The Transformer model that integrates radiomics and deep learning features can accurately distinguish between LPA and MT. Furthermore, the visual analysis of radiomics feature activation maps and Grad-CAM intuitively illustrates the distribution of radiomics and deep learning features, enhancing their potential for clinical application in preoperative assessment of adrenal tumors.</div></div>","PeriodicalId":100921,"journal":{"name":"Meta-Radiology","volume":"3 4","pages":"Article 100183"},"PeriodicalIF":0.0,"publicationDate":"2025-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145579424","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}

引用次数: 0

Frontal cortical gyrification mediates the association between hippocampal subfield microstructure and cognitive function in Parkinson's disease: A NODDI study highlighting olfactory dysfunction 前额皮质回旋介导帕金森病海马亚区微观结构和认知功能之间的关联：一项NODDI研究强调了嗅觉功能障碍

Meta-Radiology

Pub Date : 2025-09-15 DOI: 10.1016/j.metrad.2025.100182

Xin Liu , Jun Hu , Yule Zeng , Ping Yan , Ziling Huang , Liangwu Chen , Shicheng He , Yudie Xie , Beisha Tang , Weiyi Liu , Tianyao Wang , Hong Zhou

Background

Parkinson's disease (PD) is characterized by the progressive degeneration of dopaminergic neurons. Olfactory dysfunction and cognitive impairment are common and often co-occurring non-motor symptoms in PD. However, the neural mechanisms underlying their association remain unclear.

Methods

This cross-sectional study analyzed 72 patients with Parkinson's disease (PD). Based on the Hyposmia Rating Scale (HRS), they were classified into two groups: those with olfactory dysfunction (PD-OD) and those without (PD-NOD). Simple mediation models (PROCESS Model 4) were applied to both the overall PD group and the PD-OD subgroup to assess the potential mediating role of bilateral frontal cortical local gyrification index (LGI) values in the association between neurite orientation dispersion and density imaging (NODDI) parameters of the hippocampal subfields and cognitive performance.

Results

Microstructural alterations in the left subiculum and bilateral presubiculum were significantly associated with cognitive performance in PD patients. These associations were partially mediated by gyrification of the right pars opercularis, with significant indirect effects observed (c–c′ = −0.108, −0.121, and −0.126, respectively; all p < 0.05). The mediation effect was more prominent in PD-OD patients (c–c′ = −0.141, −0.138, and −0.132, respectively; all p < 0.05), indicating greater hippocampal–frontal structural disruption in this subgroup.

Conclusions

There is a significant structural association among hippocampal subfield microstructure, frontal cortical gyrification, and cognitive dysfunction in PD. Gyrification of the right pars opercularis plays a critical mediating role in the relationship between hippocampal microstructural alterations and cognitive impairment, with this mediation effect being more pronounced in PD-OD patients.

帕金森氏病（PD）以多巴胺能神经元进行性变性为特征。嗅觉功能障碍和认知障碍是PD患者常见的非运动症状。然而，它们之间联系的神经机制尚不清楚。方法对72例帕金森病（PD）患者进行横断面研究。根据嗅觉功能减退评定量表（HRS）将患者分为嗅觉功能障碍组（PD-OD）和无嗅觉功能障碍组（PD-NOD）。将简单中介模型（PROCESS Model 4）应用于PD整体组和PD- od亚组，评估双侧额叶皮质局部回化指数（LGI）值在海马亚区神经突定向分散和密度成像（NODDI）参数与认知表现之间的关联中的潜在中介作用。结果PD患者左肩带下和双侧肩带前的微结构改变与认知能力显著相关。这些关联部分由右侧包部的旋转介导，观察到显著的间接影响（c-c′分别= - 0.108,- 0.121和- 0.126；均p <； 0.05）。调解作用在PD-OD患者中更为突出（c-c′分别= - 0.141、- 0.138和- 0.132；均p <； 0.05），表明该亚组海马额叶结构破坏更严重。结论PD患者海马亚区微结构、额叶皮质回缩与认知功能障碍存在显著的结构关联。右侧包部旋回在海马微结构改变与认知功能障碍的关系中起着重要的中介作用，这种中介作用在PD-OD患者中更为明显。

