Computer methods and programs in biomedicine最新文献_第10页

GICAF-Net: A cross-attentional graph-image fusion network for hyperspectral pathological diagnosis of FNH and HCC GICAF-Net：用于FNH和HCC高光谱病理诊断的交叉注意图形图像融合网络。

IF 4.8 2区医学 Q1 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Computer methods and programs in biomedicine

Pub Date : 2026-02-01 Epub Date: 2025-11-19 DOI: 10.1016/j.cmpb.2025.109171

Yunze Li , Haiyan Chen , Baoxian Gong , Jiankang Han , Jun Cheng , Shuai Gao , Wei Li

Background and objective

Accurate intraoperative differentiation between focal nodular hyperplasia (FNH) and hepatocellular carcinoma (HCC) remains a major clinical challenge, especially in atypical cases where conventional imaging and histopathology are constrained by turnaround time, cost, or spectral resolution. This study aims to develop a novel deep learning framework to improve the precision and efficiency of hyperspectral pathological diagnosis for liver tumors.

Methods

We propose GICAF-Net, a Graph–Image Cross-Attentional Fusion Network, designed to leverage hyperspectral imaging (HSI) for capturing fine-grained spatial–spectral information. The network employs a dual-branch architecture: (1) a residual convolutional branch for extracting pseudo-color image features, and (2) a residual graph convolutional branch for modeling topological spatial–spectral features. A Topology-Aware Cross-Attention Fusion (TACA) module enables bidirectional information exchange between the two modalities, while a multi-constraint fusion loss—combining cross-entropy, prediction confidence, and cross-modal attention consistency—enhances classification stability. A balanced hyperspectral liver tumor dataset consisting of 60 HCC and 60 FNH cases was constructed and evaluated using ten-fold cross-validation.

Results

GICAF-Net achieved an AUC of 0.9571 ± 0.0068, accuracy of 88.34 % ± 1.10 %, and F1-score of 88.32 % ± 1.11 %, outperforming state-of-the-art baseline models. Ablation experiments further validated the contributions of both the TACA module and the multi-constraint loss function in enhancing cross-modal fusion and improving classification performance.

Conclusion

The integration of graph-based spectral–structural modeling with deep visual features through cross-attention provides a powerful approach for hyperspectral pathological diagnosis. The proposed GICAF-Net demonstrates strong potential for rapid, accurate, and minimally invasive intraoperative differentiation of FNH and HCC, offering valuable clinical support in liver tumor surgery.

背景和目的：术中准确区分局灶性结节增生（FNH）和肝细胞癌（HCC）仍然是一个主要的临床挑战，特别是在常规影像学和组织病理学受周转时间、成本或光谱分辨率限制的非典型病例中。本研究旨在开发一种新的深度学习框架，以提高肝脏肿瘤高光谱病理诊断的准确性和效率。方法：我们提出了图形图像交叉注意融合网络（GICAF-Net），旨在利用高光谱成像（HSI）捕获细粒度的空间光谱信息。该网络采用双分支架构：(1)残差卷积分支用于提取伪彩色图像特征，(2)残差图卷积分支用于建模拓扑空间光谱特征。拓扑感知交叉注意融合（TACA）模块实现了两种模式之间的双向信息交换，而多约束融合损失（结合交叉熵、预测置信度和交叉模态注意一致性）增强了分类稳定性。构建了一个平衡的高光谱肝脏肿瘤数据集，包括60例HCC和60例FNH病例，并使用十倍交叉验证进行评估。结果：GICAF-Net的AUC为0.9571±0.0068，准确率为88.34%±1.10%，f1评分为88.32%±1.11%，优于目前最先进的基线模型。烧蚀实验进一步验证了TACA模块和多约束损失函数在增强跨模态融合和提高分类性能方面的贡献。结论：通过交叉注意将基于图的光谱结构建模与深度视觉特征相结合，为高光谱病理诊断提供了一种强有力的方法。GICAF-Net在快速、准确、微创的术中分化FNH和HCC方面显示出强大的潜力，为肝肿瘤手术提供了宝贵的临床支持。

{"title":"GICAF-Net: A cross-attentional graph-image fusion network for hyperspectral pathological diagnosis of FNH and HCC","authors":"Yunze Li , Haiyan Chen , Baoxian Gong , Jiankang Han , Jun Cheng , Shuai Gao , Wei Li","doi":"10.1016/j.cmpb.2025.109171","DOIUrl":"10.1016/j.cmpb.2025.109171","url":null,"abstract":"<div><h3>Background and objective</h3><div>Accurate intraoperative differentiation between focal nodular hyperplasia (FNH) and hepatocellular carcinoma (HCC) remains a major clinical challenge, especially in atypical cases where conventional imaging and histopathology are constrained by turnaround time, cost, or spectral resolution. This study aims to develop a novel deep learning framework to improve the precision and efficiency of hyperspectral pathological diagnosis for liver tumors.</div></div><div><h3>Methods</h3><div>We propose GICAF-Net, a Graph–Image Cross-Attentional Fusion Network, designed to leverage hyperspectral imaging (HSI) for capturing fine-grained spatial–spectral information. The network employs a dual-branch architecture: (1) a residual convolutional branch for extracting pseudo-color image features, and (2) a residual graph convolutional branch for modeling topological spatial–spectral features. A Topology-Aware Cross-Attention Fusion (TACA) module enables bidirectional information exchange between the two modalities, while a multi-constraint fusion loss—combining cross-entropy, prediction confidence, and cross-modal attention consistency—enhances classification stability. A balanced hyperspectral liver tumor dataset consisting of 60 HCC and 60 FNH cases was constructed and evaluated using ten-fold cross-validation.</div></div><div><h3>Results</h3><div>GICAF-Net achieved an AUC of 0.9571 ± 0.0068, accuracy of 88.34 % ± 1.10 %, and F1-score of 88.32 % ± 1.11 %, outperforming state-of-the-art baseline models. Ablation experiments further validated the contributions of both the TACA module and the multi-constraint loss function in enhancing cross-modal fusion and improving classification performance.</div></div><div><h3>Conclusion</h3><div>The integration of graph-based spectral–structural modeling with deep visual features through cross-attention provides a powerful approach for hyperspectral pathological diagnosis. The proposed GICAF-Net demonstrates strong potential for rapid, accurate, and minimally invasive intraoperative differentiation of FNH and HCC, offering valuable clinical support in liver tumor surgery.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"274 ","pages":"Article 109171"},"PeriodicalIF":4.8,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145602891","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

CT-free attenuation correction of 13N-ammonia cardiac PET images using conditional denoising diffusion implicit model with logarithmic linear normalization 采用对数线性归一化条件去噪扩散隐式模型对13n -氨心脏PET图像进行无ct衰减校正。

IF 4.8 2区医学 Q1 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Computer methods and programs in biomedicine

Pub Date : 2026-02-01 Epub Date: 2025-12-01 DOI: 10.1016/j.cmpb.2025.109188

Hao Sun , Xiaotong Hong , Yijun Lu , Fanghu Wang , Amirhossein Sanaat , Weiping Xu , Shuxia Wang , Habib Zaidi , Lijun Lu

Background and objective

CT-based attenuation correction (CT-AC) is commonly used in cardiac PET but introduces additional radiation exposure, which is particularly problematic for pediatrics or repeat scans. This study aims to generate attenuation-corrected cardiac PET images from non-attenuation-corrected (NAC) PET input without CT using a conditional denoising diffusion implicit model (cDDIM).

Methods

The patient cohort included 60 rest-only and 14 rest-stress scans from Center 1, and 30 rest-stress scans from Center 2. All subjects underwent ¹³N-ammonia cardiac PET/CT scans, generating paired 3D NAC and CT-AC PET images. The middle slice and two adjacent axial slices were used as inputs to mitigate axial artifacts, forming a 2.5D cDDIM framework. Two versions were proposed: one with data logarithmic linear normalization (cDDIM_LLN) and another with linear normalization (cDDIM_LN). 2.5D and 3D generative adversarial network (GAN)-based AC methods were implemented for comparison. Model performances were evaluated using normalized mean square error (NMSE) and segment-wise absolute percent error (APE).

