Computers in biology and medicine最新文献_第9页

Unraveling blood pressure estimation with a deep learning approach using multiple embeddings 使用多个嵌入的深度学习方法解开血压估计。

IF 6.3 2区医学 Q1 BIOLOGY

Computers in biology and medicine

Pub Date : 2025-12-08 DOI: 10.1016/j.compbiomed.2025.111377

Vishal Singh Roha, Mehmet R. Yuce

We introduce a calibration-free machine learning framework for BP estimation using pulse arrival time (PAT), computed from the electrocardiogram’s R-peak and photoplethysmography P-peak. To enhance pattern recognition and unveil hidden patterns within the data samples, we introduce the use of similarity-based features based on Euclidean and Manhattan distance matrices, which are then processed by an attention-guided convolutional neural network. The model was successfully evaluated across three datasets: Cabrini Hospital, PTT PPG, and MIMIC-II, where our framework achieved a

R^{2}

values of 0.89, 0.95, and 0.92 for systolic BP (SBP) and 0.89, 0.94, and 0.91 for diastolic BP (DBP), respectively, along with mean absolute errors of 6.45, 1.31, and 2.12 mmHg for SBP and 2.92, 0.98, and 1.14 mmHg for DBP. The framework meets the Advancement of Medical Instrumentation standard on all datasets and achieves British Hypertension Society Grade ‘A’ for both BP types on the PTT PPG and MIMIC-II, and Grade ‘A’ and ‘B’ for DBP and SBP on the Cabrini, respectively. With strong generalizability, real-time compatibility, and no requirement for subject-specific calibration, the proposed framework demonstrates strong correlation, low prediction errors, and clinical applicability across diverse populations, offering a promising solution for continuous, comfortable, and reliable BP monitoring.

我们引入了一个无需校准的机器学习框架，用于使用脉冲到达时间（PAT）来估计BP，该时间是根据心电图的r峰和光容积脉搏波p峰计算的。为了增强模式识别并揭示数据样本中的隐藏模式，我们引入了基于欧几里得和曼哈顿距离矩阵的基于相似性的特征，然后通过注意引导的卷积神经网络对其进行处理。该模型成功地在三个数据集上进行了评估：Cabrini Hospital， PTT PPG和MIMIC-II，我们的框架在收缩压（SBP）和舒张压（DBP）方面分别获得了0.89,0.95和0.92的R2值，舒张压（DBP）分别获得了0.89,0.94和0.91，以及平均绝对误差为6.45,1.31和2.12 mmHg，舒张压为2.92,0.98和1.14 mmHg。该框架在所有数据集上都符合医疗器械进步标准，并在PTT PPG和MIMIC-II的两种血压类型上分别达到英国高血压协会的“A”级，在Cabrini的舒张压和收缩压方面分别达到“A”级和“B”级。该框架具有较强的通用性、实时兼容性和不需要受试者特定校准，具有较强的相关性、较低的预测误差和不同人群的临床适用性，为连续、舒适、可靠的血压监测提供了一个有希望的解决方案。

{"title":"Unraveling blood pressure estimation with a deep learning approach using multiple embeddings","authors":"Vishal Singh Roha, Mehmet R. Yuce","doi":"10.1016/j.compbiomed.2025.111377","DOIUrl":"10.1016/j.compbiomed.2025.111377","url":null,"abstract":"<div><div>We introduce a calibration-free machine learning framework for BP estimation using pulse arrival time (PAT), computed from the electrocardiogram’s R-peak and photoplethysmography P-peak. To enhance pattern recognition and unveil hidden patterns within the data samples, we introduce the use of similarity-based features based on Euclidean and Manhattan distance matrices, which are then processed by an attention-guided convolutional neural network. The model was successfully evaluated across three datasets: Cabrini Hospital, PTT PPG, and MIMIC-II, where our framework achieved a <span><math><msup><mrow><mi>R</mi></mrow><mrow><mn>2</mn></mrow></msup></math></span> values of 0.89, 0.95, and 0.92 for systolic BP (SBP) and 0.89, 0.94, and 0.91 for diastolic BP (DBP), respectively, along with mean absolute errors of 6.45, 1.31, and 2.12 mmHg for SBP and 2.92, 0.98, and 1.14 mmHg for DBP. The framework meets the Advancement of Medical Instrumentation standard on all datasets and achieves British Hypertension Society Grade ‘A’ for both BP types on the PTT PPG and MIMIC-II, and Grade ‘A’ and ‘B’ for DBP and SBP on the Cabrini, respectively. With strong generalizability, real-time compatibility, and no requirement for subject-specific calibration, the proposed framework demonstrates strong correlation, low prediction errors, and clinical applicability across diverse populations, offering a promising solution for continuous, comfortable, and reliable BP monitoring.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"200 ","pages":"Article 111377"},"PeriodicalIF":6.3,"publicationDate":"2025-12-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145713640","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

Group-level and personalized optimization for the insula and hippocampus focal electric field in transcranial temporal interferential stimulation: A computational study 群体水平和个性化优化的脑岛和海马局灶电场经颅颞叶干扰刺激：一项计算研究。

IF 6.3 2区医学 Q1 BIOLOGY

Computers in biology and medicine

Pub Date : 2025-12-08 DOI: 10.1016/j.compbiomed.2025.111374

Taiga Inoue , Naofumi Otsuru , Akimasa Hirata

Objectives

This study evaluated the efficacy of transcranial temporal interference stimulation (tTIS) for focal stimulation of the insula and hippocampus, which are clinically relevant but anatomically challenging targets. Individualized and group-level electrode optimizations were compared to determine whether generalized montages can provide reliable targeting while reducing the modeling demands.

Methods

Sixty high-resolution anatomical head models (30 individuals and their mirrored counterparts) were constructed from T1-and T2-weighted magnetic resonance images. The electric field (EF) distributions were determined using the scalar-potential finite difference method. The electrode montages and current ratios were optimized to minimize the root-mean-square error between the simulated and target EF envelope (EFE) distributions. A stimulation threshold of 0.3 V/m was applied. Subsampling analysis was performed to estimate the number of head models required for stable group-level results.

Results

For insular targeting, a novel montage combining T7–P7 and Fp1–Fp2 achieved the highest focality. The focality was comparable to most individualized configurations and reduced interindividual variability. For hippocampal targeting, a newly proposed montage combining F7–T7 and T8–P8 yielded the best group-level focality. However, individualized optimization was required in a subset of cases to achieve adequate off-target suppression. Reliable group-level EF patterns were obtained using ∼20 models for the insula and ∼9 for the hippocampus.