{"title":"Frontal cortical gyrification mediates the association between hippocampal subfield microstructure and cognitive function in Parkinson's disease: A NODDI study highlighting olfactory dysfunction","authors":"Xin Liu , Jun Hu , Yule Zeng , Ping Yan , Ziling Huang , Liangwu Chen , Shicheng He , Yudie Xie , Beisha Tang , Weiyi Liu , Tianyao Wang , Hong Zhou","doi":"10.1016/j.metrad.2025.100182","DOIUrl":"10.1016/j.metrad.2025.100182","url":null,"abstract":"<div><h3>Background</h3><div>Parkinson's disease (PD) is characterized by the progressive degeneration of dopaminergic neurons. Olfactory dysfunction and cognitive impairment are common and often co-occurring non-motor symptoms in PD. However, the neural mechanisms underlying their association remain unclear.</div></div><div><h3>Methods</h3><div>This cross-sectional study analyzed 72 patients with Parkinson's disease (PD). Based on the Hyposmia Rating Scale (HRS), they were classified into two groups: those with olfactory dysfunction (PD-OD) and those without (PD-NOD). Simple mediation models (PROCESS Model 4) were applied to both the overall PD group and the PD-OD subgroup to assess the potential mediating role of bilateral frontal cortical local gyrification index (LGI) values in the association between neurite orientation dispersion and density imaging (NODDI) parameters of the hippocampal subfields and cognitive performance.</div></div><div><h3>Results</h3><div>Microstructural alterations in the left subiculum and bilateral presubiculum were significantly associated with cognitive performance in PD patients. These associations were partially mediated by gyrification of the right pars opercularis, with significant indirect effects observed (c–c′ = −0.108, −0.121, and −0.126, respectively; all <em>p</em> < 0.05). The mediation effect was more prominent in PD-OD patients (c–c′ = −0.141, −0.138, and −0.132, respectively; all <em>p</em> < 0.05), indicating greater hippocampal–frontal structural disruption in this subgroup.</div></div><div><h3>Conclusions</h3><div>There is a significant structural association among hippocampal subfield microstructure, frontal cortical gyrification, and cognitive dysfunction in PD. Gyrification of the right pars opercularis plays a critical mediating role in the relationship between hippocampal microstructural alterations and cognitive impairment, with this mediation effect being more pronounced in PD-OD patients.</div></div>","PeriodicalId":100921,"journal":{"name":"Meta-Radiology","volume":"3 4","pages":"Article 100182"},"PeriodicalIF":0.0,"publicationDate":"2025-09-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145579425","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}

引用次数: 0

Radiomic analysis to predict arrhythmias in athletes by echocardiography: Artificial intelligence vs. formal methods 通过超声心动图预测运动员心律失常的放射组学分析：人工智能与正式方法

Meta-Radiology

Pub Date : 2025-09-02 DOI: 10.1016/j.metrad.2025.100174

Giulia Varriano , Ester Lagonigro , Giuseppe Prisco, Marialucia Spina, Klara Komici, Germano Guerra, Antonella Santone

The diagnosis and management of premature ventricular beats (PVBs) in athletes remain challenging due to the potential for underlying cardiac pathology. While the electrocardiogram is essential for detecting electrical abnormalities, echocardiography is crucial for evaluating structural heart disease. This study explores the use of radiomics analysis applied to apical 4-chamber echocardiography to automatically and preemptively identify PVB risk, better characterize athlete cardiac remodeling, and enable early detection of pathological changes.We evaluated and compared the limitations and potential of Artificial Intelligence (AI) and Formal Methods (FM) for this task. Using data from 723 athletes, we processed echocardiography videos and extracted over 100 features per athlete to develop robust classifiers.Our findings demonstrate that radiomics can power automated decision-support systems for diagnosis. While AI-based models offer powerful predictive capabilities, FM presents a compelling, mathematically rigorous, and explainable alternative. Both modalities are useful for the early detection of structural substrates underlying PVBs. The collaboration between physicians and computer scientists is crucial to unlocking new frontiers in radiomics and advancing personalized medicine.