Results

Visual analysis indicated that cDDIM_LLN and cDDIM_LN outperformed GAN-based AC methods for both rest and stress cardiac PET images. cDDIM_LLN achieved significantly lower NMSE than cDDIM_LN in Center 1 (1.87 ± 1.25 % vs. 2.77 ± 2.37 %, p < 0.001) and Center 2 (4.63 ± 3.71 % vs. 5.67 ± 7.88 %, p < 0.001). cDDIM_LLN also showed smaller APE than other methods for both centers (Center 1: 5.78 ± 1.43 %, p < 0.001; Center 2: 9.10 ± 5.59 %, p < 0.001). cDDIM_LN demonstrated lower APE than GAN-based AC methods across both centers.

Conclusions

cDDIM-based AC methods synthesized tracer distributions highly similar to clinical CT-AC. Among the evaluated methods, cDDIM_LLN demonstrated the best performance. Overall, cDDIM-based AC outperformed traditional GAN-based methods.

背景和目的：基于ct的衰减校正（CT-AC）通常用于心脏PET，但会引入额外的辐射暴露，这对儿科或重复扫描尤其有问题。本研究旨在使用条件去噪扩散隐式模型（cDDIM）从无CT的非衰减校正（NAC） PET输入生成衰减校正的心脏PET图像。方法：患者队列包括来自中心1的60例静息扫描和14例静息应激扫描，以及来自中心2的30例静息应激扫描。所有受试者进行13n -氨氮心脏PET/CT扫描，生成配对的3D NAC和CT- ac PET图像。中间切片和两个相邻的轴向切片作为输入，以减轻轴向伪影，形成2.5D cDDIM框架。提出了数据对数线性归一化（cDDIM_LLN）和线性归一化（cDDIM_LN）两个版本。采用基于2.5D和3D生成对抗网络（GAN）的AC方法进行比较。使用归一化均方误差（NMSE）和分段绝对百分比误差（APE）评估模型性能。结果：视觉分析表明cDDIM_LLN和cDDIM_LN在休息和应激心脏PET图像上优于基于gan的AC方法。cDDIM_LLN在中心1（1.87±1.25%比2.77±2.37%,p < 0.001）和中心2（4.63±3.71%比5.67±7.88%,p < 0.001）的NMSE显著低于cDDIM_LN。cDDIM_LLN在两个中心的APE均小于其他方法（中心1:5.78±1.43%,p < 0.001；中心2:9.10±5.59%,p < 0.001）。cDDIM_LN在两个中心的APE均低于基于gan的AC方法。结论：基于cddim的AC方法合成的示踪剂分布与临床CT-AC高度相似。在评价的方法中，cDDIM_LLN的性能最好。总体而言，基于cddim的AC优于传统的基于gan的AC。

{"title":"CT-free attenuation correction of 13N-ammonia cardiac PET images using conditional denoising diffusion implicit model with logarithmic linear normalization","authors":"Hao Sun , Xiaotong Hong , Yijun Lu , Fanghu Wang , Amirhossein Sanaat , Weiping Xu , Shuxia Wang , Habib Zaidi , Lijun Lu","doi":"10.1016/j.cmpb.2025.109188","DOIUrl":"10.1016/j.cmpb.2025.109188","url":null,"abstract":"<div><h3>Background and objective</h3><div>CT-based attenuation correction (CT-AC) is commonly used in cardiac PET but introduces additional radiation exposure, which is particularly problematic for pediatrics or repeat scans. This study aims to generate attenuation-corrected cardiac PET images from non-attenuation-corrected (NAC) PET input without CT using a conditional denoising diffusion implicit model (cDDIM).</div></div><div><h3>Methods</h3><div>The patient cohort included 60 rest-only and 14 rest-stress scans from Center 1, and 30 rest-stress scans from Center 2. All subjects underwent <sup>13</sup>N-ammonia cardiac PET/CT scans, generating paired 3D NAC and CT-AC PET images. The middle slice and two adjacent axial slices were used as inputs to mitigate axial artifacts, forming a 2.5D cDDIM framework. Two versions were proposed: one with data logarithmic linear normalization (cDDIM_LLN) and another with linear normalization (cDDIM_LN). 2.5D and 3D generative adversarial network (GAN)-based AC methods were implemented for comparison. Model performances were evaluated using normalized mean square error (NMSE) and segment-wise absolute percent error (APE).</div></div><div><h3>Results</h3><div>Visual analysis indicated that cDDIM_LLN and cDDIM_LN outperformed GAN-based AC methods for both rest and stress cardiac PET images. cDDIM_LLN achieved significantly lower NMSE than cDDIM_LN in Center 1 (1.87 ± 1.25 % vs. 2.77 ± 2.37 %, <em>p</em> < 0.001) and Center 2 (4.63 ± 3.71 % vs. 5.67 ± 7.88 %, <em>p</em> < 0.001). cDDIM_LLN also showed smaller APE than other methods for both centers (Center 1: 5.78 ± 1.43 %, <em>p</em> < 0.001; Center 2: 9.10 ± 5.59 %, <em>p</em> < 0.001). cDDIM_LN demonstrated lower APE than GAN-based AC methods across both centers.</div></div><div><h3>Conclusions</h3><div>cDDIM-based AC methods synthesized tracer distributions highly similar to clinical CT-AC. Among the evaluated methods, cDDIM_LLN demonstrated the best performance. Overall, cDDIM-based AC outperformed traditional GAN-based methods.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"275 ","pages":"Article 109188"},"PeriodicalIF":4.8,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145713201","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

A SAS macro for multilevel Cosinor analysis 用于多级余弦分析的SAS宏

IF 4.8 2区医学 Q1 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Computer methods and programs in biomedicine

Pub Date : 2026-02-01 Epub Date: 2025-11-14 DOI: 10.1016/j.cmpb.2025.109167

Margaret M. Doyle , Terrence E. Murphy , Brienne Miner , Melissa P. Knauert

Background and objective

Cosinor analysis allows for the fitting of a cosine curve to describe cyclical variation in periodic data. The analysis provides an intuitive set of estimates that includes the MESOR (Midline Estimating Statistic of Rhythm), i.e., the mid-point of the fitted outcome, the amplitude, i.e., one-half the distance between the MESOR and the peak for normally distributed outcomes, and the acrophase, i.e. the time at which the outcome reaches its peak. Traditionally, most published cosinor analyses were generated though a two-stage approach in which a curve was fit to each individual’s data and differences in the estimated cosinor parameters were compared in downstream analyses. More recently multilevel cosinor modeling software has been developed which allows for the simultaneous modeling of data from multiple individuals. In addition to simplifying the model building process, the advantage of multilevel vs. two-stage cosinor analysis includes the option to fit more complex models and, likely, an improvement in fit for each individual’s data. However, to our knowledge, there are no SAS procedures or macros that assist users with this analytical approach.

Methods

In this paper we introduce multilevel cosinor models and SAS macros we have developed to perform these analyses. In addition, we compare model fit between the multilevel and two-stage methods.

Results

The SAS macros presented in this paper allow users to select the best random variable specification for the unconditional cosinor model and add a dichotomous grouping variable to detect differences in parameters across groups. At each step of model building, parameter estimates, measures of model fit and graphical output help the user understand the model derived and its appropriateness for their data. Results of cross-validation analyses are presented that illustrate the superior fit of the multilevel over the single-level approach for the dataset utilized in the examples.

Conclusions

Multilevel cosinor analysis extends the single subject cosinor model by allowing for more convenient model selection and may provide a better fit for each individual’s data. We are hopeful that this manuscript will introduce more researchers to this analytical technique and allow them to apply it in their own research.