Conclusions

The findings show optimal transcranial stimulation montages depend on the target's anatomical depth. For cortical targets, including deep areas like the insula, group-level montages derived from sufficiently diverse anatomical models can achieve both high focality and applicability. However, for subcortical targets like the hippocampus, individualized optimization is recommended to maximize focality and minimize off-target activation, despite requiring fewer models to achieve stable group-level patterns.

目的：本研究评估经颅颞叶干扰刺激（tTIS）对脑岛和海马局灶性刺激的疗效，这是临床相关但解剖学上具有挑战性的靶点。比较了个性化和群体水平的电极优化，以确定广义蒙太奇是否可以在减少建模需求的同时提供可靠的目标。方法：利用t1和t2加权磁共振图像构建60个高分辨率头颅解剖模型（30个个体及其镜像模型）。采用标势有限差分法确定了电场分布。优化了电极蒙太奇和电流比，以最小化模拟和目标EF包络分布之间的均方根误差。刺激阈值为0.3 V/m。进行了次抽样分析，以估计稳定组水平结果所需的头部模型数量。结果：对于胰岛靶向，结合T7-P7和Fp1-Fp2的新型蒙太奇具有最高的聚焦性。焦点性与大多数个体化配置相当，减少了个体间的可变性。对于海马靶向，新提出的结合F7-T7和T8-P8的蒙太奇产生了最佳的组水平聚焦。然而，在一些情况下，需要个性化的优化来实现足够的脱靶抑制。使用~ 20个脑岛模型和~ 9个海马模型获得了可靠的组水平EF模式。结论：最佳的经颅刺激蒙太奇取决于目标的解剖深度。对于皮层目标，包括脑岛等深部区域，从足够多样化的解剖模型中获得的群体水平蒙太奇可以实现高聚焦性和适用性。然而，对于海马等皮质下靶点，尽管需要较少的模型来获得稳定的群体水平模式，但建议进行个体化优化，以最大化聚焦并最小化脱靶激活。

{"title":"Group-level and personalized optimization for the insula and hippocampus focal electric field in transcranial temporal interferential stimulation: A computational study","authors":"Taiga Inoue , Naofumi Otsuru , Akimasa Hirata","doi":"10.1016/j.compbiomed.2025.111374","DOIUrl":"10.1016/j.compbiomed.2025.111374","url":null,"abstract":"<div><h3>Objectives</h3><div>This study evaluated the efficacy of transcranial temporal interference stimulation (tTIS) for focal stimulation of the insula and hippocampus, which are clinically relevant but anatomically challenging targets. Individualized and group-level electrode optimizations were compared to determine whether generalized montages can provide reliable targeting while reducing the modeling demands.</div></div><div><h3>Methods</h3><div>Sixty high-resolution anatomical head models (30 individuals and their mirrored counterparts) were constructed from T1-and T2-weighted magnetic resonance images. The electric field (EF) distributions were determined using the scalar-potential finite difference method. The electrode montages and current ratios were optimized to minimize the root-mean-square error between the simulated and target EF envelope (EFE) distributions. A stimulation threshold of 0.3 V/m was applied. Subsampling analysis was performed to estimate the number of head models required for stable group-level results.</div></div><div><h3>Results</h3><div>For insular targeting, a novel montage combining T7–P7 and Fp1–Fp2 achieved the highest focality. The focality was comparable to most individualized configurations and reduced interindividual variability. For hippocampal targeting, a newly proposed montage combining F7–T7 and T8–P8 yielded the best group-level focality. However, individualized optimization was required in a subset of cases to achieve adequate off-target suppression. Reliable group-level EF patterns were obtained using ∼20 models for the insula and ∼9 for the hippocampus.</div></div><div><h3>Conclusions</h3><div>The findings show optimal transcranial stimulation montages depend on the target's anatomical depth. For cortical targets, including deep areas like the insula, group-level montages derived from sufficiently diverse anatomical models can achieve both high focality and applicability. However, for subcortical targets like the hippocampus, individualized optimization is recommended to maximize focality and minimize off-target activation, despite requiring fewer models to achieve stable group-level patterns.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"200 ","pages":"Article 111374"},"PeriodicalIF":6.3,"publicationDate":"2025-12-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145713649","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

Computational investigation of the effect of pre-existing silent myocardial infarction on acute myocardial infarction 既往无症状心肌梗死对急性心肌梗死影响的计算研究。

IF 6.3 2区医学 Q1 BIOLOGY

Computers in biology and medicine

Pub Date : 2025-12-08 DOI: 10.1016/j.compbiomed.2025.111364

Md Shah Wali Ullah , Jijo Derick Abraham , Samayala Rahman Oyshie , Sheikh Mohammad Shavik

Recent clinical studies have reported that the pre-existence of unnoticed silent myocardial infarction (SMI) in patients hospitalized with acute myocardial infarction (AMI) significantly worsens morbidity and increases mortality. SMI patients exhibit infarcts of different sizes and locations. This condition occurs in conjunction with left ventricular (LV) hypertrophy. The distinct roles of these abnormalities on LV functions and their contribution to AMI prognosis are poorly understood. To address these gaps, this study utilizes a coupled finite element (FE) framework to investigate the effect of the size and location of isolated SMI and AMI on LV performance. Furthermore, the influence of the simultaneous occurrence of SMI and hypertrophy alongside AMI is evaluated. The results show that under the simulation conditions, the peak global longitudinal strain is influenced by the variation in the location and size of the SMI and AMI. However, ejection fraction (EF), end-diastolic pressure (EDP), and peak global circumferential strain are influenced by the size of the SMI. The findings indicate that the presence of pre-existing SMI and hypertrophy will notably impair the EF and global circumferential strain, resulting in worse morbidity in an AMI patient.