由于潜在的心脏病理，运动员室性早搏（pvb）的诊断和管理仍然具有挑战性。虽然心电图对检测电异常至关重要，但超声心动图对评估结构性心脏病至关重要。本研究探索将放射组学分析应用于根尖4室超声心动图，以自动和先发制人地识别PVB风险，更好地表征运动员心脏重构，并能够早期发现病理变化。我们评估并比较了人工智能（AI）和形式化方法（FM）在这项任务中的局限性和潜力。使用来自723名运动员的数据，我们处理了超声心动图视频，并提取了每个运动员超过100个特征，以开发健壮的分类器。我们的研究结果表明，放射组学可以为诊断的自动决策支持系统提供动力。虽然基于人工智能的模型提供了强大的预测能力，但FM提供了一个引人注目的、数学上严谨的、可解释的替代方案。这两种方式都有助于早期检测PVBs下的结构基底。医生和计算机科学家之间的合作对于打开放射组学的新领域和推进个性化医疗至关重要。

{"title":"Radiomic analysis to predict arrhythmias in athletes by echocardiography: Artificial intelligence vs. formal methods","authors":"Giulia Varriano , Ester Lagonigro , Giuseppe Prisco, Marialucia Spina, Klara Komici, Germano Guerra, Antonella Santone","doi":"10.1016/j.metrad.2025.100174","DOIUrl":"10.1016/j.metrad.2025.100174","url":null,"abstract":"<div><div>The diagnosis and management of premature ventricular beats (PVBs) in athletes remain challenging due to the potential for underlying cardiac pathology. While the electrocardiogram is essential for detecting electrical abnormalities, echocardiography is crucial for evaluating structural heart disease. This study explores the use of radiomics analysis applied to apical 4-chamber echocardiography to automatically and preemptively identify PVB risk, better characterize athlete cardiac remodeling, and enable early detection of pathological changes.We evaluated and compared the limitations and potential of Artificial Intelligence (AI) and Formal Methods (FM) for this task. Using data from 723 athletes, we processed echocardiography videos and extracted over 100 features per athlete to develop robust classifiers.Our findings demonstrate that radiomics can power automated decision-support systems for diagnosis. While AI-based models offer powerful predictive capabilities, FM presents a compelling, mathematically rigorous, and explainable alternative. Both modalities are useful for the early detection of structural substrates underlying PVBs. The collaboration between physicians and computer scientists is crucial to unlocking new frontiers in radiomics and advancing personalized medicine.</div></div>","PeriodicalId":100921,"journal":{"name":"Meta-Radiology","volume":"3 4","pages":"Article 100174"},"PeriodicalIF":0.0,"publicationDate":"2025-09-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145425575","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}

引用次数: 0

Advances in neuroimaging applications of quantitative susceptibility mapping 定量易感图谱在神经影像学中的应用进展

Meta-Radiology

Pub Date : 2025-09-01 DOI: 10.1016/j.metrad.2025.100148

Shuxin Ma , Wencan Fu , Chao Chai , Huiying Wang , Ke Lv , Chenxi Zhao , E. Mark Haacke , Sagar Buch , Shuang Xia

This review article delves into the advancements of quantitative susceptibility mapping (QSM) in neuroimaging, highlighting its utility in detecting and quantifying magnetic susceptibility differences in tissues, particularly for paramagnetic substances like iron and diamagnetic substances such as calcifications in the brain. QSM has revolutionized the diagnosis and monitoring of neurodegenerative diseases by enabling the precise measurement of brain iron deposition and blood oxygen saturation. The review is partitioned into three sections. The first section underscores QSM's role in clinical applications related to microhemorrhages, cerebral amyloidosis, intracranial hematomas, and cerebrovascular malformations. The second section focuses on QSM's application in mapping iron content in neurodegenerative diseases such as Parkinson's disease and Alzheimer's disease. The final section discusses QSM's potential in assessing stroke by measuring oxygen saturation. The article also outlines the basic theory and development of QSM, emphasizing the importance of echo time selection for accurate QSM results. Challenges in clinical applications and future directions, including the integration of AI technology for image reconstruction and data analysis, are also discussed. QSM's ability to differentiate between microbleeds and calcifications, assess dynamic susceptibility changes in intracranial hematomas, and guide thrombolytic strategies in acute cerebrovascular disease is highlighted. The review concludes by emphasizing the need for further optimization of QSM algorithms and the expansion of its applications in biomedical imaging.