背景和目的余弦分析允许拟合余弦曲线来描述周期性数据的周期性变化。该分析提供了一组直观的估计，其中包括MESOR（节奏中线估计统计量），即拟合结果的中点，幅度，即正态分布结果的MESOR与峰值之间距离的一半，以及顶相，即结果达到峰值的时间。传统上，大多数已发表的余弦分析是通过两阶段方法生成的，其中曲线拟合每个个体的数据，并在下游分析中比较估计的余弦参数的差异。最近已经开发了多级余弦建模软件，它允许同时对来自多个个体的数据进行建模。除了简化模型构建过程之外，多层余弦分析与两阶段余弦分析的优势还包括适合更复杂模型的选项，并且可能改善每个个体数据的拟合。然而，据我们所知，没有SAS过程或宏可以帮助用户使用这种分析方法。方法在本文中，我们介绍了我们开发的用于这些分析的多层余弦模型和SAS宏。此外，我们比较了多级方法和两阶段方法的模型拟合。结果本文提出的SAS宏允许用户为无条件余弦模型选择最佳随机变量规格，并添加一个二分类分组变量来检测组间参数的差异。在模型构建的每个步骤中，参数估计、模型拟合度量和图形输出帮助用户理解所导出的模型及其对其数据的适当性。交叉验证分析的结果表明，对于示例中使用的数据集，多级方法优于单级方法的拟合。结论多层余弦分析扩展了单主体余弦模型，使模型选择更加方便，可以更好地拟合每个个体的数据。我们希望这篇文章能让更多的研究人员了解这种分析技术，并将其应用到自己的研究中。

{"title":"A SAS macro for multilevel Cosinor analysis","authors":"Margaret M. Doyle , Terrence E. Murphy , Brienne Miner , Melissa P. Knauert","doi":"10.1016/j.cmpb.2025.109167","DOIUrl":"10.1016/j.cmpb.2025.109167","url":null,"abstract":"<div><h3>Background and objective</h3><div>Cosinor analysis allows for the fitting of a cosine curve to describe cyclical variation in periodic data. The analysis provides an intuitive set of estimates that includes the MESOR (Midline Estimating Statistic of Rhythm), i.e., the mid-point of the fitted outcome, the amplitude, i.e., one-half the distance between the MESOR and the peak for normally distributed outcomes, and the acrophase, i.e. the time at which the outcome reaches its peak. Traditionally, most published cosinor analyses were generated though a two-stage approach in which a curve was fit to each individual’s data and differences in the estimated cosinor parameters were compared in downstream analyses. More recently multilevel cosinor modeling software has been developed which allows for the simultaneous modeling of data from multiple individuals. In addition to simplifying the model building process, the advantage of multilevel vs. two-stage cosinor analysis includes the option to fit more complex models and, likely, an improvement in fit for each individual’s data. However, to our knowledge, there are no SAS procedures or macros that assist users with this analytical approach.</div></div><div><h3>Methods</h3><div>In this paper we introduce multilevel cosinor models and SAS macros we have developed to perform these analyses. In addition, we compare model fit between the multilevel and two-stage methods.</div></div><div><h3>Results</h3><div>The SAS macros presented in this paper allow users to select the best random variable specification for the unconditional cosinor model and add a dichotomous grouping variable to detect differences in parameters across groups. At each step of model building, parameter estimates, measures of model fit and graphical output help the user understand the model derived and its appropriateness for their data. Results of cross-validation analyses are presented that illustrate the superior fit of the multilevel over the single-level approach for the dataset utilized in the examples.</div></div><div><h3>Conclusions</h3><div>Multilevel cosinor analysis extends the single subject cosinor model by allowing for more convenient model selection and may provide a better fit for each individual’s data. We are hopeful that this manuscript will introduce more researchers to this analytical technique and allow them to apply it in their own research.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"274 ","pages":"Article 109167"},"PeriodicalIF":4.8,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145615377","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Triggers and maintenance of idiopathic atrial fibrillation: A multiscale computational simulation study 特发性心房颤动的触发和维持：一项多尺度计算模拟研究

IF 4.8 2区医学 Q1 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Computer methods and programs in biomedicine

Pub Date : 2026-02-01 Epub Date: 2025-11-19 DOI: 10.1016/j.cmpb.2025.109173

Lian Xin , Li Haiying , Chen Yanhong , Yu Shiqi , He Linsheng , Wu Jian

Background and Objective

Idiopathic atrial fibrillation (IAF) is linked to electrical remodeling, yet prospective clinical cohort data isolating the individual contributions of candidate risk factors remain scarce. To address the gap, we used in-silico simulations to delineate the mechanisms and interactions underlying IAF.

Methods

We built a multi-scale atrial computational model on a 3D atrial anatomy integrating ion-channel kinetics, cellular electrophysiology, and tissue-level propagation. Atrial tissue conductivity (ATC), atrial effective refractory period (AERP), and sinus cycle length (SCL) were systematically varied at 6, 4, and 4 levels, respectively, within physiologic ranges. We recorded parameter sets that precipitated IAF and quantified dynamics by complexity, stability, and spatial disorder.

Results

In our simulations, ATC was represented by the diffusion coefficient D_scalar, and the action potential duration at −60 mV (APD-60mV) served as a surrogate for AERP. At SCL = 1000 ms, IAF initiated only when APD-60mV = 94 ms and D_scalar ≤ 0.0002 mm²/ms. With SCL shortening, the APD-60mV threshold for initiation decreased. SCL modulated susceptibility but was not sufficient alone. If either the AERP or ATC criterion was unmet, IAF did not initiate at any SCL. After initiation, both slow conduction and high frequency pacing increased arrhythmic complexity (spiral-wave count +63% and +129%); high frequency pacing enhanced stability (spiral-wave lifetimes up to +48%), whereas slow conduction worsened spatial disorder (organization index −27%). Spiral waves preferentially clustered along the interatrial septum.

Conclusions

IAF initiation requires both shortened AERP and reduced ATC, while maintenance is promoted by high-frequency pacing and slowed conduction. The interatrial septum emerges as a leading non–pulmonary-vein source. These findings provide mechanistic insight into IAF initiation and persistence and may inform early prevention.

背景与目的特发性心房颤动（IAF）与电重构有关，但分离候选危险因素个体贡献的前瞻性临床队列数据仍然很少。为了解决这一差距，我们使用了计算机模拟来描述IAF的机制和相互作用。方法建立三维心房解剖的多尺度心房计算模型，结合离子通道动力学、细胞电生理和组织水平的传播。心房组织电导率（ATC）、心房有效不应期（AERP）和窦循环长度（SCL）在生理范围内分别系统变化为6、4和4个水平。我们记录了沉淀IAF的参数集，并通过复杂性、稳定性和空间无序性量化了动态。结果在我们的模拟中，ATC用扩散系数d标量表示，动作电位持续时间在−60 mV （APD-60mV）时代表AERP。在SCL = 1000 ms时，只有当APD-60mV = 94 ms且Dscalar≤0.0002 mm2/ms时才会启动IAF。随着SCL的缩短，APD-60mV起始阈值降低。SCL可调节敏感性，但单独作用是不够的。如果AERP或ATC标准未满足，则IAF不会在任何SCL启动。起搏后，慢传导和高频起搏均增加心律失常复杂性（螺旋波计数分别为+63%和+129%）；高频起搏增强了稳定性（螺旋波寿命可达+48%），而慢传导恶化了空间紊乱（组织指数- 27%）。螺旋波优先沿心房间隔聚集。结论siaf的起始需要缩短AERP和降低ATC，而维持需要高频起搏和减慢传导。房间隔是主要的非肺静脉源。这些发现为IAF的发生和持续提供了机制见解，并可能为早期预防提供信息。