最近的临床研究报道，急性心肌梗死（AMI）住院患者先前存在未被注意到的无症状心肌梗死（SMI）可显著加重发病率并增加死亡率。重度精神分裂症患者表现出不同大小和部位的梗死。这种情况与左心室肥厚同时发生。这些异常对左室功能的不同作用及其对AMI预后的贡献尚不清楚。为了解决这些差距，本研究利用耦合有限元（FE）框架来研究孤立的SMI和AMI的大小和位置对左室性能的影响。此外，我们还评估了AMI同时发生重度精神损伤和心肌肥厚的影响。结果表明：在模拟条件下，整体纵向应变峰值受SMI和AMI位置和尺寸变化的影响；然而，射血分数（EF）、舒张末压（EDP）和峰值总周向应变受SMI大小的影响。研究结果表明，先前存在的SMI和肥厚会显著损害EF和全周应变，导致AMI患者更严重的发病率。

{"title":"Computational investigation of the effect of pre-existing silent myocardial infarction on acute myocardial infarction","authors":"Md Shah Wali Ullah , Jijo Derick Abraham , Samayala Rahman Oyshie , Sheikh Mohammad Shavik","doi":"10.1016/j.compbiomed.2025.111364","DOIUrl":"10.1016/j.compbiomed.2025.111364","url":null,"abstract":"<div><div>Recent clinical studies have reported that the pre-existence of unnoticed silent myocardial infarction (SMI) in patients hospitalized with acute myocardial infarction (AMI) significantly worsens morbidity and increases mortality. SMI patients exhibit infarcts of different sizes and locations. This condition occurs in conjunction with left ventricular (LV) hypertrophy. The distinct roles of these abnormalities on LV functions and their contribution to AMI prognosis are poorly understood. To address these gaps, this study utilizes a coupled finite element (FE) framework to investigate the effect of the size and location of isolated SMI and AMI on LV performance. Furthermore, the influence of the simultaneous occurrence of SMI and hypertrophy alongside AMI is evaluated. The results show that under the simulation conditions, the peak global longitudinal strain is influenced by the variation in the location and size of the SMI and AMI. However, ejection fraction (EF), end-diastolic pressure (EDP), and peak global circumferential strain are influenced by the size of the SMI. The findings indicate that the presence of pre-existing SMI and hypertrophy will notably impair the EF and global circumferential strain, resulting in worse morbidity in an AMI patient.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"200 ","pages":"Article 111364"},"PeriodicalIF":6.3,"publicationDate":"2025-12-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145713621","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

AI-driven transfer learning and classical molecular dynamics for strategic therapeutic repurposing and rational design of antiviral peptides targeting monkeypox virus DNA polymerase 人工智能驱动的迁移学习和经典分子动力学用于猴痘病毒DNA聚合酶抗病毒肽的策略性治疗再利用和合理设计。

IF 6.3 2区医学 Q1 BIOLOGY

Computers in biology and medicine

Pub Date : 2025-12-07 DOI: 10.1016/j.compbiomed.2025.111372

Leona Dcunha , Sonet Daniel Thomas , Levin John , Amritha Thaikkad , Dileep Francis , Rajesh Raju , Nik Hirani , Abhithaj Jayanandan

The emergence of monkeypox virus (MPXV) as a global health threat has necessitated the rapid identification of novel antiviral therapeutics. Currently, no FDA-approved drugs are specifically designed against the disease. We used an in-house deep learning pharmacophore model for screening a library of 1974 FDA-approved drugs targeting the active site of MPXV DNA polymerase. Three drugs exhibited the strongest binding affinities, outperforming the control drug, Cidofovir diphosphate, and forming stable interactions with key active site residues. Among them, Paromomycin emerged as the most favourable drug, demonstrating stable, persistent, and adaptable interactions in molecular dynamics simulation. In parallel, we developed a novel automated peptide-generating AI pipeline that integrates active-site residues with knowledge-guided amino acid selection to generate and evaluate synthetic peptides. Cysteine-Phenylalanine-Cysteine (CFC), together with a panel of candidates, emerged through rational balancing of physicochemical properties and drug-likeness for accelerated therapeutic discovery. Synthetic peptides were evaluated to further understand the binding efficacies with DNA polymerase. CFC peptide demonstrated strong binding affinity (−8.08 kcal/mol) through stable interactions with key catalytic residues ASP549, ARG634 and LYS661, while MMGBSA analysis confirmed favourable binding energy (−33.02 kcal/mol). Consistent results in MD simulations indicate functional binding without destabilisation. Although ADMET predictions for CFC revealed limitations in permeability and oral bioavailability, its favourable binding profile and reduced predicted toxicity support its potential as a novel antiviral lead.

猴痘病毒（MPXV）作为一种全球健康威胁的出现，要求迅速确定新的抗病毒治疗方法。目前，没有fda批准的药物是专门针对这种疾病设计的。我们使用内部深度学习药效团模型筛选了1974年fda批准的靶向MPXV DNA聚合酶活性位点的药物库。三种药物表现出最强的结合亲和力，优于对照药物西多福韦二磷酸，并与关键活性位点残基形成稳定的相互作用。其中，Paromomycin在分子动力学模拟中表现出稳定、持久和适应性强的相互作用，是最受欢迎的药物。同时，我们开发了一种新的自动化肽生成AI管道，该管道将活性位点残基与知识引导的氨基酸选择相结合，以生成和评估合成肽。半胱氨酸-苯丙氨酸-半胱氨酸（CFC），连同一组候选药物，通过合理平衡物理化学性质和药物相似性来加速治疗发现。对合成肽进行评价，以进一步了解其与DNA聚合酶的结合效果。CFC肽通过与关键催化残基ASP549、ARG634和LYS661的稳定相互作用，显示出较强的结合亲和力（-8.08 kcal/mol），而MMGBSA分析证实了较好的结合能（-33.02 kcal/mol）。MD模拟的一致结果表明没有不稳定的功能结合。尽管ADMET预测CFC的渗透性和口服生物利用度存在局限性，但其良好的结合特性和较低的预测毒性支持其作为新型抗病毒先导物的潜力。