本文综述了定量易感性制图（QSM）在神经影像学中的研究进展，重点介绍了其在检测和定量组织磁化率差异方面的应用，特别是对铁等顺磁性物质和脑钙化等反磁性物质。QSM通过精确测量脑铁沉积和血氧饱和度，彻底改变了神经退行性疾病的诊断和监测。这篇综述分为三个部分。第一部分强调了QSM在与微出血、脑淀粉样变性、颅内血肿和脑血管畸形相关的临床应用中的作用。第二部分重点介绍QSM在帕金森病和阿尔茨海默病等神经退行性疾病中铁含量制图中的应用。最后一节讨论了通过测量血氧饱和度来评估脑卒中的QSM的潜力。本文还概述了QSM的基本理论和发展，强调了回波时间选择对准确的QSM结果的重要性。还讨论了临床应用中的挑战和未来方向，包括将人工智能技术集成到图像重建和数据分析中。QSM能够区分微出血和钙化，评估颅内血肿的动态易感性变化，并指导急性脑血管疾病的溶栓策略。最后强调了QSM算法的进一步优化和在生物医学成像中的应用。

{"title":"Advances in neuroimaging applications of quantitative susceptibility mapping","authors":"Shuxin Ma , Wencan Fu , Chao Chai , Huiying Wang , Ke Lv , Chenxi Zhao , E. Mark Haacke , Sagar Buch , Shuang Xia","doi":"10.1016/j.metrad.2025.100148","DOIUrl":"10.1016/j.metrad.2025.100148","url":null,"abstract":"<div><div>This review article delves into the advancements of quantitative susceptibility mapping (QSM) in neuroimaging, highlighting its utility in detecting and quantifying magnetic susceptibility differences in tissues, particularly for paramagnetic substances like iron and diamagnetic substances such as calcifications in the brain. QSM has revolutionized the diagnosis and monitoring of neurodegenerative diseases by enabling the precise measurement of brain iron deposition and blood oxygen saturation. The review is partitioned into three sections. The first section underscores QSM's role in clinical applications related to microhemorrhages, cerebral amyloidosis, intracranial hematomas, and cerebrovascular malformations. The second section focuses on QSM's application in mapping iron content in neurodegenerative diseases such as Parkinson's disease and Alzheimer's disease. The final section discusses QSM's potential in assessing stroke by measuring oxygen saturation. The article also outlines the basic theory and development of QSM, emphasizing the importance of echo time selection for accurate QSM results. Challenges in clinical applications and future directions, including the integration of AI technology for image reconstruction and data analysis, are also discussed. QSM's ability to differentiate between microbleeds and calcifications, assess dynamic susceptibility changes in intracranial hematomas, and guide thrombolytic strategies in acute cerebrovascular disease is highlighted. The review concludes by emphasizing the need for further optimization of QSM algorithms and the expansion of its applications in biomedical imaging.</div></div>","PeriodicalId":100921,"journal":{"name":"Meta-Radiology","volume":"3 3","pages":"Article 100148"},"PeriodicalIF":0.0,"publicationDate":"2025-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144922991","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}

引用次数: 0

Cardiac ECV mapping: Underlying concepts and clinical applications 心脏ECV制图：基本概念和临床应用

Meta-Radiology

Pub Date : 2025-09-01 DOI: 10.1016/j.metrad.2025.100168

Guo-Jun Zhu , Simran Qureshi , Ward Hedges , Chong-Wen Wu , Lian-Ming Wu

Myocardial Extracellular volume fraction (ECV) mapping, based on cardiac magnetic resonance (CMR), is a crucial technique for assessing myocardial histological changes by evaluating extracellular matrix expansion. In recent years, cardiac ECV mapping has gained significant attention in both basic research and clinical settings, particularly for its role in myocardial fibrosis and other cardiovascular diseases. This review explores the measurement of ECV mapping, its diagnostic and prognostic value in ischemic and non-ischemic cardiovascular diseases, and emerging advancements in ECV mapping techniques.