{"title":"Triggers and maintenance of idiopathic atrial fibrillation: A multiscale computational simulation study","authors":"Lian Xin , Li Haiying , Chen Yanhong , Yu Shiqi , He Linsheng , Wu Jian","doi":"10.1016/j.cmpb.2025.109173","DOIUrl":"10.1016/j.cmpb.2025.109173","url":null,"abstract":"<div><h3>Background and Objective</h3><div>Idiopathic atrial fibrillation (IAF) is linked to electrical remodeling, yet prospective clinical cohort data isolating the individual contributions of candidate risk factors remain scarce. To address the gap, we used in-silico simulations to delineate the mechanisms and interactions underlying IAF.</div></div><div><h3>Methods</h3><div>We built a multi-scale atrial computational model on a 3D atrial anatomy integrating ion-channel kinetics, cellular electrophysiology, and tissue-level propagation. Atrial tissue conductivity (ATC), atrial effective refractory period (AERP), and sinus cycle length (SCL) were systematically varied at 6, 4, and 4 levels, respectively, within physiologic ranges. We recorded parameter sets that precipitated IAF and quantified dynamics by complexity, stability, and spatial disorder.</div></div><div><h3>Results</h3><div>In our simulations, ATC was represented by the diffusion coefficient <em>D</em><sub>scalar</sub>, and the action potential duration at −60 mV (APD-60mV) served as a surrogate for AERP. At SCL = 1000 ms, IAF initiated only when APD-60mV = 94 ms and <em>D</em><sub>scalar</sub> ≤ 0.0002 mm<sup>2</sup>/ms. With SCL shortening, the APD-60mV threshold for initiation decreased. SCL modulated susceptibility but was not sufficient alone. If either the AERP or ATC criterion was unmet, IAF did not initiate at any SCL. After initiation, both slow conduction and high frequency pacing increased arrhythmic complexity (spiral-wave count +63% and +129%); high frequency pacing enhanced stability (spiral-wave lifetimes up to +48%), whereas slow conduction worsened spatial disorder (organization index −27%). Spiral waves preferentially clustered along the interatrial septum.</div></div><div><h3>Conclusions</h3><div>IAF initiation requires both shortened AERP and reduced ATC, while maintenance is promoted by high-frequency pacing and slowed conduction. The interatrial septum emerges as a leading non–pulmonary-vein source. These findings provide mechanistic insight into IAF initiation and persistence and may inform early prevention.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"274 ","pages":"Article 109173"},"PeriodicalIF":4.8,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145615447","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Integration of quantum artificial intelligence in disease diagnosis: A review of methods and applications 量子人工智能在疾病诊断中的集成：方法与应用综述

IF 4.8 2区医学 Q1 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Computer methods and programs in biomedicine

Pub Date : 2026-02-01 Epub Date: 2025-11-19 DOI: 10.1016/j.cmpb.2025.109175

Shobha Sharma , Lokesh Sharma , Tapan Kumar Gandhi

Background and objective

Accurate disease diagnosis is vital for effective treatment and improved patient outcomes. While artificial intelligence (AI) has advanced medical diagnostics, conventional AI approaches often face limitations in real-time data processing, scalability, and managing high-dimensional biomedical data. Quantum Artificial Intelligence (QAI) integrates quantum computing with AI to address these challenges. This study explores

QAI models in disease diagnosis, highlighting their advantages over classical AI, their applications across diseases, and integration possibilities within diagnostic workflows.

Methods

A structured literature review was conducted using Scopus, PubMed, IEEE Xplore, and Google Scholar databases. A total of 37 peer-reviewed articles were selected based on relevance, methodological quality, and focus on QAI applications in diagnostics. The review analyzed key quantum machine learning (QML) models, including hybrid and quantum inspired techniques.

Results

The findings indicate that QAI demonstrates promising applications in diagnosing cancer, neurodegenerative disorders, cardiovascular diseases, COVID-19, and other conditions. Quantum algorithms enable faster and more accurate pattern recognition in complex medical datasets. Additionally, QAI can be integrated into various stages of the diagnostic pipeline, from feature engineering to optimization to provide clinical decision support. However, technical challenges such as quantum noise, hardware instability, and limited algorithm maturity were frequently noted.

Conclusions

QAI has the potential to revolutionize disease diagnosis by overcoming many limitations of classical AI systems. While significant progress has been made, real-world clinical integration requires further advancements in algorithm development and hardware scalability. Future research should focus on closing the gap between theoretical models and clinical implementation to fully realize the benefits of QAI in healthcare.

背景与目的准确的疾病诊断对于有效治疗和改善患者预后至关重要。虽然人工智能（AI）具有先进的医疗诊断技术，但传统的人工智能方法在实时数据处理、可扩展性和高维生物医学数据管理方面往往面临限制。量子人工智能（QAI）将量子计算与人工智能相结合，以应对这些挑战。本研究探讨了qai模型在疾病诊断中的应用，强调了它们相对于经典AI的优势，它们在疾病中的应用，以及在诊断工作流程中的集成可能性。方法采用Scopus、PubMed、IEEE Xplore和谷歌Scholar数据库进行结构化文献综述。根据相关性、方法学质量和对QAI在诊断中的应用的关注，共选择了37篇同行评议的文章。该综述分析了关键的量子机器学习（QML）模型，包括混合和量子启发技术。结果QAI在诊断癌症、神经退行性疾病、心血管疾病、COVID-19等疾病方面具有广阔的应用前景。量子算法能够在复杂的医疗数据集中实现更快、更准确的模式识别。此外，QAI可以集成到诊断流程的各个阶段，从特征工程到优化，以提供临床决策支持。然而，量子噪声、硬件不稳定性和有限的算法成熟度等技术挑战经常被注意到。结论人工智能克服了传统人工智能系统的诸多局限性，具有革新疾病诊断的潜力。虽然已经取得了重大进展，但现实世界的临床整合需要在算法开发和硬件可扩展性方面取得进一步的进展。未来的研究应着眼于缩小理论模型与临床实施之间的差距，以充分实现质量评价在医疗保健中的益处。

{"title":"Integration of quantum artificial intelligence in disease diagnosis: A review of methods and applications","authors":"Shobha Sharma , Lokesh Sharma , Tapan Kumar Gandhi","doi":"10.1016/j.cmpb.2025.109175","DOIUrl":"10.1016/j.cmpb.2025.109175","url":null,"abstract":"<div><h3>Background and objective</h3><div>Accurate disease diagnosis is vital for effective treatment and improved patient outcomes. While artificial intelligence (AI) has advanced medical diagnostics, conventional AI approaches often face limitations in real-time data processing, scalability, and managing high-dimensional biomedical data. Quantum Artificial Intelligence (QAI) integrates quantum computing with AI to address these challenges. This study explores</div><div>QAI models in disease diagnosis, highlighting their advantages over classical AI, their applications across diseases, and integration possibilities within diagnostic workflows.</div></div><div><h3>Methods</h3><div>A structured literature review was conducted using Scopus, PubMed, IEEE Xplore, and Google Scholar databases. A total of 37 peer-reviewed articles were selected based on relevance, methodological quality, and focus on QAI applications in diagnostics. The review analyzed key quantum machine learning (QML) models, including hybrid and quantum inspired techniques.</div></div><div><h3>Results</h3><div>The findings indicate that QAI demonstrates promising applications in diagnosing cancer, neurodegenerative disorders, cardiovascular diseases, COVID-19, and other conditions. Quantum algorithms enable faster and more accurate pattern recognition in complex medical datasets. Additionally, QAI can be integrated into various stages of the diagnostic pipeline, from feature engineering to optimization to provide clinical decision support. However, technical challenges such as quantum noise, hardware instability, and limited algorithm maturity were frequently noted.</div></div><div><h3>Conclusions</h3><div>QAI has the potential to revolutionize disease diagnosis by overcoming many limitations of classical AI systems. While significant progress has been made, real-world clinical integration requires further advancements in algorithm development and hardware scalability. Future research should focus on closing the gap between theoretical models and clinical implementation to fully realize the benefits of QAI in healthcare.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"274 ","pages":"Article 109175"},"PeriodicalIF":4.8,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145615453","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Patients visits forecasting in emergency departments: Self-adapting LSTM models for evolving data distributions 急诊科患者就诊预测：数据分布演化的自适应LSTM模型

IF 4.8 2区医学 Q1 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Computer methods and programs in biomedicine

Pub Date : 2026-02-01 Epub Date: 2025-11-08 DOI: 10.1016/j.cmpb.2025.109153

Nicolas Haxaire , Farah Mourad-Chehade , Alice Yalaoui , Patrick Nader , Hicham Chehade

Background and Objective:

Overcrowding in emergency departments (EDs) remains a critical challenge in healthcare systems. Accurate forecasting of patient visits is essential to optimize the allocation of human and material resources. However, traditional forecasting models often fail when abrupt distribution shifts occur, such as during pandemics or organizational changes, because they require frequent manual retraining. The objective of this study is to design a deep learning-based forecasting framework that maintains accuracy under evolving conditions without manual intervention or full retraining.