{"title":"AI-driven transfer learning and classical molecular dynamics for strategic therapeutic repurposing and rational design of antiviral peptides targeting monkeypox virus DNA polymerase","authors":"Leona Dcunha , Sonet Daniel Thomas , Levin John , Amritha Thaikkad , Dileep Francis , Rajesh Raju , Nik Hirani , Abhithaj Jayanandan","doi":"10.1016/j.compbiomed.2025.111372","DOIUrl":"10.1016/j.compbiomed.2025.111372","url":null,"abstract":"<div><div>The emergence of monkeypox virus (MPXV) as a global health threat has necessitated the rapid identification of novel antiviral therapeutics. Currently, no FDA-approved drugs are specifically designed against the disease. We used an in-house deep learning pharmacophore model for screening a library of 1974 FDA-approved drugs targeting the active site of MPXV DNA polymerase. Three drugs exhibited the strongest binding affinities, outperforming the control drug, Cidofovir diphosphate, and forming stable interactions with key active site residues. Among them, Paromomycin emerged as the most favourable drug, demonstrating stable, persistent, and adaptable interactions in molecular dynamics simulation. In parallel, we developed a novel automated peptide-generating AI pipeline that integrates active-site residues with knowledge-guided amino acid selection to generate and evaluate synthetic peptides. Cysteine-Phenylalanine-Cysteine (CFC), together with a panel of candidates, emerged through rational balancing of physicochemical properties and drug-likeness for accelerated therapeutic discovery. Synthetic peptides were evaluated to further understand the binding efficacies with DNA polymerase. CFC peptide demonstrated strong binding affinity (−8.08 kcal/mol) through stable interactions with key catalytic residues ASP549, ARG634 and LYS661, while MMGBSA analysis confirmed favourable binding energy (−33.02 kcal/mol). Consistent results in MD simulations indicate functional binding without destabilisation. Although ADMET predictions for CFC revealed limitations in permeability and oral bioavailability, its favourable binding profile and reduced predicted toxicity support its potential as a novel antiviral lead.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"200 ","pages":"Article 111372"},"PeriodicalIF":6.3,"publicationDate":"2025-12-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145707635","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

Bayesian parameter inference and uncertainty-informed sensitivity analysis in a 0D cardiovascular model for intraoperative hypotension 术中低血压的0D心血管模型的贝叶斯参数推断和不确定性敏感性分析

IF 6.3 2区医学 Q1 BIOLOGY

Computers in biology and medicine

Pub Date : 2025-12-06 DOI: 10.1016/j.compbiomed.2025.111371

Jan-Niklas Thiel , Marko Zlicar , Ulrich Steinseifer , Borut Kirn , Michael Neidlin

Computational cardiovascular models are promising tools for clinical decision support, particularly in complex conditions, such as intraoperative hypotension (IOH). IOH arises from different mechanisms, making treatment selection non-trivial. Patient-specific predictions require calibration, typically performed using classical optimization prone to parameter non-identifiability and lacking uncertainty quantification, hindering clinical translation. Consequently, Bayesian approaches are needed that facilitate parameter inference, sensitivity analysis, and uncertainty quantification in cardiovascular models.

We utilize Bayesian Markov chain Monte Carlo (MCMC) to estimate parameter distributions of a cardiovascular lumped parameter model (LPM) across different IOH scenarios. We demonstrate parameter non-uniqueness and its impact on sensitivity indices. We improve parameter reliability by incorporating clinical knowledge and measurement uncertainties. Continual learning of the model is achieved by sequential parameter updating as new patient data become available. We introduce an uncertainty-aware sensitivity analysis and compare it with a classical approach.

Classical optimization yielded many local solutions for IOH, with notably different sensitivities. MCMC distinguished different hypotension scenarios, such as those induced by impaired contractility or hypovolemia. Parameter uncertainty decreased by about 70 % with additional data, and by up to 94 % with sequential updating. Propagating uncertainties from MCMC through sensitivity analysis provided tighter credible intervals, resulting in more stable parameter rankings than the classical approach. The Bayesian approach revealed differences in model sensitivity and treatment suggestions across patient conditions, highlighting the potential to inform therapy planning.

Combining Bayesian inference with sequential updating and sensitivity analysis improves the reliability and identifiability of parameter estimates, enhancing the clinical utility of LPMs for therapy guidance.

计算心血管模型是临床决策支持的有前途的工具，特别是在复杂的情况下，如术中低血压（IOH）。IOH由不同的机制产生，使得治疗选择变得非常重要。特定患者的预测需要校准，通常使用经典优化执行，容易出现参数不可识别和缺乏不确定性量化，阻碍了临床翻译。因此，需要贝叶斯方法来促进心血管模型的参数推断、敏感性分析和不确定性量化。我们利用贝叶斯马尔可夫链蒙特卡罗（MCMC）来估计心血管集总参数模型（LPM）在不同IOH场景下的参数分布。论证了参数的非唯一性及其对灵敏度指标的影响。我们通过结合临床知识和测量不确定性来提高参数的可靠性。当新的患者数据可用时，通过顺序参数更新来实现模型的持续学习。我们引入了一种不确定性感知灵敏度分析方法，并将其与经典方法进行了比较。经典优化得到了IOH的许多局部解，灵敏度明显不同。MCMC区分了不同的低血压情况，如收缩力受损或低血容量引起的低血压。使用附加数据时，参数不确定性降低约70%，使用顺序更新时，参数不确定性降低高达94%。通过灵敏度分析从MCMC传播不确定性提供了更紧密的可信区间，从而产生比经典方法更稳定的参数排名。贝叶斯方法揭示了不同患者情况下模型敏感性和治疗建议的差异，强调了为治疗计划提供信息的潜力。贝叶斯推理与序列更新和敏感性分析相结合，提高了参数估计的可靠性和可识别性，增强了lpm在治疗指导中的临床应用。

{"title":"Bayesian parameter inference and uncertainty-informed sensitivity analysis in a 0D cardiovascular model for intraoperative hypotension","authors":"Jan-Niklas Thiel , Marko Zlicar , Ulrich Steinseifer , Borut Kirn , Michael Neidlin","doi":"10.1016/j.compbiomed.2025.111371","DOIUrl":"10.1016/j.compbiomed.2025.111371","url":null,"abstract":"<div><div>Computational cardiovascular models are promising tools for clinical decision support, particularly in complex conditions, such as intraoperative hypotension (IOH). IOH arises from different mechanisms, making treatment selection non-trivial. Patient-specific predictions require calibration, typically performed using classical optimization prone to parameter non-identifiability and lacking uncertainty quantification, hindering clinical translation. Consequently, Bayesian approaches are needed that facilitate parameter inference, sensitivity analysis, and uncertainty quantification in cardiovascular models.</div><div>We utilize Bayesian Markov chain Monte Carlo (MCMC) to estimate parameter distributions of a cardiovascular lumped parameter model (LPM) across different IOH scenarios. We demonstrate parameter non-uniqueness and its impact on sensitivity indices. We improve parameter reliability by incorporating clinical knowledge and measurement uncertainties. Continual learning of the model is achieved by sequential parameter updating as new patient data become available. We introduce an uncertainty-aware sensitivity analysis and compare it with a classical approach.</div><div>Classical optimization yielded many local solutions for IOH, with notably different sensitivities. MCMC distinguished different hypotension scenarios, such as those induced by impaired contractility or hypovolemia. Parameter uncertainty decreased by about 70 % with additional data, and by up to 94 % with sequential updating. Propagating uncertainties from MCMC through sensitivity analysis provided tighter credible intervals, resulting in more stable parameter rankings than the classical approach. The Bayesian approach revealed differences in model sensitivity and treatment suggestions across patient conditions, highlighting the potential to inform therapy planning.</div><div>Combining Bayesian inference with sequential updating and sensitivity analysis improves the reliability and identifiability of parameter estimates, enhancing the clinical utility of LPMs for therapy guidance.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"200 ","pages":"Article 111371"},"PeriodicalIF":6.3,"publicationDate":"2025-12-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145682752","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

