Computer methods and programs in biomedicine最新文献

A comprehensive benchmarking of a U-Net based model for midbrain auto-segmentation on transcranial sonography 基于 U-Net 的经颅超声中脑自动分割模型的综合基准测试。

IF 4.9 2区医学 Q1 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Computer methods and programs in biomedicine

Pub Date : 2024-11-13 DOI: 10.1016/j.cmpb.2024.108494

Hong-yu Kang , Wei Zhang , Shuai Li , Xinyi Wang , Yu Sun , Xin Sun , Fang-Xian Li , Chao Hou , Sai-kit Lam , Yong-ping Zheng

Background and objective

Transcranial sonography-based grading of Parkinson's Disease has gained increasing attention in recent years, and it is currently used for assistive differential diagnosis in some specialized centers. To this end, accurate midbrain segmentation is considered an important initial step. However, current practice is manual, time-consuming, and bias-prone due to the subjective nature. Relevant studies in the literature are scarce and lacks comprehensive model evaluations from application perspectives. Herein, we aimed to benchmark the best-performing U-Net model for objective, stable and robust midbrain auto-segmentation using transcranial sonography images.

Methods

A total of 584 patients who were suspected of Parkinson's Disease were retrospectively enrolled from Beijing Tiantan Hospital. The dataset was divided into training (n = 416), validation (n = 104), and testing (n = 64) sets. Three state-of-the-art deep-learning networks (U-Net, U-Net+++, and nnU-Net) were utilized to develop segmentation models, under 5-fold cross-validation and three randomization seeds for safeguarding model validity and stability. Model evaluation was conducted in testing set in three key aspects: (i) segmentation agreement using DICE coefficients (DICE), Intersection over Union (IoU), and Hausdorff Distance (HD); (ii) model stability using standard deviations of segmentation agreement metrics; (iii) prediction time efficiency, and (iv) model robustness against various degrees of ultrasound imaging noise produced by the salt-and-pepper noise and Gaussian noise.

Results

The nnU-Net achieved the best segmentation agreement (averaged DICE: 0.910, IoU: 0.836, HD: 2.793-mm) and time efficiency (1.456-s). Under mild noise corruption, the nnU-Net outperformed others with averaged scores of DICE (0.904), IoU (0.827), HD (2.941 mm) in the salt-and-pepper noise (signal-to-noise ratio, SNR = 0.95), and DICE (0.906), IoU (0.830), HD (2.967 mm) in the Gaussian noise (sigma value, σ = 0.1); by contrast, intriguingly, performance of the U-Net and U-Net+++ models were remarkably degraded. Under increasing levels of simulated noise corruption (SNR decreased from 0.95 to 0.75; σ increased from 0.1 to 0.5), the nnU-Net network exhibited marginal decline in segmentation agreement meanwhile yielding decent performance as if there were absence of noise corruption.

Conclusions

The nnU-Net model was the best-performing midbrain segmentation model in terms of segmentation agreement, stability, time efficiency and robustness, providing the community with an objective, effective and automated alternative. Moving forward, a multi-center multi-vendor study is warranted when it comes to clinical implementation.

背景和目的：近年来，基于经颅超声的帕金森病分级越来越受到关注，目前在一些专业中心用于辅助鉴别诊断。为此，准确的中脑分割被认为是重要的第一步。然而，目前的做法是手工操作，耗时长，而且由于主观性，容易出现偏差。文献中的相关研究很少，而且缺乏从应用角度对模型的全面评估。在此，我们旨在为使用经颅超声图像进行客观、稳定和稳健的中脑自动分割的最佳 U-Net 模型设定基准：方法：我们从北京天坛医院回顾性招募了 584 名疑似帕金森病患者。数据集分为训练集（n = 416）、验证集（n = 104）和测试集（n = 64）。利用三种最先进的深度学习网络（U-Net、U-Net+++ 和 nnU-Net）开发分割模型，并进行 5 倍交叉验证和三种随机化种子，以保障模型的有效性和稳定性。在测试集中从三个关键方面对模型进行了评估：(i) 使用 DICE 系数（DICE）、Intersection over Union（IoU）和 Hausdorff Distance（HD）进行的分割一致性评估；(ii) 使用分割一致性指标的标准偏差进行的模型稳定性评估；(iii) 预测时间效率评估；(iv) 模型对由椒盐噪声和高斯噪声产生的不同程度超声成像噪声的鲁棒性评估：nnU-Net 获得了最佳的分割一致性（平均 DICE：0.910，IoU：0.836，HD：2.793-mm）和时间效率（1.456-s）。在轻度噪声破坏情况下，nnU-Net 的表现优于其他网络，在椒盐噪声（信噪比 SNR = 0.95）中的平均得分分别为 DICE (0.904)、IoU (0.827)、HD (2.941 mm)。95），以及高斯噪声（西格玛值，σ = 0.1）下的 DICE (0.906)、IoU (0.830)、HD (2.967 mm)。在模拟噪声损坏水平不断提高的情况下（信噪比从 0.95 降至 0.75；σ 从 0.1 升至 0.5），nnU-Net 网络的分割一致性略有下降，但其性能却与没有噪声损坏时相当：nnU-Net模型是在分割一致性、稳定性、时间效率和鲁棒性方面表现最好的中脑分割模型，为社会提供了一个客观、有效和自动化的替代方案。展望未来，在临床应用方面，有必要进行多中心多供应商研究。