Methods:

An adaptive forecasting approach is proposed, built on a Seq2Seq LSTM architecture, enhanced with two independent strategies: Per-Sequence Scaling (PSS), which preserves sequence-specific statistics, and Continual Learning (CL), which incrementally updates the model with new data through replay buffers. These methods are evaluated separately against a standard LSTM, ARIMA, and Prophet. Historical daily ED visit data covering pre-COVID, COVID, and post-COVID phases are used. Performance is assessed across 1-day, 7-day, and 30-day horizons using MSE, MAE, MAPE, and MASE.

Results:

The adaptive LSTM significantly reduces forecasting errors compared to both statistical and deep learning baselines on short and medium horizons (1-day and 1-week), achieving MASE values consistently below 1. For 1-month forecasts, performance converges towards the seasonal naïve benchmark, with ARIMA occasionally achieving superior accuracy during abrupt COVID-related shifts. Importantly, the adaptive LSTM regains superiority in post-COVID conditions and maintains stable accuracy by continuously adjusting to new distributions without requiring full retraining, a property particularly valuable under hospital privacy and IT constraints, thereby ensuring robustness across changing regimes and reducing operational burden in hospital settings.

Conclusions:

The adaptive Seq2Seq LSTM framework offers a robust and operationally viable solution for ED forecasting under evolving conditions. It consistently outperforms Prophet and ARIMA once data distributions stabilize, while only temporarily lagging behind ARIMA during structural shocks. Its ability to self-adjust incrementally removes the need for manual reconfiguration when data distributions change, making it particularly suitable for real-time hospital deployment under privacy and IT constraints. Future applications may extend beyond ED forecasting to other hospital departments facing similar variability in demand.

背景和目的：急诊科（EDs）人满为患仍然是医疗保健系统的一个关键挑战。准确预测患者就诊对于优化人力和物力资源的分配至关重要。然而，传统的预测模型往往在发生突然的分布变化时失效，例如在大流行或组织变化期间，因为它们需要频繁的人工再培训。本研究的目的是设计一个基于深度学习的预测框架，在不需要人工干预或完全再培训的情况下保持不断变化的条件下的准确性。方法：提出了一种基于Seq2Seq LSTM架构的自适应预测方法，并通过两种独立策略进行增强：保留序列特定统计信息的按序列缩放（PSS）和通过重放缓冲区使用新数据增量更新模型的持续学习（CL）。这些方法分别针对标准LSTM、ARIMA和Prophet进行评估。使用涵盖COVID前、COVID和COVID后阶段的历史每日ED访问数据。使用MSE、MAE、MAPE和MASE对1天、7天和30天的绩效进行评估。结果：与统计基线和深度学习基线相比，自适应LSTM在中短期（1天和1周）显著降低了预测误差，实现了MASE值始终低于1。对于1个月的预测，性能趋近于季节性naïve基准，在与covid相关的突然变化期间，ARIMA偶尔会达到更高的准确性。重要的是，自适应LSTM在后covid条件下恢复了优势，并通过不断调整以适应新的分布而保持稳定的准确性，而无需进行全面的再培训，这在医院隐私和IT约束下特别有价值，从而确保了在不断变化的制度下的鲁棒性，并减少了医院环境中的运营负担。结论：自适应Seq2Seq LSTM框架为不断变化的条件下ED预测提供了一个强大且操作可行的解决方案。一旦数据分布稳定，它的表现就会持续优于Prophet和ARIMA，而在结构性冲击期间，它只会暂时落后于ARIMA。它能够逐步自我调整，因此在数据分布发生变化时无需手动重新配置，因此特别适合在隐私和it限制下进行实时医院部署。未来的应用可能会超越急诊科预测到其他面临类似需求变化的医院部门。

{"title":"Patients visits forecasting in emergency departments: Self-adapting LSTM models for evolving data distributions","authors":"Nicolas Haxaire , Farah Mourad-Chehade , Alice Yalaoui , Patrick Nader , Hicham Chehade","doi":"10.1016/j.cmpb.2025.109153","DOIUrl":"10.1016/j.cmpb.2025.109153","url":null,"abstract":"<div><h3>Background and Objective:</h3><div>Overcrowding in emergency departments (EDs) remains a critical challenge in healthcare systems. Accurate forecasting of patient visits is essential to optimize the allocation of human and material resources. However, traditional forecasting models often fail when abrupt distribution shifts occur, such as during pandemics or organizational changes, because they require frequent manual retraining. The objective of this study is to design a deep learning-based forecasting framework that maintains accuracy under evolving conditions without manual intervention or full retraining.</div></div><div><h3>Methods:</h3><div>An adaptive forecasting approach is proposed, built on a Seq2Seq LSTM architecture, enhanced with two independent strategies: Per-Sequence Scaling (PSS), which preserves sequence-specific statistics, and Continual Learning (CL), which incrementally updates the model with new data through replay buffers. These methods are evaluated separately against a standard LSTM, ARIMA, and Prophet. Historical daily ED visit data covering pre-COVID, COVID, and post-COVID phases are used. Performance is assessed across 1-day, 7-day, and 30-day horizons using MSE, MAE, MAPE, and MASE.</div></div><div><h3>Results:</h3><div>The adaptive LSTM significantly reduces forecasting errors compared to both statistical and deep learning baselines on short and medium horizons (1-day and 1-week), achieving MASE values consistently below 1. For 1-month forecasts, performance converges towards the seasonal naïve benchmark, with ARIMA occasionally achieving superior accuracy during abrupt COVID-related shifts. Importantly, the adaptive LSTM regains superiority in post-COVID conditions and maintains stable accuracy by continuously adjusting to new distributions without requiring full retraining, a property particularly valuable under hospital privacy and IT constraints, thereby ensuring robustness across changing regimes and reducing operational burden in hospital settings.</div></div><div><h3>Conclusions:</h3><div>The adaptive Seq2Seq LSTM framework offers a robust and operationally viable solution for ED forecasting under evolving conditions. It consistently outperforms Prophet and ARIMA once data distributions stabilize, while only temporarily lagging behind ARIMA during structural shocks. Its ability to self-adjust incrementally removes the need for manual reconfiguration when data distributions change, making it particularly suitable for real-time hospital deployment under privacy and IT constraints. Future applications may extend beyond ED forecasting to other hospital departments facing similar variability in demand.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"274 ","pages":"Article 109153"},"PeriodicalIF":4.8,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145494802","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Intelligent ensemble learning-enhanced finite element modeling for precision thermal ablation in cancer therapy 基于智能集成学习的有限元模型在癌症治疗中的精确热消融。

IF 4.8 2区医学 Q1 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Computer methods and programs in biomedicine

Pub Date : 2026-02-01 Epub Date: 2025-11-12 DOI: 10.1016/j.cmpb.2025.109157

Hongfei Qiao , Ayesha Sohail , Peter Kim , Frank Hübner , Thomas Josef Vogl

Background:

Microwave ablation (MWA) is a minimally invasive treatment for liver tumors, yet accurate prediction of ablation zones remains challenging due to tissue heterogeneity, uncertain antenna placement, and complex thermal dynamics.

Objective:

This study develops a hybrid computational framework that integrates finite element modeling (FEM) with supervised machine learning to improve the prediction and optimization of MWA-induced tissue ablation.