Machine learning approaches to reveal pinealocyte changes in ageing and Alzheimer's disease 机器学习方法揭示衰老和阿尔茨海默病中松果体细胞的变化

IF 6.3 2区医学 Q1 BIOLOGY

Computers in biology and medicine

Pub Date : 2025-12-06 DOI: 10.1016/j.compbiomed.2025.111362

Inna Bukreeva , Olga Junemann , Francesca Palermo , Michela Fratini , Giuseppe Gigli , Dmitry Karpov , Sergey V. Saveliev , Alessia Cedola

This study investigates age-related changes in human pinealocytes and their association with Alzheimer's disease (AD). We analyzed calcified deposits in the pineal gland (PG) using a novel approach combining X-ray nano-holotomography and convolutional neural network-based image processing. Our approach used a U-net architecture for PG morphological feature segmentation, with particular emphasis on micro-calcifications in the pinealocyte cytoplasm, identified as primary calcification sites. The ratio of cytoplasmic deposits to number of pinealocytes in tissue volume showed a weak negative age-related tendency, but was not associated with AD. Our results show that pineal calcification may serve as a biomarker for synthetic activity, which declines with age. In addition, pathology-specific factors associated with AD may modulate pineal calcification patterns, potentially confounding age-related trends. Our findings contribute to a broader understanding of age-related neuropathology by providing insight into pineal alterations at the cellular level.

本研究探讨了人类松果体细胞的年龄相关变化及其与阿尔茨海默病（AD）的关系。我们使用一种结合x射线纳米全息摄影和基于卷积神经网络的图像处理的新方法分析了松果体（PG）中的钙化沉积物。我们的方法使用U-net架构进行PG形态特征分割，特别强调松果体细胞质中的微钙化，确定为初级钙化位点。细胞质沉积物与组织体积中松果体细胞数量的比值呈弱负的年龄相关趋势，但与AD无关。我们的研究结果表明，松果体钙化可能作为合成活性的生物标志物，随着年龄的增长而下降。此外，与AD相关的病理特异性因素可能会调节松果体钙化模式，潜在地混淆与年龄相关的趋势。我们的研究结果有助于更广泛地了解与年龄相关的神经病理学，提供了松果体在细胞水平上的改变。

{"title":"Machine learning approaches to reveal pinealocyte changes in ageing and Alzheimer's disease","authors":"Inna Bukreeva , Olga Junemann , Francesca Palermo , Michela Fratini , Giuseppe Gigli , Dmitry Karpov , Sergey V. Saveliev , Alessia Cedola","doi":"10.1016/j.compbiomed.2025.111362","DOIUrl":"10.1016/j.compbiomed.2025.111362","url":null,"abstract":"<div><div>This study investigates age-related changes in human pinealocytes and their association with Alzheimer's disease (AD). We analyzed calcified deposits in the pineal gland (PG) using a novel approach combining X-ray nano-holotomography and convolutional neural network-based image processing. Our approach used a U-net architecture for PG morphological feature segmentation, with particular emphasis on micro-calcifications in the pinealocyte cytoplasm, identified as primary calcification sites. The ratio of cytoplasmic deposits to number of pinealocytes in tissue volume showed a weak negative age-related tendency, but was not associated with AD. Our results show that pineal calcification may serve as a biomarker for synthetic activity, which declines with age. In addition, pathology-specific factors associated with AD may modulate pineal calcification patterns, potentially confounding age-related trends. Our findings contribute to a broader understanding of age-related neuropathology by providing insight into pineal alterations at the cellular level.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"200 ","pages":"Article 111362"},"PeriodicalIF":6.3,"publicationDate":"2025-12-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145682763","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

TinyML-enabled wearable system for early detection of knee osteoarthritis using ensemble gait classification 支持tinyml的可穿戴系统，用于使用整体步态分类早期检测膝关节骨关节炎

IF 6.3 2区医学 Q1 BIOLOGY

Computers in biology and medicine

Pub Date : 2025-12-06 DOI: 10.1016/j.compbiomed.2025.111345

Madhavan Bharanidivya , Samiappan Dhanalakshmi

Gait analysis commonly uses Inertial measurement unit (IMU) sensors because of their mobility, reliability, and applicability for real-time applications. Current research on gait phase identification mostly concentrates on enhancing classification accuracy through the use of many sensors or intricate algorithms, but often overlooks challenges such as sensor drift and recalibration in wearable medical settings. To perform real-time gait phase classification and identify irregularities linked to knee osteoarthritis (KOA), this work proposes a wearable sensor-based system that uses dual IMU sensors mounted on the femur and tibia. Walking, stair climbing, and sit-to-stand exercises were observed in a total of 15 participants, including both healthy people and KOA patients, resulting in 637 gait cycle samples. MPU9250 sensors were used to gather IMU data at 100 Hz, and min-max normalization and outlier elimination were used during processing. Real-time gait features (orientation, angular velocity, and acceleration) were extracted and used to train eight machine learning classifiers. Out of the above models being analysed, ensemble classifiers exhibited high performance. The classification accuracy of Random Forest was 97 %, that of Gb and Stacking was also 96 %. Statistical analysis using the Friedman test (χ² = 18.9, p < 0.01) and post-hoc Nemenyi comparisons confirmed Random Forest's significant advantage. After training, the models were paired with TinyML-ready hardware to ensure gait phase classification operates reliably and efficiently. Results demonstrate the feasibility of affordable, real-time wearable IMU devices for accurate gait monitoring and early KOA detection. The proposed method has immense potential for applications in online surveillance, clinical rehabilitation, and personalized mobility assessments.