{"title":"A comprehensive benchmarking of a U-Net based model for midbrain auto-segmentation on transcranial sonography","authors":"Hong-yu Kang , Wei Zhang , Shuai Li , Xinyi Wang , Yu Sun , Xin Sun , Fang-Xian Li , Chao Hou , Sai-kit Lam , Yong-ping Zheng","doi":"10.1016/j.cmpb.2024.108494","DOIUrl":"10.1016/j.cmpb.2024.108494","url":null,"abstract":"<div><h3>Background and objective</h3><div>Transcranial sonography-based grading of Parkinson's Disease has gained increasing attention in recent years, and it is currently used for assistive differential diagnosis in some specialized centers. To this end, accurate midbrain segmentation is considered an important initial step. However, current practice is manual, time-consuming, and bias-prone due to the subjective nature. Relevant studies in the literature are scarce and lacks comprehensive model evaluations from application perspectives. Herein, we aimed to benchmark the best-performing U-Net model for objective, stable and robust midbrain auto-segmentation using transcranial sonography images.</div></div><div><h3>Methods</h3><div>A total of 584 patients who were suspected of Parkinson's Disease were retrospectively enrolled from Beijing Tiantan Hospital. The dataset was divided into training (<em>n</em> = 416), validation (<em>n</em> = 104), and testing (<em>n</em> = 64) sets. Three state-of-the-art deep-learning networks (U-Net, U-Net+++, and nnU-Net) were utilized to develop segmentation models, under 5-fold cross-validation and three randomization seeds for safeguarding model validity and stability. Model evaluation was conducted in testing set in three key aspects: (i) segmentation agreement using DICE coefficients (DICE), Intersection over Union (IoU), and Hausdorff Distance (HD); (ii) model stability using standard deviations of segmentation agreement metrics; (iii) prediction time efficiency, and (iv) model robustness against various degrees of ultrasound imaging noise produced by the salt-and-pepper noise and Gaussian noise.</div></div><div><h3>Results</h3><div>The nnU-Net achieved the best segmentation agreement (averaged DICE: 0.910, IoU: 0.836, HD: 2.793-mm) and time efficiency (1.456-s). Under mild noise corruption, the nnU-Net outperformed others with averaged scores of DICE (0.904), IoU (0.827), HD (2.941 mm) in the salt-and-pepper noise (signal-to-noise ratio, SNR = 0.95), and DICE (0.906), IoU (0.830), HD (2.967 mm) in the Gaussian noise (sigma value, σ = 0.1); by contrast, intriguingly, performance of the U-Net and U-Net+++ models were remarkably degraded. Under increasing levels of simulated noise corruption (SNR decreased from 0.95 to 0.75; σ increased from 0.1 to 0.5), the nnU-Net network exhibited marginal decline in segmentation agreement meanwhile yielding decent performance as if there were absence of noise corruption.</div></div><div><h3>Conclusions</h3><div>The nnU-Net model was the best-performing midbrain segmentation model in terms of segmentation agreement, stability, time efficiency and robustness, providing the community with an objective, effective and automated alternative. Moving forward, a multi-center multi-vendor study is warranted when it comes to clinical implementation.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"258 ","pages":"Article 108494"},"PeriodicalIF":4.9,"publicationDate":"2024-11-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142616196","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

DeepForest-HTP: A novel deep forest approach for predicting antihypertensive peptides DeepForest-HTP：预测抗高血压肽的新型深度森林方法。

IF 4.9 2区医学 Q1 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Computer methods and programs in biomedicine

Pub Date : 2024-11-12 DOI: 10.1016/j.cmpb.2024.108514

Qiyuan Bai , Hao Chen , Wenshuo Li , Lei Li , Junhao Li , Zhen Gao , Yuan Li , Xuhua Li , Bing Song

Hypertension is a major preventable risk factor for cardiovascular disease, affecting over 1.5 billion adults worldwide. Antihypertensive peptides (AHTPs) have gained attention as a natural therapeutic option with minimal side effects. This study proposes a Deep Forest-based machine learning framework for AHTP prediction, leveraging a multi-granularity cascade structure to enhance classification accuracy. We integrated data from BIOPEP-UWM and three previously used datasets, totaling 2000 peptide sequences, and introduced novel feature extraction methods to build a comprehensive dataset for model training.