Methods:

Ex vivo porcine liver experiments were conducted at varying power (40, 60, 80 W) and duration (3, 5, 10 min) settings, measuring long-axis, short-axis, and volumetric ablation outcomes. FEM simulations captured the coupled electromagnetic and thermal processes, while a Random Forest regression model was trained on FEM-generated data to optimize antenna insertion depth and predict ablation geometry with high fidelity. The framework enabled systematic exploration of the parameter space and reduced reliance on iterative laboratory experiments.

Results:

The integrated FEM-ML approach accurately predicted ablation dimensions, showing strong agreement with experimental measurements. Optimized antenna placement enhanced spatial and temporal temperature predictions, allowing precise estimation of lesion size, shape, and volume. The method supports rapid, patient-specific treatment planning and minimizes collateral tissue damage.

Conclusions:

Combining FEM simulations with supervised learning provides a scalable, data-driven framework for precision MWA. This approach improves predictive reliability, accelerates treatment planning, and enables adaptive, optimized thermal therapies, offering potential for enhanced clinical outcomes in minimally invasive oncological interventions.

背景：微波消融（MWA）是肝脏肿瘤的一种微创治疗方法，但由于组织异质性、天线位置不确定和复杂的热动力学，准确预测消融区域仍然具有挑战性。目的：本研究开发了一种混合计算框架，将有限元建模（FEM）与监督机器学习相结合，以提高mwa诱导组织消融的预测和优化。方法：在不同功率（40、60、80 W）和持续时间（3、5、10 min）设置下进行离体猪肝实验，测量长轴、短轴和体积消融结果。FEM模拟捕获了电磁和热耦合过程，同时基于FEM生成的数据训练随机森林回归模型，以优化天线插入深度并高保真地预测烧蚀几何形状。该框架能够系统地探索参数空间，减少对迭代实验室实验的依赖。结果：综合FEM-ML方法准确预测烧蚀尺寸，与实验测量结果一致。优化的天线放置增强了空间和时间温度预测，允许精确估计病变的大小、形状和体积。该方法支持快速，针对患者的治疗计划，并最大限度地减少附带组织损伤。结论：将FEM模拟与监督学习相结合，为精确MWA提供了一个可扩展的、数据驱动的框架。这种方法提高了预测的可靠性，加速了治疗计划，并实现了自适应、优化的热疗法，为微创肿瘤干预提供了潜在的增强临床结果。

{"title":"Intelligent ensemble learning-enhanced finite element modeling for precision thermal ablation in cancer therapy","authors":"Hongfei Qiao , Ayesha Sohail , Peter Kim , Frank Hübner , Thomas Josef Vogl","doi":"10.1016/j.cmpb.2025.109157","DOIUrl":"10.1016/j.cmpb.2025.109157","url":null,"abstract":"<div><h3>Background:</h3><div>Microwave ablation (MWA) is a minimally invasive treatment for liver tumors, yet accurate prediction of ablation zones remains challenging due to tissue heterogeneity, uncertain antenna placement, and complex thermal dynamics.</div></div><div><h3>Objective:</h3><div>This study develops a hybrid computational framework that integrates finite element modeling (FEM) with supervised machine learning to improve the prediction and optimization of MWA-induced tissue ablation.</div></div><div><h3>Methods:</h3><div>Ex vivo porcine liver experiments were conducted at varying power (40, 60, 80 W) and duration (3, 5, 10 min) settings, measuring long-axis, short-axis, and volumetric ablation outcomes. FEM simulations captured the coupled electromagnetic and thermal processes, while a Random Forest regression model was trained on FEM-generated data to optimize antenna insertion depth and predict ablation geometry with high fidelity. The framework enabled systematic exploration of the parameter space and reduced reliance on iterative laboratory experiments.</div></div><div><h3>Results:</h3><div>The integrated FEM-ML approach accurately predicted ablation dimensions, showing strong agreement with experimental measurements. Optimized antenna placement enhanced spatial and temporal temperature predictions, allowing precise estimation of lesion size, shape, and volume. The method supports rapid, patient-specific treatment planning and minimizes collateral tissue damage.</div></div><div><h3>Conclusions:</h3><div>Combining FEM simulations with supervised learning provides a scalable, data-driven framework for precision MWA. This approach improves predictive reliability, accelerates treatment planning, and enables adaptive, optimized thermal therapies, offering potential for enhanced clinical outcomes in minimally invasive oncological interventions.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"274 ","pages":"Article 109157"},"PeriodicalIF":4.8,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145548553","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Magnetic resonance-based computational modelling of healthy and prolapsing mitral valves to quantify the load transfer between the mitral apparatus and the ventricular myocardium 基于磁共振的健康二尖瓣和脱垂二尖瓣的计算模型，量化二尖瓣和心室心肌之间的负荷转移。

IF 4.8 2区医学 Q1 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Computer methods and programs in biomedicine

Pub Date : 2026-02-01 Epub Date: 2025-11-05 DOI: 10.1016/j.cmpb.2025.109151

Davide Tondi , Francesco Sturla , Mateo Marin-Cuartas , Maja-Theresa Dieterlen , Cosima Jahnke , Ingo Paetsch , Ricardo A Spampinato , Michael A. Borger , Emiliano Votta

Replacement myocardial fibrosis of the left ventricle (LV) is frequently observed in mitral valve prolapse (MVP) patients, possibly due to altered load transfer between the mitral valve (MV) and LV wall. Leveraging cardiac magnetic resonance (CMR) imaging, we propose an improved finite element (FE) model to quantify in vivo this load transfer in both healthy and MVP-affected valves.

The framework was tested on 6 subjects (3 healthy, 3 with MVP). CMR images were realigned via normalized cross-correlation; MV leaflets were reconstructed in a stress-free state, discretized and paired with a functionally equivalent chordal model. Chordae were partially removed in prolapsing segments and their length tuned. Valve closure was simulated under physiological pressure, incorporating anisotropic, hyperelastic tissue properties and subject-specific annular and papillary muscle (PM) motion. PMs head anatomy was separately simulated to assess stress transfer to the LV wall.

Simulations reliably reproduced in vivo MV closure, with median discrepancies in the order of imaging resolution. FE results suggest that at peak systole MVP is associated to increased stress at the base of the anterolateral and posteromedial PMs (values averaged over the subgroups: σ_AL = 0.16 MPa, σ_PM = 0.20 MPa vs. σ_AL = 0.11 MPa, σ_PM = 0.12 MPa) and to greater annular force at the insertion of the prolapsing mid-posterior cusp (values averaged over the subgroups: RF_P2 = 0.03 N vs. RF_P2 = 0.02 N) vs. healthy controls.

The framework reliably quantified MV–LV load transfer and its application to larger cohorts could provide valuable clinical insights.