步态分析通常使用惯性测量单元（IMU）传感器，因为它们具有移动性、可靠性和实时应用的适用性。目前对步态相位识别的研究主要集中在通过使用多种传感器或复杂的算法来提高分类精度，但往往忽视了可穿戴医疗环境中传感器漂移和重新校准等挑战。为了进行实时步态阶段分类并识别与膝骨关节炎（KOA）相关的不规则性，本研究提出了一种基于可穿戴传感器的系统，该系统使用安装在股骨和胫骨上的双IMU传感器。在包括健康人和KOA患者在内的15名参与者中，共观察了步行、爬楼梯和坐立运动，产生了637个步态周期样本。使用MPU9250传感器采集100 Hz的IMU数据，处理过程中采用最小-最大归一化和异常值消除。提取实时步态特征（方向、角速度和加速度）并用于训练8个机器学习分类器。在所分析的上述模型中，集成分类器表现出较高的性能。随机森林的分类准确率为97%，Gb和Stacking的分类准确率为96%。采用Friedman检验（χ2 = 18.9, p < 0.01）和事后Nemenyi比较的统计分析证实了Random Forest的显著优势。训练后，将模型与TinyML-ready硬件配对，以确保步态相位分类可靠高效地运行。结果表明，经济实惠的实时可穿戴IMU设备可用于准确的步态监测和早期KOA检测。该方法在在线监测、临床康复和个性化移动能力评估方面具有巨大的应用潜力。

{"title":"TinyML-enabled wearable system for early detection of knee osteoarthritis using ensemble gait classification","authors":"Madhavan Bharanidivya , Samiappan Dhanalakshmi","doi":"10.1016/j.compbiomed.2025.111345","DOIUrl":"10.1016/j.compbiomed.2025.111345","url":null,"abstract":"<div><div>Gait analysis commonly uses Inertial measurement unit (IMU) sensors because of their mobility, reliability, and applicability for real-time applications. Current research on gait phase identification mostly concentrates on enhancing classification accuracy through the use of many sensors or intricate algorithms, but often overlooks challenges such as sensor drift and recalibration in wearable medical settings. To perform real-time gait phase classification and identify irregularities linked to knee osteoarthritis (KOA), this work proposes a wearable sensor-based system that uses dual IMU sensors mounted on the femur and tibia. Walking, stair climbing, and sit-to-stand exercises were observed in a total of 15 participants, including both healthy people and KOA patients, resulting in 637 gait cycle samples. MPU9250 sensors were used to gather IMU data at 100 Hz, and min-max normalization and outlier elimination were used during processing. Real-time gait features (orientation, angular velocity, and acceleration) were extracted and used to train eight machine learning classifiers. Out of the above models being analysed, ensemble classifiers exhibited high performance. The classification accuracy of Random Forest was 97 %, that of Gb and Stacking was also 96 %. Statistical analysis using the Friedman test (χ<sup>2</sup> = 18.9, p < 0.01) and post-hoc Nemenyi comparisons confirmed Random Forest's significant advantage. After training, the models were paired with TinyML-ready hardware to ensure gait phase classification operates reliably and efficiently. Results demonstrate the feasibility of affordable, real-time wearable IMU devices for accurate gait monitoring and early KOA detection. The proposed method has immense potential for applications in online surveillance, clinical rehabilitation, and personalized mobility assessments.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"200 ","pages":"Article 111345"},"PeriodicalIF":6.3,"publicationDate":"2025-12-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145682751","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

Statistical, multi-scale and attention-based layer pooling of Wav2Vec-2 speech embeddings for Parkinson's disease detection 用于帕金森病检测的Wav2Vec-2语音嵌入的统计、多尺度和基于注意力的层池化

IF 6.3 2区医学 Q1 BIOLOGY

Computers in biology and medicine

Pub Date : 2025-12-05 DOI: 10.1016/j.compbiomed.2025.111368

Ondrej Klempir , Juliana Grand Mullerova , Radim Krupicka

Self-supervised pre-trained speech models such as wav2vec 2.0 provide rich frame-level embeddings that are increasingly used for clinical voice screening, including Parkinson's disease (PD). Optimally aggregating the frame-level embeddings is a significant task, with the underexplored question of how to best aggregate the frame-level embeddings into fixed-length utterance descriptors for the downstream task of binary classification (healthy controls (HC) vs. PD). To address this, our study compared three wav2vec 2.0 variants, the base model and two fine-tuned variants (one adapted for dysarthric corpora), across multiple layer depths, ten statistical aggregation functions (e.g., mean, median, quantiles), and two proposed advanced schemes (attention-based and multi-scale average pooling). We used the MDVR-KCL as a read-speech corpus (16 PD, 21 HC). Contrary to the expectation that sophisticated pooling would help, statistical aggregations such as mean pooling consistently provided better performance and robustness. Early representations (pre-Transformer and 1st Transformer block) were often the most informative, and mean aggregation produced relatively high, low-variance scores across models and depths. Attention and multi-scale pooling did not yield consistent gains. Moreover, wav2vec-based embeddings outperformed traditional acoustic baselines. Applying the supervised feature selection (ANOVA F-value) further improved performance, i.e. conservative selection (12 features) achieved a mean balanced accuracy of 0.87 and precision of 0.92, with top configurations exceeding a 0.93 balanced accuracy and had a precision of 1.0. The findings empirically support the continued use of mean pooling as a viable strategy for temporal aggregation of latent features for PD wav2vec-based detection.

自我监督的预训练语音模型，如wav2vec 2.0，提供丰富的帧级嵌入，越来越多地用于临床语音筛查，包括帕金森病（PD）。最佳地聚合帧级嵌入是一项重要的任务，如何最好地将帧级嵌入聚合到固定长度的话语描述符中，以用于二分类的下游任务（健康对照（HC）与PD），这一问题尚未得到充分的探讨。为了解决这个问题，我们的研究比较了三个wav2vec 2.0版本，基本模型和两个微调版本（一个适用于困难语料库），跨多层深度，十个统计聚合函数（例如，平均值，中位数，分位数），以及两个提出的高级方案（基于注意力和多尺度平均池化）。我们使用MDVR-KCL作为读-语音语料库（16 PD, 21 HC）。与期望复杂的池化会有所帮助相反，统计聚合（如平均池化）始终提供更好的性能和健壮性。早期的表示（pre-Transformer和1st Transformer块）通常是最具信息量的，并且平均聚合在模型和深度上产生相对较高，低方差的分数。注意和多尺度池化不能产生一致的收益。此外，基于wav2vec的嵌入优于传统的声学基线。应用监督特征选择（ANOVA f值）进一步提高了性能，即保守选择（12个特征）的平均平衡精度为0.87，精度为0.92，顶部配置的平衡精度超过0.93，精度为1.0。研究结果从经验上支持继续使用均值池作为一种可行的策略，用于PD wav2vec检测的潜在特征的时间聚合。