This study represents the first application of Deep Forest for AHTP identification, demonstrating substantial classification performance advantages over traditional methods (e.g., SVM, CNN, and XGBoost) as well as recent mainstream prediction models (Ensemble-AHTPpred, CNN-SVM Ensemble, and mAHTPred). Requiring no complex manual feature engineering, the model adapts flexibly to various data needs, offering a novel perspective for efficient AHTP prediction and promising utility in hypertension management.

On the benchmark dataset, the model achieved high accuracy, sensitivity, and AUC, providing a robust tool for identifying safe and effective AHTPs. However, future efforts should incorporate larger and more diverse independent validation datasets to further improve the model and enhance its generalizability. Additionally, the model's predictive accuracy relies on known AHTP targets and sequence features, potentially limiting its ability to detect AHTPs with uncharacterized or atypical properties.

高血压是心血管疾病的主要可预防风险因素，影响着全球超过 15 亿成年人。抗高血压肽（AHTPs）作为一种副作用极小的天然疗法受到了人们的关注。本研究提出了一种基于深林的机器学习框架来预测 AHTP，利用多粒度级联结构来提高分类准确性。我们整合了来自 BIOPEP-UWM 和之前使用过的三个数据集的数据，共计 2000 个肽序列，并引入了新颖的特征提取方法，以建立一个用于模型训练的综合数据集。这项研究是 Deep Forest 在 AHTP 鉴定中的首次应用，与传统方法（如 SVM、CNN 和 XGBoost）以及最近的主流预测模型（Ensemble-AHTPpred、CNN-SVM Ensemble 和 mAHTPred）相比，Deep Forest 的分类性能具有很大优势。该模型无需复杂的人工特征工程，可灵活适应各种数据需求，为高效的 AHTP 预测提供了新的视角，并有望在高血压管理中发挥作用。在基准数据集上，该模型实现了较高的准确度、灵敏度和 AUC，为识别安全有效的 AHTPs 提供了一个强大的工具。不过，今后的工作应纳入更大、更多样化的独立验证数据集，以进一步改进该模型并提高其通用性。此外，该模型的预测准确性依赖于已知的 AHTP 靶点和序列特征，可能会限制其检测具有未表征或非典型特性的 AHTP 的能力。

{"title":"DeepForest-HTP: A novel deep forest approach for predicting antihypertensive peptides","authors":"Qiyuan Bai , Hao Chen , Wenshuo Li , Lei Li , Junhao Li , Zhen Gao , Yuan Li , Xuhua Li , Bing Song","doi":"10.1016/j.cmpb.2024.108514","DOIUrl":"10.1016/j.cmpb.2024.108514","url":null,"abstract":"<div><div>Hypertension is a major preventable risk factor for cardiovascular disease, affecting over 1.5 billion adults worldwide. Antihypertensive peptides (AHTPs) have gained attention as a natural therapeutic option with minimal side effects. This study proposes a Deep Forest-based machine learning framework for AHTP prediction, leveraging a multi-granularity cascade structure to enhance classification accuracy. We integrated data from BIOPEP-UWM and three previously used datasets, totaling 2000 peptide sequences, and introduced novel feature extraction methods to build a comprehensive dataset for model training.</div><div>This study represents the first application of Deep Forest for AHTP identification, demonstrating substantial classification performance advantages over traditional methods (e.g., SVM, CNN, and XGBoost) as well as recent mainstream prediction models (Ensemble-AHTPpred, CNN-SVM Ensemble, and mAHTPred). Requiring no complex manual feature engineering, the model adapts flexibly to various data needs, offering a novel perspective for efficient AHTP prediction and promising utility in hypertension management.</div><div>On the benchmark dataset, the model achieved high accuracy, sensitivity, and AUC, providing a robust tool for identifying safe and effective AHTPs. However, future efforts should incorporate larger and more diverse independent validation datasets to further improve the model and enhance its generalizability. Additionally, the model's predictive accuracy relies on known AHTP targets and sequence features, potentially limiting its ability to detect AHTPs with uncharacterized or atypical properties.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"258 ","pages":"Article 108514"},"PeriodicalIF":4.9,"publicationDate":"2024-11-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142643498","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Positional encoding-guided transformer-based multiple instance learning for histopathology whole slide images classification 基于位置编码引导变换器的多实例学习，用于组织病理学整张切片图像分类。