在二尖瓣脱垂（MVP）患者中经常观察到左心室（LV）的替代心肌纤维化，可能是由于二尖瓣（MV）和左室壁之间负荷转移的改变。利用心脏磁共振（CMR）成像，我们提出了一种改进的有限元（FE）模型来量化健康和mvp影响瓣膜的体内负荷转移。该框架在6名受试者（3名健康受试者，3名MVP受试者）上进行了测试。通过归一化互相关对CMR图像进行重新对齐；在无应力状态下重建MV小叶，离散化并与功能等效的弦模型配对。在脱垂的节段中部分去除索索，并调整其长度。在生理压力下模拟瓣膜关闭，包括各向异性、超弹性组织特性和受试者特异性环和乳头肌（PM）运动。分别模拟pm头部解剖以评估左室壁的应力传递。模拟可靠地再现了体内的MV闭合，在成像分辨率的顺序上有中位数差异。结果表明，与健康对照组相比，收缩期MVP与前外侧和后内侧pm基部的应力增加有关（σAL = 0.16 MPa, σPM = 0.20 MPa, σAL = 0.11 MPa, σPM = 0.12 MPa），与脱垂的中后尖插入处的环力增加有关（各亚组平均值：RFP2 = 0.03 N, RFP2 = 0.02 N）。该框架可靠地量化了MV-LV负荷转移，并将其应用于更大的队列，可以提供有价值的临床见解。

{"title":"Magnetic resonance-based computational modelling of healthy and prolapsing mitral valves to quantify the load transfer between the mitral apparatus and the ventricular myocardium","authors":"Davide Tondi , Francesco Sturla , Mateo Marin-Cuartas , Maja-Theresa Dieterlen , Cosima Jahnke , Ingo Paetsch , Ricardo A Spampinato , Michael A. Borger , Emiliano Votta","doi":"10.1016/j.cmpb.2025.109151","DOIUrl":"10.1016/j.cmpb.2025.109151","url":null,"abstract":"<div><div>Replacement myocardial fibrosis of the left ventricle (LV) is frequently observed in mitral valve prolapse (MVP) patients, possibly due to altered load transfer between the mitral valve (MV) and LV wall. Leveraging cardiac magnetic resonance (CMR) imaging, we propose an improved finite element (FE) model to quantify <em>in vivo</em> this load transfer in both healthy and MVP-affected valves.</div><div>The framework was tested on 6 subjects (3 healthy, 3 with MVP). CMR images were realigned via normalized cross-correlation; MV leaflets were reconstructed in a stress-free state, discretized and paired with a functionally equivalent chordal model. Chordae were partially removed in prolapsing segments and their length tuned. Valve closure was simulated under physiological pressure, incorporating anisotropic, hyperelastic tissue properties and subject-specific annular and papillary muscle (PM) motion. PMs head anatomy was separately simulated to assess stress transfer to the LV wall.</div><div>Simulations reliably reproduced <em>in vivo</em> MV closure, with median discrepancies in the order of imaging resolution. FE results suggest that at peak systole MVP is associated to increased stress at the base of the anterolateral and posteromedial PMs (values averaged over the subgroups: σ<sub><em>AL</em></sub> = 0.16 MPa, σ<sub><em>PM</em></sub> = 0.20 MPa vs. σ<sub><em>AL</em></sub> = 0.11 MPa, σ<sub><em>PM</em></sub> = 0.12 MPa) and to greater annular force at the insertion of the prolapsing mid-posterior cusp (values averaged over the subgroups: <em>RF</em><sub><em>P</em>2</sub> = 0.03 N vs. <em>RF</em><sub><em>P</em>2</sub> = 0.02 N) vs. healthy controls.</div><div>The framework reliably quantified MV–LV load transfer and its application to larger cohorts could provide valuable clinical insights.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"274 ","pages":"Article 109151"},"PeriodicalIF":4.8,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145563116","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Prediction of respiratory rate from schlieren images using artificial intelligence 利用人工智能从纹影图像预测呼吸频率

IF 4.8 2区医学 Q1 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Computer methods and programs in biomedicine

Pub Date : 2026-02-01 Epub Date: 2025-11-29 DOI: 10.1016/j.cmpb.2025.109183

Byung Jun Kim , Woo Sang Cho , Hyungsoo Lim , Jung Chan Lee , Taegyun Kim , Hyeongkyu Kwak , Jieun Shin , Gil Joon Suh , Woon Yong Kwon , Kyung Su Kim , Yoon Sun Jung , Hayoung Kim , Heesu Park

Background and Objective

Respiratory rate is a fundamental physiological parameter and one of the earliest indicators that plays a crucial role in assessing a patient’s condition. However, existing respiratory rate measurement methods face challenges related to stability and invasiveness. In this study, we developed a noninvasive and stable method for measuring respiratory rate using the schlieren imaging technique combined with artificial intelligence (AI).

Methods

A system capable of accurately assessing respiratory rate was established using the schlieren imaging technique to visualize respiration-induced airflow, combined with AI. Schlieren-based respiratory images were acquired from 40 healthy subjects under three conditions: normal respiratory rate (2 min), fast respiratory rate (1 min), and slow respiratory rate (1 min). Using these images, three AI models (ResNet, InceptionNet, and EfficientNet) were trained to classify respiration and reconstruct respiratory signals by applying sine-fitting techniques to the classification results. Afterward, respiratory rates were estimated, and their performances were compared against those obtained using the standard tool called capnography.

Results

The accuracies of ResNet, InceptionNet, and EfficientNet were 0.82, 0.91, and 0.92, respectively. Among them, EfficientNet demonstrated the superior performance (accuracy: 0.92; precision: 0.90; sensitivity: 0.93; F1-score: 0.92). Intra-class correlation values between the respiratory rates predicted by the three models (ResNet, InceptionNet, and EfficientNet) and the actual respiratory rate were 0.99, 0.99, and 0.98, respectively.

Conclusions

These results suggest that our AI-based respiratory monitoring system using the schlieren imaging technique provides a stable and non-invasive method for accurate respiratory assessment.

背景与目的呼吸频率是一项基本的生理参数，是早期评估患者病情的重要指标之一。然而，现有的呼吸频率测量方法面临着稳定性和侵入性方面的挑战。在本研究中，我们开发了一种使用纹影成像技术结合人工智能（AI）的无创、稳定的呼吸频率测量方法。方法采用纹影成像技术，结合人工智能技术，建立一套能够准确评估呼吸频率的呼吸诱导气流可视化系统。对40名健康受试者在正常呼吸速率（2 min）、快速呼吸速率（1 min）和缓慢呼吸速率（1 min）三种情况下进行纹影成像。利用这些图像，训练三个人工智能模型（ResNet、InceptionNet和EfficientNet）对呼吸进行分类，并通过对分类结果应用正弦拟合技术重建呼吸信号。之后，评估呼吸频率，并将其与使用称为二氧化碳成像的标准工具获得的结果进行比较。结果ResNet、InceptionNet和effentnet的准确率分别为0.82、0.91和0.92。其中，效率网表现出较好的性能（准确度：0.92，精密度：0.90，灵敏度：0.93,F1-score: 0.92）。三种模型（ResNet、InceptionNet和EfficientNet）预测的呼吸速率与实际呼吸速率的类内相关值分别为0.99、0.99和0.98。结论基于纹影成像技术的人工智能呼吸监测系统为准确的呼吸评估提供了一种稳定、无创的方法。

{"title":"Prediction of respiratory rate from schlieren images using artificial intelligence","authors":"Byung Jun Kim , Woo Sang Cho , Hyungsoo Lim , Jung Chan Lee , Taegyun Kim , Hyeongkyu Kwak , Jieun Shin , Gil Joon Suh , Woon Yong Kwon , Kyung Su Kim , Yoon Sun Jung , Hayoung Kim , Heesu Park","doi":"10.1016/j.cmpb.2025.109183","DOIUrl":"10.1016/j.cmpb.2025.109183","url":null,"abstract":"<div><h3>Background and Objective</h3><div>Respiratory rate is a fundamental physiological parameter and one of the earliest indicators that plays a crucial role in assessing a patient’s condition. However, existing respiratory rate measurement methods face challenges related to stability and invasiveness. In this study, we developed a noninvasive and stable method for measuring respiratory rate using the schlieren imaging technique combined with artificial intelligence (AI).</div></div><div><h3>Methods</h3><div>A system capable of accurately assessing respiratory rate was established using the schlieren imaging technique to visualize respiration-induced airflow, combined with AI. Schlieren-based respiratory images were acquired from 40 healthy subjects under three conditions: normal respiratory rate (2 min), fast respiratory rate (1 min), and slow respiratory rate (1 min). Using these images, three AI models (ResNet, InceptionNet, and EfficientNet) were trained to classify respiration and reconstruct respiratory signals by applying sine-fitting techniques to the classification results. Afterward, respiratory rates were estimated, and their performances were compared against those obtained using the standard tool called capnography.</div></div><div><h3>Results</h3><div>The accuracies of ResNet, InceptionNet, and EfficientNet were 0.82, 0.91, and 0.92, respectively. Among them, EfficientNet demonstrated the superior performance (accuracy: 0.92; precision: 0.90; sensitivity: 0.93; F1-score: 0.92). Intra-class correlation values between the respiratory rates predicted by the three models (ResNet, InceptionNet, and EfficientNet) and the actual respiratory rate were 0.99, 0.99, and 0.98, respectively.</div></div><div><h3>Conclusions</h3><div>These results suggest that our AI-based respiratory monitoring system using the schlieren imaging technique provides a stable and non-invasive method for accurate respiratory assessment.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"275 ","pages":"Article 109183"},"PeriodicalIF":4.8,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145682463","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