心肌细胞外体积分数（ECV）制图，基于心脏磁共振（CMR），是通过评估细胞外基质扩张来评估心肌组织学变化的关键技术。近年来，心脏ECV制图在基础研究和临床环境中都受到了极大的关注，特别是其在心肌纤维化和其他心血管疾病中的作用。本文综述了ECV作图的测量方法，在缺血性和非缺血性心血管疾病中的诊断和预后价值，以及ECV作图技术的新进展。

引用次数: 0

2-[18F]FDG PET/CT identifies metabolic activity linked to myocardial deformation in hypertrophic cardiomyopathy 2-[18F]FDG PET/CT鉴定肥厚性心肌病患者与心肌变形相关的代谢活动

Meta-Radiology

Pub Date : 2025-08-26 DOI: 10.1016/j.metrad.2025.100173

Patrícia Marques-Alves , Maria João Ferreira , Rodolfo Silva , João Borges-Rosa , Andreia Gomes , Antero Abrunhosa , Miguel Castelo-Branco , Wael Jaber , Lino Gonçalves

Purpose

Hypertrophic cardiomyopathy (HCM) is a highly heterogeneous disease, which makes prognostic assessment challenging. We aim to explore the relationship between myocardial metabolism, evaluated via 2-[18F]FDG PET/CT, and myocardial deformation, through strain analysis, in HCM patients.

Methods

We included 30 patients with non-obstructive HCM who underwent echocardiography with left ventricular (LV-GLS), left atrial (LA) strain and LV myocardial work (MW) analysis. Myocardial metabolism through 2-[¹⁸F]FDG PET/CT was evaluated using dedicated software. Cardiac magnetic resonance was also performed, evaluating fibrosis with late gadolinium enhancement (LGE). After a follow-up period of 2 ± 0.5 years, a subsequent echocardiography was performed.

Results

Mean age was 54 years and 57 % patients were male. Mean myocardial maximal wall thickness (MMWT) was 20.3 ± 4.4 mm and correlated with impaired myocardial deformation (LA and LV strain, LV-MW indexes). Increased 2-[¹⁸F]FDG uptake was present in 53 % patients. Hypermetabolism extension was higher in patients with higher MMWT (18.4 ± 19.7 % vs 3.07 ± 7.1 %, p = 0.013; rho = 0.46, p = 0.012), impaired LV-GLS (rho 0.37, p = 0.042) and impaired MW indexes (rho = −0.41, p = 0.034). Hypermetabolism was also correlated with LGE (rho 0.47, p = 0.028). At follow-up, there was a significant decrease in LV ejection fraction (61.2 ± 4.5 % to 55.1 ± 6.7 %, p = 0.014), anterior myocardial strain (−13.2 ± 4.4 to −9.5 ± 4.4, p = 0.035) and LA strain. This was particularly noted in patients that presented with 2-[¹⁸F]FDG uptake and also LGE.

Conclusions

Our study highlights key correlations between 2-[¹⁸F]FDG uptake, myocardial hypertrophy, and impaired deformation. These findings suggest a progression from hypermetabolism to fibrosis and deformation impairment, potentially driving functional decline in HCM.

目的肥厚性心肌病（HCM）是一种高度异质性的疾病，这使得预后评估具有挑战性。我们旨在探讨心肌代谢（通过2-[18F]FDG PET/CT评估）与心肌变形（通过应变分析）在HCM患者中的关系。方法30例非阻塞性HCM患者行超声心动图左室（LV- gls）、左房（LA）应变和左室心肌功（MW）分析。使用专用软件通过2-[18F]FDG PET/CT评估心肌代谢。同时进行心脏磁共振，用晚期钆增强（LGE）评估纤维化。随访2±0.5年后，进行超声心动图检查。结果患者平均年龄54岁，男性占57%。平均心肌最大壁厚（MMWT）为20.3±4.4 mm，与心肌变形受损（左室、左室应变、LV- mw指数）相关。53%的患者出现2-[18F]FDG摄取增加。MMWT高（18.4±19.7% vs 3.07±7.1%,p = 0.013; rho = 0.46, p = 0.012）、LV-GLS受损（rho = 0.37, p = 0.042）和MW指数受损（rho = - 0.41, p = 0.034）患者的高代谢延伸程度更高。高代谢也与LGE相关（rho 0.47, p = 0.028）。随访时左室射血分数（61.2±4.5%至55.1±6.7%,p = 0.014）、前心肌应变（- 13.2±4.4至- 9.5±4.4,p = 0.035）和左室应变均显著降低。这在出现2-[18F]FDG摄取和LGE的患者中尤为明显。结论我们的研究强调了2-[18F]FDG摄取与心肌肥大和变形受损之间的关键相关性。这些发现提示从高代谢到纤维化和变形损伤的进展，可能导致HCM功能下降。