IF 4.9 2区医学 Q1 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Computer methods and programs in biomedicine

Pub Date : 2024-11-09 DOI: 10.1016/j.cmpb.2024.108491

Jun Shi , Dongdong Sun , Kun Wu , Zhiguo Jiang , Xue Kong , Wei Wang , Haibo Wu , Yushan Zheng

Background and objectives:

Whole slide image (WSI) classification is of great clinical significance in computer-aided pathological diagnosis. Due to the high cost of manual annotation, weakly supervised WSI classification methods have gained more attention. As the most representative, multiple instance learning (MIL) generally aggregates the predictions or features of the patches within a WSI to achieve the slide-level classification under the weak supervision of WSI labels. However, most existing MIL methods ignore spatial position relationships of the patches, which is likely to strengthen the discriminative ability of WSI-level features.

Methods:

In this paper, we propose a novel positional encoding-guided transformer-based multiple instance learning (PEGTB-MIL) method for histopathology WSI classification. It aims to encode the spatial positional property of the patch into its corresponding semantic features and explore the potential correlation among the patches for improving the WSI classification performance. Concretely, the deep features of the patches in WSI are first extracted and simultaneously a position encoder is used to encode the spatial 2D positional information of the patches into the spatial-aware features. After incorporating the semantic features and spatial embeddings, multi-head self-attention (MHSA) is applied to explore the contextual and spatial dependencies of the fused features. Particularly, we introduce an auxiliary reconstruction task to enhance the spatial–semantic consistency and generalization ability of features.

Results:

The proposed method is evaluated on two public benchmark TCGA datasets (TCGA-LUNG and TCGA-BRCA) and two in-house clinical datasets (USTC-EGFR and USTC-GIST). Experimental results validate it is effective in the tasks of cancer subtyping and gene mutation status prediction. In the test stage, the proposed PEGTB-MIL outperforms the other state-of-the-art methods and respectively achieves 97.13±0.34%, 86.74±2.64%, 83.25±1.65%, and 72.52±1.63% of the area under the receiver operating characteristic (ROC) curve (AUC).

Conclusion:

PEGTB-MIL utilizes positional encoding to effectively guide and reinforce MIL, leading to enhanced performance on downstream WSI classification tasks. Specifically, the introduced auxiliary reconstruction module adeptly preserves the spatial–semantic consistency of patch features. More significantly, this study investigates the relationship between position information and disease diagnosis and presents a promising avenue for further research.

背景和目的：全切片图像（WSI）分类在计算机辅助病理诊断中具有重要的临床意义。由于人工标注成本高昂，弱监督 WSI 分类方法受到越来越多的关注。最具代表性的是多实例学习（Multiple instance learning，MIL），一般是在 WSI 标签的弱监督下，将 WSI 中的斑块预测或特征汇总，实现玻片级分类。然而，现有的 MIL 方法大多忽略了斑块的空间位置关系，而这很可能会加强 WSI 级特征的判别能力：本文提出了一种用于组织病理学 WSI 分类的新型位置编码引导的基于变换器的多实例学习（PEGTB-MIL）方法。该方法旨在将斑块的空间位置属性编码为其相应的语义特征，并探索斑块之间潜在的相关性，以提高 WSI 分类性能。具体来说，首先提取 WSI 中补丁的深层特征，同时使用位置编码器将补丁的空间二维位置信息编码为空间感知特征。在整合语义特征和空间嵌入后，多头自注意（MHSA）将用于探索融合特征的上下文和空间依赖关系。特别是，我们引入了一项辅助重建任务，以增强特征的空间语义一致性和泛化能力：结果：我们在两个公共基准 TCGA 数据集（TCGA-LUNG 和 TCGA-BRCA）和两个内部临床数据集（USTC-EGFR 和 USTC-GIST）上对所提出的方法进行了评估。实验结果验证了它在癌症亚型划分和基因突变状态预测任务中的有效性。在测试阶段，所提出的 PEGTB-MIL 优于其他最先进的方法，分别达到了 97.13±0.34%、86.74±2.64%、83.25±1.65% 和 72.52±1.63% 的接收者操作特征曲线（ROC）下面积（AUC）：结论：PEGTB-MIL 利用位置编码有效地指导和加强了 MIL，从而提高了下游 WSI 分类任务的性能。具体来说，引入的辅助重构模块能够很好地保留斑块特征的空间语义一致性。更重要的是，这项研究探讨了位置信息与疾病诊断之间的关系，为进一步的研究提供了一个前景广阔的途径。