TiDE-Net: A time-guided dual-encoder ResUNet for Positron Emission Tomography (PET) image denoising 潮汐网：用于正电子发射断层扫描（PET）图像去噪的时间导向双编码器重构网

IF 4.8 2区医学 Q1 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Computer methods and programs in biomedicine

Pub Date : 2026-02-01 Epub Date: 2025-12-02 DOI: 10.1016/j.cmpb.2025.109200

Boyuan Tan , Yuxin Xue , Lei Bi , Michael Fulham , Jinman Kim

Background and Objective:

Positron emission tomography (PET) imaging enables quantitative assessment of metabolic activity in the human body and is widely used for clinical diagnosis and monitoring treatment. PET imaging, however, is inherently affected by noise due to a number of factors including radioactive decay from the injected radiotracer, the amount of radiotracer injected, attenuation of photons, patient characteristics (e.g Body Mass Index), the underlying condition and the characteristics of the instrumentation employed (the type of PET-CT scanner). Low-dose imaging protocols have been in vogue to minimize patient radiation exposure. As such, extensive research has been focused on reducing noise levels while minimizing the administered radiotracer dose. State-of-the-art denoising methods are trained on data with fixed noise levels, i.e., noise with consistent statistical properties across all samples. This assumption limits their ability to generalize to real-world PET imaging scenarios, where noise levels differ from those seen during training for the reason outlined above.

Methods:

We propose a time-guided dual-encoder ResUNet (TiDE-Net) to remove reliance on fixed noise levels at both training and inference stages for PET image denoising. TiDE-Net incorporates a Dual-Encoder with Complementary Attention (DECA) to extract global context and local structure, enhancing feature representation under varying noise conditions. To enhance adaptability, we introduce a Time-Guided Mechanism (TGM), where a scalar timestep variable represents and controls the level of simulated noise, where higher timesteps correspond to higher noise levels. By simulating varying noise levels from standard-dose PET (stndPET) data, TiDE-Net learns a continuous representation of noise characteristics.

Results:

In our experiments, TiDE-Net outperformed state-of-the-art comparison methods on two benchmark datasets (664 patient studies from Siemens Biograph Vision Quadra and United Imaging uEXPLORER total body scanners), across all noise levels. Ablation studies further demonstrated that our proposed TGM and DECA modules markedly improved performance, particularly on dose levels not seen during training. Clinical evaluation through SUV analysis further demonstrates the model’s ability to preserve quantitative accuracy across different regions.

Conclusion:

Our experiments demonstrate that conditioning on a timestep enables TiDE-Net to generalize across unseen noise levels, achieving consistent denoising performance across different noise conditions.

背景与目的：正电子发射断层扫描（PET）成像能够定量评估人体的代谢活动，广泛应用于临床诊断和监测治疗。然而，由于注入的放射性示踪剂的放射性衰变、注入的放射性示踪剂的量、光子的衰减、患者特征（如体重指数）、潜在条件和所使用仪器的特征（PET- ct扫描仪的类型）等许多因素，PET成像本质上受到噪声的影响。低剂量成像方案已经流行，以尽量减少病人的辐射暴露。因此，广泛的研究集中在降低噪音水平的同时尽量减少放射性示踪剂的剂量。最先进的去噪方法是在具有固定噪声水平的数据上进行训练的，即在所有样本中具有一致统计特性的噪声。这一假设限制了他们推广到真实世界PET成像场景的能力，在真实世界中，由于上述原因，噪声水平与训练期间看到的不同。方法：我们提出了一个时间导向的双编码器ResUNet (TiDE-Net)，以消除PET图像去噪在训练和推理阶段对固定噪声水平的依赖。TiDE-Net采用了具有互补注意（DECA）的双编码器来提取全局上下文和局部结构，增强了在不同噪声条件下的特征表示。为了增强适应性，我们引入了时间引导机制（TGM），其中标量时间步长变量表示并控制模拟噪声的水平，其中越高的时间步长对应越高的噪声水平。通过模拟来自标准剂量PET （stndPET）数据的不同噪声水平，TiDE-Net学习噪声特征的连续表示。结果：在我们的实验中，TiDE-Net在两个基准数据集（来自Siemens Biograph Vision Quadra和United Imaging uEXPLORER全身扫描仪的664例患者研究）上的所有噪声水平上都优于最先进的比较方法。消融研究进一步表明，我们提出的TGM和DECA模块显著提高了性能，特别是在训练期间未见的剂量水平上。通过SUV分析的临床评估进一步证明了该模型在不同区域保持定量准确性的能力。结论：我们的实验表明，对时间步长的调节使TiDE-Net能够在看不见的噪声水平上进行泛化，从而在不同的噪声条件下实现一致的去噪性能。

{"title":"TiDE-Net: A time-guided dual-encoder ResUNet for Positron Emission Tomography (PET) image denoising","authors":"Boyuan Tan , Yuxin Xue , Lei Bi , Michael Fulham , Jinman Kim","doi":"10.1016/j.cmpb.2025.109200","DOIUrl":"10.1016/j.cmpb.2025.109200","url":null,"abstract":"<div><h3>Background and Objective:</h3><div>Positron emission tomography (PET) imaging enables quantitative assessment of metabolic activity in the human body and is widely used for clinical diagnosis and monitoring treatment. PET imaging, however, is inherently affected by noise due to a number of factors including radioactive decay from the injected radiotracer, the amount of radiotracer injected, attenuation of photons, patient characteristics (e.g Body Mass Index), the underlying condition and the characteristics of the instrumentation employed (the type of PET-CT scanner). Low-dose imaging protocols have been in vogue to minimize patient radiation exposure. As such, extensive research has been focused on reducing noise levels while minimizing the administered radiotracer dose. State-of-the-art denoising methods are trained on data with fixed noise levels, i.e., noise with consistent statistical properties across all samples. This assumption limits their ability to generalize to real-world PET imaging scenarios, where noise levels differ from those seen during training for the reason outlined above.</div></div><div><h3>Methods:</h3><div>We propose a time-guided dual-encoder ResUNet (TiDE-Net) to remove reliance on fixed noise levels at both training and inference stages for PET image denoising. TiDE-Net incorporates a Dual-Encoder with Complementary Attention (DECA) to extract global context and local structure, enhancing feature representation under varying noise conditions. To enhance adaptability, we introduce a Time-Guided Mechanism (TGM), where a scalar timestep variable represents and controls the level of simulated noise, where higher timesteps correspond to higher noise levels. By simulating varying noise levels from standard-dose PET (stndPET) data, TiDE-Net learns a continuous representation of noise characteristics.</div></div><div><h3>Results:</h3><div>In our experiments, TiDE-Net outperformed state-of-the-art comparison methods on two benchmark datasets (664 patient studies from Siemens Biograph Vision Quadra and United Imaging uEXPLORER total body scanners), across all noise levels. Ablation studies further demonstrated that our proposed TGM and DECA modules markedly improved performance, particularly on dose levels not seen during training. Clinical evaluation through SUV analysis further demonstrates the model’s ability to preserve quantitative accuracy across different regions.</div></div><div><h3>Conclusion:</h3><div>Our experiments demonstrate that conditioning on a timestep enables TiDE-Net to generalize across unseen noise levels, achieving consistent denoising performance across different noise conditions.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"275 ","pages":"Article 109200"},"PeriodicalIF":4.8,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145682477","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0