{"title":"Statistical, multi-scale and attention-based layer pooling of Wav2Vec-2 speech embeddings for Parkinson's disease detection","authors":"Ondrej Klempir , Juliana Grand Mullerova , Radim Krupicka","doi":"10.1016/j.compbiomed.2025.111368","DOIUrl":"10.1016/j.compbiomed.2025.111368","url":null,"abstract":"<div><div>Self-supervised pre-trained speech models such as wav2vec 2.0 provide rich frame-level embeddings that are increasingly used for clinical voice screening, including Parkinson's disease (PD). Optimally aggregating the frame-level embeddings is a significant task, with the underexplored question of how to best aggregate the frame-level embeddings into fixed-length utterance descriptors for the downstream task of binary classification (healthy controls (HC) vs. PD). To address this, our study compared three wav2vec 2.0 variants, the base model and two fine-tuned variants (one adapted for dysarthric corpora), across multiple layer depths, ten statistical aggregation functions (e.g., mean, median, quantiles), and two proposed advanced schemes (attention-based and multi-scale average pooling). We used the MDVR-KCL as a read-speech corpus (16 PD, 21 HC). Contrary to the expectation that sophisticated pooling would help, statistical aggregations such as mean pooling consistently provided better performance and robustness. Early representations (pre-Transformer and 1st Transformer block) were often the most informative, and mean aggregation produced relatively high, low-variance scores across models and depths. Attention and multi-scale pooling did not yield consistent gains. Moreover, wav2vec-based embeddings outperformed traditional acoustic baselines. Applying the supervised feature selection (ANOVA F-value) further improved performance, i.e. conservative selection (12 features) achieved a mean balanced accuracy of 0.87 and precision of 0.92, with top configurations exceeding a 0.93 balanced accuracy and had a precision of 1.0. The findings empirically support the continued use of mean pooling as a viable strategy for temporal aggregation of latent features for PD wav2vec-based detection.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"200 ","pages":"Article 111368"},"PeriodicalIF":6.3,"publicationDate":"2025-12-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145682764","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

Metabolic signatures of immune checkpoint inhibitor response in gynecologic cancers: Insights from flux balance analysis 妇科癌症中免疫检查点抑制剂反应的代谢特征：来自通量平衡分析的见解

IF 6.3 2区医学 Q1 BIOLOGY

Computers in biology and medicine

Pub Date : 2025-12-05 DOI: 10.1016/j.compbiomed.2025.111366

Gideon Idumah , Lin Li , Lamis Yehia , Haider Mahdi , Ying Ni

Modifiers of immune checkpoint inhibitor (ICI) responses in cancer patients are complex and remain poorly characterized, especially in gynecologic cancers. In this study, we explored fluxomic biomarkers that differentiate responders from non-responders to ICIs in a series of 49 patients with gynecologic cancers, including ovarian, cervical, and endometrial cancers. By applying metabolic enzyme expression as constraints, we utilized an objective-customizable flux balance analysis within a genome-scale metabolic model to predict the metabolic flux differences between responders versus non-responders of ICI treatment. We identified three reactions with consistent differential activity across all ten different optimization objectives: Succinate Dehydrogenase (SUCD1m) in the citric acid cycle, NADH: Guanosine-5-Phosphate Oxidoreductase (r0276) involved in purine catabolism, and Ornithine Transaminase Reversible, Mitochondrial (ORNTArm) in the urea cycle. Additionally, reactions within the folate cycle subsystem, particularly involving MTHFD2, demonstrated significance in distinguishing treatment responses, aligning with previous findings linking MTHFD2 to immune evasion and tumor progression. To further analyze the association between metabolic features and survival outcomes, we implemented machine learning models that integrate multi-omics data. Our model included clinical-pathologic, molecular-genomic features (gene expression, TGF-β score, immune cell abundance from transcriptomic deconvolution), and significant reaction fluxes. Our findings suggest that SUCD1m, MTHFDm and ORNTArm are important metabolic biomarkers that could serve as predictive indicators for ICI response and, if validated in a larger cohort, may guide the development of targeted therapies to enhance treatment efficacy for gynecologic cancer patients. This study highlights the use of genome-scale metabolic modeling to identify clinically relevant biomarkers and improve therapeutic strategies.

癌症患者免疫检查点抑制剂（ICI）反应的修饰因子是复杂的，并且仍然缺乏特征，特别是在妇科癌症中。在这项研究中，我们探索了49例妇科癌症（包括卵巢癌、子宫癌和子宫内膜癌）患者中区分ICIs应答者和无应答者的通量组生物标志物。通过将代谢酶表达作为约束条件，我们在基因组尺度代谢模型中利用可客观定制的通量平衡分析来预测ICI治疗反应者与无反应者之间的代谢通量差异。我们确定了在所有10个不同优化目标中具有一致差异活性的三个反应：柠檬酸循环中的琥珀酸脱氢酶（SUCD1m），参与嘌呤分解代谢的NADH：鸟苷-5-磷酸氧化还原酶（r0276），以及尿素循环中的鸟氨酸转氨酶可逆线粒体（ORNTArm）。此外，叶酸循环子系统中的反应，特别是涉及MTHFD2的反应，在区分治疗反应中具有重要意义，这与先前将MTHFD2与免疫逃避和肿瘤进展联系起来的研究结果一致。为了进一步分析代谢特征与生存结果之间的关系，我们实施了集成多组学数据的机器学习模型。我们的模型包括临床病理、分子基因组特征（基因表达、TGF-β评分、转录组反褶积产生的免疫细胞丰度）和显著的反应通量。我们的研究结果表明，SUCD1m、MTHFDm和ORNTArm是重要的代谢生物标志物，可以作为ICI反应的预测指标，如果在更大的队列中得到验证，可能会指导靶向治疗的开发，以提高妇科癌症患者的治疗效果。这项研究强调了基因组尺度代谢模型的使用，以确定临床相关的生物标志物和改进治疗策略。