{"title":"Positional encoding-guided transformer-based multiple instance learning for histopathology whole slide images classification","authors":"Jun Shi , Dongdong Sun , Kun Wu , Zhiguo Jiang , Xue Kong , Wei Wang , Haibo Wu , Yushan Zheng","doi":"10.1016/j.cmpb.2024.108491","DOIUrl":"10.1016/j.cmpb.2024.108491","url":null,"abstract":"<div><h3>Background and objectives:</h3><div>Whole slide image (WSI) classification is of great clinical significance in computer-aided pathological diagnosis. Due to the high cost of manual annotation, weakly supervised WSI classification methods have gained more attention. As the most representative, multiple instance learning (MIL) generally aggregates the predictions or features of the patches within a WSI to achieve the slide-level classification under the weak supervision of WSI labels. However, most existing MIL methods ignore spatial position relationships of the patches, which is likely to strengthen the discriminative ability of WSI-level features.</div></div><div><h3>Methods:</h3><div>In this paper, we propose a novel positional encoding-guided transformer-based multiple instance learning (PEGTB-MIL) method for histopathology WSI classification. It aims to encode the spatial positional property of the patch into its corresponding semantic features and explore the potential correlation among the patches for improving the WSI classification performance. Concretely, the deep features of the patches in WSI are first extracted and simultaneously a position encoder is used to encode the spatial 2D positional information of the patches into the spatial-aware features. After incorporating the semantic features and spatial embeddings, multi-head self-attention (MHSA) is applied to explore the contextual and spatial dependencies of the fused features. Particularly, we introduce an auxiliary reconstruction task to enhance the spatial–semantic consistency and generalization ability of features.</div></div><div><h3>Results:</h3><div>The proposed method is evaluated on two public benchmark TCGA datasets (TCGA-LUNG and TCGA-BRCA) and two in-house clinical datasets (USTC-EGFR and USTC-GIST). Experimental results validate it is effective in the tasks of cancer subtyping and gene mutation status prediction. In the test stage, the proposed PEGTB-MIL outperforms the other state-of-the-art methods and respectively achieves 97.13±0.34%, 86.74±2.64%, 83.25±1.65%, and 72.52±1.63% of the area under the receiver operating characteristic (ROC) curve (AUC).</div></div><div><h3>Conclusion:</h3><div>PEGTB-MIL utilizes positional encoding to effectively guide and reinforce MIL, leading to enhanced performance on downstream WSI classification tasks. Specifically, the introduced auxiliary reconstruction module adeptly preserves the spatial–semantic consistency of patch features. More significantly, this study investigates the relationship between position information and disease diagnosis and presents a promising avenue for further research.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"258 ","pages":"Article 108491"},"PeriodicalIF":4.9,"publicationDate":"2024-11-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142643509","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

Integrating radiomic and 3D autoencoder-based features for Non-Small Cell Lung Cancer survival analysis 整合放射线组学和基于三维自动编码器的非小细胞肺癌生存分析特征。

IF 4.9 2区医学 Q1 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Computer methods and programs in biomedicine

Pub Date : 2024-11-08 DOI: 10.1016/j.cmpb.2024.108496

Meri Ferretti , Valentina D.A. Corino

Background and objectives

The aim of this study is to develop a radiomic and deep learning-based signature for survival analysis of patients with Non-Small Cell Lung Cancer.

Methods

Four-hundred twenty-two patients from “Lung1” dataset were included in the study. A 3D convolutional autoencoder (AE) was built and features from the latent space extracted for further analysis. Radiomic features were derived from the 3D volume of the tumor region using PyRadiomics. Both radiomic and AE-based features underwent feature selection, by removing: i) highly correlated and ii) constant features. The selected variables were then used to derive both mono-domain (radiomics, AE and clinic) and multi-domain signatures fitting a Cox Proportional Hazard model with LASSO penalization and evaluated considering the concordance (C)-index as performance metric.