{"title":"2-[18F]FDG PET/CT identifies metabolic activity linked to myocardial deformation in hypertrophic cardiomyopathy","authors":"Patrícia Marques-Alves , Maria João Ferreira , Rodolfo Silva , João Borges-Rosa , Andreia Gomes , Antero Abrunhosa , Miguel Castelo-Branco , Wael Jaber , Lino Gonçalves","doi":"10.1016/j.metrad.2025.100173","DOIUrl":"10.1016/j.metrad.2025.100173","url":null,"abstract":"<div><h3>Purpose</h3><div>Hypertrophic cardiomyopathy (HCM) is a highly heterogeneous disease, which makes prognostic assessment challenging. We aim to explore the relationship between myocardial metabolism, evaluated via 2-[18F]FDG PET/CT, and myocardial deformation, through strain analysis, in HCM patients.</div></div><div><h3>Methods</h3><div>We included 30 patients with non-obstructive HCM who underwent echocardiography with left ventricular (LV-GLS), left atrial (LA) strain and LV myocardial work (MW) analysis. Myocardial metabolism through 2-[<sup>18</sup>F]FDG PET/CT was evaluated using dedicated software. Cardiac magnetic resonance was also performed, evaluating fibrosis with late gadolinium enhancement (LGE). After a follow-up period of 2 ± 0.5 years, a subsequent echocardiography was performed.</div></div><div><h3>Results</h3><div>Mean age was 54 years and 57 % patients were male. Mean myocardial maximal wall thickness (MMWT) was 20.3 ± 4.4 mm and correlated with impaired myocardial deformation (LA and LV strain, LV-MW indexes). Increased 2-[<sup>18</sup>F]FDG uptake was present in 53 % patients. Hypermetabolism extension was higher in patients with higher MMWT (18.4 ± 19.7 % vs 3.07 ± 7.1 %, p = 0.013; rho = 0.46, p = 0.012), impaired LV-GLS (rho 0.37, p = 0.042) and impaired MW indexes (rho = −0.41, p = 0.034). Hypermetabolism was also correlated with LGE (rho 0.47, p = 0.028). At follow-up, there was a significant decrease in LV ejection fraction (61.2 ± 4.5 % to 55.1 ± 6.7 %, p = 0.014), anterior myocardial strain (−13.2 ± 4.4 to −9.5 ± 4.4, p = 0.035) and LA strain. This was particularly noted in patients that presented with 2-[<sup>18</sup>F]FDG uptake and also LGE.</div></div><div><h3>Conclusions</h3><div>Our study highlights key correlations between 2-[<sup>18</sup>F]FDG uptake, myocardial hypertrophy, and impaired deformation. These findings suggest a progression from hypermetabolism to fibrosis and deformation impairment, potentially driving functional decline in HCM.</div></div>","PeriodicalId":100921,"journal":{"name":"Meta-Radiology","volume":"3 4","pages":"Article 100173"},"PeriodicalIF":0.0,"publicationDate":"2025-08-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145398607","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}

引用次数: 0

Intelligent analysis of chest X-ray based on multi-modal instruction tuning 基于多模态指令调谐的胸部x射线智能分析