{"title":"Metabolic signatures of immune checkpoint inhibitor response in gynecologic cancers: Insights from flux balance analysis","authors":"Gideon Idumah , Lin Li , Lamis Yehia , Haider Mahdi , Ying Ni","doi":"10.1016/j.compbiomed.2025.111366","DOIUrl":"10.1016/j.compbiomed.2025.111366","url":null,"abstract":"<div><div>Modifiers of immune checkpoint inhibitor (ICI) responses in cancer patients are complex and remain poorly characterized, especially in gynecologic cancers. In this study, we explored fluxomic biomarkers that differentiate responders from non-responders to ICIs in a series of 49 patients with gynecologic cancers, including ovarian, cervical, and endometrial cancers. By applying metabolic enzyme expression as constraints, we utilized an objective-customizable flux balance analysis within a genome-scale metabolic model to predict the metabolic flux differences between responders versus non-responders of ICI treatment. We identified three reactions with consistent differential activity across all ten different optimization objectives: Succinate Dehydrogenase (SUCD1m) in the citric acid cycle, NADH: Guanosine-5-Phosphate Oxidoreductase (r0276) involved in purine catabolism, and Ornithine Transaminase Reversible, Mitochondrial (ORNTArm) in the urea cycle. Additionally, reactions within the folate cycle subsystem, particularly involving MTHFD2, demonstrated significance in distinguishing treatment responses, aligning with previous findings linking MTHFD2 to immune evasion and tumor progression. To further analyze the association between metabolic features and survival outcomes, we implemented machine learning models that integrate multi-omics data. Our model included clinical-pathologic, molecular-genomic features (gene expression, TGF-β score, immune cell abundance from transcriptomic deconvolution), and significant reaction fluxes. Our findings suggest that SUCD1m, MTHFDm and ORNTArm are important metabolic biomarkers that could serve as predictive indicators for ICI response and, if validated in a larger cohort, may guide the development of targeted therapies to enhance treatment efficacy for gynecologic cancer patients. This study highlights the use of genome-scale metabolic modeling to identify clinically relevant biomarkers and improve therapeutic strategies.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"200 ","pages":"Article 111366"},"PeriodicalIF":6.3,"publicationDate":"2025-12-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145682749","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

Human brain hemodynamics for 3D ultrasound localization microscopy benchmarking 人脑血流动力学的三维超声定位显微镜标杆

IF 6.3 2区医学 Q1 BIOLOGY

Computers in biology and medicine

Pub Date : 2025-12-05 DOI: 10.1016/j.compbiomed.2025.111370

Juliette Reydet, Hugues Favre, Alexandre Dizeux, Nabil Haidour, Mathieu Pernot, Mickael Tanter, Clément Papadacci

Ultrasound Localization Microscopy (ULM) has emerged as a promising technique for imaging microvascular networks at subwavelength resolution. However, its 3D translation in clinics for complex organs like the brain remains limited due to technological and experimental challenges, including probe design constraints, motions artifacts, and both acoustic attenuation and aberrations caused by the skull. In this study, we present a fast and versatile simulation framework for 3D ULM based on a realistic human brain vasculature model that includes small vessels down to the precapillary scale, along with its hemodynamics driven by conservation and Murray's law. This novel framework enables comparison of ultrasound probe configurations, ULM algorithm performance, and experimental parameters such as microbubble (MB) concentration, subpixel motion, and skull-induced aberrations in transcranial imaging conditions. To illustrate a range of case scenarios, three matrix array probes were evaluated including a matrix probe with large elements combined with diverging lenses. We evaluated localization accuracy, tracking performance, velocity distribution and the extent of the field of view. As expected, the simulations also highlighted the negative impact of high MB concentration and motions artifacts on detection performance, as well as the significant effect of skull-induced aberrations. The proposed framework provides a robust interface for developing, testing and optimizing 3D ULM systems, with potential applications extending to other organs and clinical scenarios.

超声定位显微镜（ULM）已成为一种很有前途的亚波长分辨率微血管网络成像技术。然而，由于技术和实验方面的挑战，包括探头设计限制、运动人工制品以及头骨引起的声学衰减和畸变，它在临床上用于复杂器官（如大脑）的3D翻译仍然有限。在这项研究中，我们提出了一个快速和通用的3D ULM模拟框架，该框架基于现实的人类脑血管模型，包括小血管到毛细血管前尺度，以及由守恒和默里定律驱动的血流动力学。这种新颖的框架可以比较超声探头配置、ULM算法性能和实验参数，如微泡（MB）浓度、亚像素运动和经颅成像条件下的头骨诱导畸变。为了说明一系列的情况下，三个矩阵阵列探头进行了评估，包括矩阵探头与大元素与发散透镜相结合。我们评估了定位精度、跟踪性能、速度分布和视野范围。正如预期的那样，模拟还突出了高MB浓度和运动伪影对检测性能的负面影响，以及头骨诱导的畸变的显著影响。提出的框架为开发、测试和优化3D ULM系统提供了一个强大的接口，具有扩展到其他器官和临床场景的潜在应用。

{"title":"Human brain hemodynamics for 3D ultrasound localization microscopy benchmarking","authors":"Juliette Reydet, Hugues Favre, Alexandre Dizeux, Nabil Haidour, Mathieu Pernot, Mickael Tanter, Clément Papadacci","doi":"10.1016/j.compbiomed.2025.111370","DOIUrl":"10.1016/j.compbiomed.2025.111370","url":null,"abstract":"<div><div>Ultrasound Localization Microscopy (ULM) has emerged as a promising technique for imaging microvascular networks at subwavelength resolution. However, its 3D translation in clinics for complex organs like the brain remains limited due to technological and experimental challenges, including probe design constraints, motions artifacts, and both acoustic attenuation and aberrations caused by the skull. In this study, we present a fast and versatile simulation framework for 3D ULM based on a realistic human brain vasculature model that includes small vessels down to the precapillary scale, along with its hemodynamics driven by conservation and Murray's law. This novel framework enables comparison of ultrasound probe configurations, ULM algorithm performance, and experimental parameters such as microbubble (MB) concentration, subpixel motion, and skull-induced aberrations in transcranial imaging conditions. To illustrate a range of case scenarios, three matrix array probes were evaluated including a matrix probe with large elements combined with diverging lenses. We evaluated localization accuracy, tracking performance, velocity distribution and the extent of the field of view. As expected, the simulations also highlighted the negative impact of high MB concentration and motions artifacts on detection performance, as well as the significant effect of skull-induced aberrations. The proposed framework provides a robust interface for developing, testing and optimizing 3D ULM systems, with potential applications extending to other organs and clinical scenarios.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"200 ","pages":"Article 111370"},"PeriodicalIF":6.3,"publicationDate":"2025-12-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145682750","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