Results

Both mono-domain and multi-domain signatures could significantly differentiate high risk from low risk patients. Among the mono-domain signatures, the highest hazard ratio (HR) in the test set was obtained using radiomics (HR = 1.5428) followed by the AE-based signature (HR = 1.5012) and the clinical signature (HR = 1.4770). The best overall performance was achieved by combining all three signatures, resulting in the highest HR (HR = 1.7383), while the combination of AE-based and clinical signatures yielded the highest C-index (C-index = 0.6309).

Conclusions

These preliminary results show that combining information carried by AE, radiomic and clinical domain shows potential for improving the prediction of overall survival in NSCLC patients.

背景与目的本研究旨在为非小细胞肺癌患者的生存分析开发一种基于放射学和深度学习的特征：研究纳入了 "Lung1 "数据集中的 422 名患者。建立了一个三维卷积自动编码器（AE），并从潜在空间中提取特征进行进一步分析。使用 PyRadiomics 从肿瘤区域的三维体积中提取放射线特征。基于辐射组学和 AE 的特征都经过了特征选择，去除：i）高度相关特征和 ii）恒定特征。然后利用所选变量得出单域（放射组学、AE和临床）和多域特征，并利用LASSO惩罚拟合考克斯比例危险模型，以一致性（C）指数作为性能指标进行评估：单域和多域特征都能明显区分高风险和低风险患者。在单域特征中，放射组学在测试集中获得的危险比（HR）最高（HR = 1.5428），其次是基于 AE 的特征（HR = 1.5012）和临床特征（HR = 1.4770）。将所有三种特征结合在一起可获得最佳整体性能，从而产生最高的 HR（HR = 1.7383），而将基于 AE 的特征和临床特征结合在一起可产生最高的 C 指数（C 指数 = 0.6309）：这些初步结果表明，将AE、放射组学和临床领域所携带的信息结合起来，在提高NSCLC患者的总生存率预测方面具有潜力。

{"title":"Integrating radiomic and 3D autoencoder-based features for Non-Small Cell Lung Cancer survival analysis","authors":"Meri Ferretti , Valentina D.A. Corino","doi":"10.1016/j.cmpb.2024.108496","DOIUrl":"10.1016/j.cmpb.2024.108496","url":null,"abstract":"<div><h3>Background and objectives</h3><div>The aim of this study is to develop a radiomic and deep learning-based signature for survival analysis of patients with Non-Small Cell Lung Cancer.</div></div><div><h3>Methods</h3><div>Four-hundred twenty-two patients from “Lung1” dataset were included in the study. A 3D convolutional autoencoder (AE) was built and features from the latent space extracted for further analysis. Radiomic features were derived from the 3D volume of the tumor region using PyRadiomics. Both radiomic and AE-based features underwent feature selection, by removing: i) highly correlated and ii) constant features. The selected variables were then used to derive both mono-domain (radiomics, AE and clinic) and multi-domain signatures fitting a Cox Proportional Hazard model with LASSO penalization and evaluated considering the concordance (C)-index as performance metric.</div></div><div><h3>Results</h3><div>Both mono-domain and multi-domain signatures could significantly differentiate high risk from low risk patients. Among the mono-domain signatures, the highest hazard ratio (HR) in the test set was obtained using radiomics (HR = 1.5428) followed by the AE-based signature (HR = 1.5012) and the clinical signature (HR = 1.4770). The best overall performance was achieved by combining all three signatures, resulting in the highest HR (HR = 1.7383), while the combination of AE-based and clinical signatures yielded the highest C-index (C-index = 0.6309).</div></div><div><h3>Conclusions</h3><div>These preliminary results show that combining information carried by AE, radiomic and clinical domain shows potential for improving the prediction of overall survival in NSCLC patients.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"258 ","pages":"Article 108496"},"PeriodicalIF":4.9,"publicationDate":"2024-11-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142647119","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Computer model coupling hemodynamics and oxygen transport in the coronary capillary network: Pulsatile vs. non-pulsatile analysis 冠状动脉毛细血管网络中血液动力学与氧运输耦合的计算机模型：脉动与非脉动分析。

IF 4.9 2区医学 Q1 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Computer methods and programs in biomedicine

Pub Date : 2024-11-08 DOI: 10.1016/j.cmpb.2024.108486

Haifeng Wang , Jenny S. Choy , Ghassan S. Kassab , Lik-Chuan Lee

Background and Objective:

Oxygen transport in the heart is crucial, and its impairment can lead to pathological conditions such as hypoxia, ischemia, and heart failure. However, investigating oxygen transport in the heart using in vivo measurements is difficult due to the small size of the coronary capillaries and their deep embedding within the heart wall.