Meta-Radiology

Pub Date : 2025-08-19 DOI: 10.1016/j.metrad.2025.100172

Junjie Yao , Junhao Wang , Zhenxiang Xiao , Xinlin Hao , Xi Jiang

Chest X-ray plays a crucial role in the screening and diagnosis of chest diseases. Due to the complexity of pathological manifestations and limitations of radiologists' experience, the accuracy and efficiency of diagnosing chest diseases need to be further improved. In recent years, deep learning has made significant progress in chest X-ray image analysis, while existing methods mainly rely on uni-modal visual information, overlooking the prior knowledge related to disease category descriptions embedded in medical text data, making it challenging to fully understand the deep semantics of chest X-ray images. To address these challenges, inspired by the Instruction-ViT model, we adopt instruction tuning techniques to integrate medical textual information into the fine-tuning process of the visual model. Furthermore, a contrastive learning loss is employed to align textual and visual features, thereby enhancing the model's capacity to understand and differentiate complex pathological patterns. Experimental results demonstrate that the model integrating medical text information outperforms uni-modal models in various evaluation metrics, confirming that with instruction tuning, our model can effectively utilize medical text as prior knowledge to improve the performance of visual models in chest disease diagnosis. Furthermore, we conduct an interpretability analysis of the model's decision-making process, revealing that the regions attended to by the model highly correspond to the radiographic manifestations of different diseases, demonstrating the model's interpretability to a certain degree.

胸部x线在胸部疾病的筛查和诊断中起着至关重要的作用。由于病理表现的复杂性和放射科医师经验的局限性，胸部疾病诊断的准确性和效率有待进一步提高。近年来，深度学习在胸部x线图像分析方面取得了重大进展，但现有方法主要依赖于单模态视觉信息，忽略了医学文本数据中嵌入的疾病类别描述相关的先验知识，难以充分理解胸部x线图像的深度语义。为了解决这些挑战，受instruction - vit模型的启发，我们采用指令调优技术将医学文本信息集成到视觉模型的微调过程中。此外，使用对比学习损失来对齐文本和视觉特征，从而增强模型理解和区分复杂病理模式的能力。实验结果表明，集成医学文本信息的模型在各种评价指标上优于单模态模型，验证了通过指令调优，我们的模型可以有效地利用医学文本作为先验知识来提高视觉模型在胸部疾病诊断中的性能。进一步，我们对模型的决策过程进行了可解释性分析，发现模型所关注的区域与不同疾病的影像学表现高度对应，说明模型具有一定的可解释性。

{"title":"Intelligent analysis of chest X-ray based on multi-modal instruction tuning","authors":"Junjie Yao , Junhao Wang , Zhenxiang Xiao , Xinlin Hao , Xi Jiang","doi":"10.1016/j.metrad.2025.100172","DOIUrl":"10.1016/j.metrad.2025.100172","url":null,"abstract":"<div><div>Chest X-ray plays a crucial role in the screening and diagnosis of chest diseases. Due to the complexity of pathological manifestations and limitations of radiologists' experience, the accuracy and efficiency of diagnosing chest diseases need to be further improved. In recent years, deep learning has made significant progress in chest X-ray image analysis, while existing methods mainly rely on uni-modal visual information, overlooking the prior knowledge related to disease category descriptions embedded in medical text data, making it challenging to fully understand the deep semantics of chest X-ray images. To address these challenges, inspired by the Instruction-ViT model, we adopt instruction tuning techniques to integrate medical textual information into the fine-tuning process of the visual model. Furthermore, a contrastive learning loss is employed to align textual and visual features, thereby enhancing the model's capacity to understand and differentiate complex pathological patterns. Experimental results demonstrate that the model integrating medical text information outperforms uni-modal models in various evaluation metrics, confirming that with instruction tuning, our model can effectively utilize medical text as prior knowledge to improve the performance of visual models in chest disease diagnosis. Furthermore, we conduct an interpretability analysis of the model's decision-making process, revealing that the regions attended to by the model highly correspond to the radiographic manifestations of different diseases, demonstrating the model's interpretability to a certain degree.</div></div>","PeriodicalId":100921,"journal":{"name":"Meta-Radiology","volume":"3 3","pages":"Article 100172"},"PeriodicalIF":0.0,"publicationDate":"2025-08-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144904346","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}

引用次数: 0