Methods:

In this study, we developed a novel computational modeling framework that integrates a 0-D hemodynamic model with a 1-D mass transport model to simulate oxygen transport in/across the coronary capillary network.

Results:

The model predictions agree with analytical solutions and experimental measurements. The framework is used to simulate the effects of pulsatile vs. non-pulsatile behavior of the capillary hemodynamics on oxygen-related metrics such as the myocardial oxygen consumption (

{MVO}_{2}

) and oxygen extraction ratio (OER). Compared to simulations that consider (physiological) pulsatile behaviors of the capillary hemodynamics, the OER is underestimated by less than 9% and the

{MVO}_{2}

is overestimated by less than 5% when the pulsatile behaviors are ignored in the simulations. Statistical analyses show that model predictions of oxygen-related quantities and spatial distribution of oxygen without consideration of the pulsatile behaviors do not significantly differ from those that considered such behaviors (p-values

> 0.05

).

Conclusions:

This finding provides the basis for reducing the model complexity by ignoring the pulsatility of coronary capillary hemodynamics in the computational framework without a substantial loss of accuracy when predicting oxygen-related metrics.

背景和目的：心脏中的氧输送至关重要，其受损可导致缺氧、缺血和心力衰竭等病理状况。然而，由于冠状动脉毛细血管较小，且深埋于心壁内，利用体内测量来研究心脏内的氧输送十分困难：在这项研究中，我们开发了一个新颖的计算建模框架，将 0-D 血流动力学模型与 1-D 质量传输模型相结合，模拟冠状动脉毛细血管网络内/外的氧气传输：结果：模型预测结果与分析解法和实验测量结果一致。该框架用于模拟毛细血管血流动力学的搏动与非搏动行为对心肌耗氧量（MVO2）和氧萃取率（OER）等氧相关指标的影响。与考虑了毛细血管血液动力学（生理）搏动行为的模拟相比，如果在模拟中忽略搏动行为，OER 被低估了不到 9%，MVO2 被高估了不到 5%。统计分析显示，未考虑脉动行为的模型对氧气相关量和氧气空间分布的预测与考虑了脉动行为的模型没有显著差异（P 值大于 0.05）：这一发现为在计算框架中忽略冠状动脉毛细血管血流动力学的搏动性以降低模型的复杂性提供了依据，而不会对预测氧相关指标的准确性造成重大损失。

{"title":"Computer model coupling hemodynamics and oxygen transport in the coronary capillary network: Pulsatile vs. non-pulsatile analysis","authors":"Haifeng Wang , Jenny S. Choy , Ghassan S. Kassab , Lik-Chuan Lee","doi":"10.1016/j.cmpb.2024.108486","DOIUrl":"10.1016/j.cmpb.2024.108486","url":null,"abstract":"<div><h3>Background and Objective:</h3><div>Oxygen transport in the heart is crucial, and its impairment can lead to pathological conditions such as hypoxia, ischemia, and heart failure. However, investigating oxygen transport in the heart using <em>in vivo</em> measurements is difficult due to the small size of the coronary capillaries and their deep embedding within the heart wall.</div></div><div><h3>Methods:</h3><div>In this study, we developed a novel computational modeling framework that integrates a 0-D hemodynamic model with a 1-D mass transport model to simulate oxygen transport in/across the coronary capillary network.</div></div><div><h3>Results:</h3><div>The model predictions agree with analytical solutions and experimental measurements. The framework is used to simulate the effects of pulsatile vs. non-pulsatile behavior of the capillary hemodynamics on oxygen-related metrics such as the myocardial oxygen consumption (<span><math><msub><mrow><mtext>MVO</mtext></mrow><mrow><mn>2</mn></mrow></msub></math></span>) and oxygen extraction ratio (OER). Compared to simulations that consider (physiological) pulsatile behaviors of the capillary hemodynamics, the OER is underestimated by less than 9% and the <span><math><msub><mrow><mtext>MVO</mtext></mrow><mrow><mn>2</mn></mrow></msub></math></span> is overestimated by less than 5% when the pulsatile behaviors are ignored in the simulations. Statistical analyses show that model predictions of oxygen-related quantities and spatial distribution of oxygen without consideration of the pulsatile behaviors do not significantly differ from those that considered such behaviors (p-values <span><math><mrow><mo>></mo><mn>0</mn><mo>.</mo><mn>05</mn></mrow></math></span>).</div></div><div><h3>Conclusions:</h3><div>This finding provides the basis for reducing the model complexity by ignoring the pulsatility of coronary capillary hemodynamics in the computational framework without a substantial loss of accuracy when predicting oxygen-related metrics.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"258 ","pages":"Article 108486"},"PeriodicalIF":4.9,"publicationDate":"2024-11-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142643484","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 computational method to predict cerebral perfusion flow after endovascular treatment based on invasive pressure and resistance 基于侵入压力和阻力预测血管内治疗后脑灌注流量的计算方法。

IF 4.9 2区医学 Q1 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Computer methods and programs in biomedicine

Pub Date : 2024-11-08 DOI: 10.1016/j.cmpb.2024.108510

Xi Zhao , Li Bai , Raynald , Jie He , Bin Han , Xiaotong Xu , Zhongrong Miao , Dapeng Mo

Background and objective

Predicting post-operative flow is essential for assessing the risk of adverse events in cerebrovascular stenosis patients following endovascular treatment (EVT). This study aimed to evaluate the accuracy of the CFD simulation model in predicting post-operative velocity, flow and pressure distal to a stenosis, based on cerebrovascular microcirculatory resistance.

Methods

The patient-specific models of the extracranial and intracranial arteries were reconstructed. The cerebrovascular microcirculatory resistance was applied to estimate post-operative blood velocity and flow rates. Pearson correlation and Bland-Altman analyses were used to evaluate the correlation and agreement between CFD calculations and transcranial Doppler (TCD) measurements.

Results

There was a strong correlation between CFD- and TCD-based mean velocities (r = 0.7733; P = 0.0002), with volume flow measured by both methods also showing robust correlation (r = 0.8621; P < 0.0001). Additionally, agreement was found between mean velocities determined by CFD simulation and those estimated by TCD (P = 0.2446, mean difference −4.2089; limits of agreement -11.5764 to 3.1586). However, agreement between volume flow from CFD simulations and TCD was less consistent (P = 0.0387, mean difference -0.3272, limits of agreement -0.9276 to 0.2731).

Conclusions

The computational method used in this study enables the prediction of hemodynamic changes and offers valuable support in tailoring treatment strategies for cerebrovascular stenosis lesions.

背景和目的：预测术后血流对于评估脑血管狭窄患者接受血管内治疗（EVT）后发生不良事件的风险至关重要。本研究旨在评估基于脑血管微循环阻力的 CFD 模拟模型在预测术后狭窄远端速度、流量和压力方面的准确性：方法：重建了颅外动脉和颅内动脉的患者特异性模型。应用脑血管微循环阻力估算术后血流速度和流速。采用皮尔逊相关分析和布兰德-阿尔特曼分析评估 CFD 计算与经颅多普勒（TCD）测量之间的相关性和一致性：基于 CFD 和 TCD 的平均速度之间存在很强的相关性（r = 0.7733；P = 0.0002），两种方法测量的血流量也显示出很强的相关性（r = 0.8621；P < 0.0001）。此外，CFD 模拟确定的平均速度与 TCD 估算的平均速度之间也存在一致性（P = 0.2446，平均差 -4.2089；一致性范围 -11.5764 至 3.1586）。然而，CFD 模拟和 TCD 估算的体积流量之间的一致性较差（P = 0.0387，平均差 -0.3272，一致性范围 -0.9276 至 0.2731）：本研究中使用的计算方法可以预测血流动力学变化，为制定脑血管狭窄病变的治疗策略提供有价值的支持。

{"title":"A computational method to predict cerebral perfusion flow after endovascular treatment based on invasive pressure and resistance","authors":"Xi Zhao , Li Bai , Raynald , Jie He , Bin Han , Xiaotong Xu , Zhongrong Miao , Dapeng Mo","doi":"10.1016/j.cmpb.2024.108510","DOIUrl":"10.1016/j.cmpb.2024.108510","url":null,"abstract":"<div><h3>Background and objective</h3><div>Predicting post-operative flow is essential for assessing the risk of adverse events in cerebrovascular stenosis patients following endovascular treatment (EVT). This study aimed to evaluate the accuracy of the CFD simulation model in predicting post-operative velocity, flow and pressure distal to a stenosis, based on cerebrovascular microcirculatory resistance.</div></div><div><h3>Methods</h3><div>The patient-specific models of the extracranial and intracranial arteries were reconstructed. The cerebrovascular microcirculatory resistance was applied to estimate post-operative blood velocity and flow rates. Pearson correlation and Bland-Altman analyses were used to evaluate the correlation and agreement between CFD calculations and transcranial Doppler (TCD) measurements.</div></div><div><h3>Results</h3><div>There was a strong correlation between CFD- and TCD-based mean velocities (<em>r</em> = 0.7733; <em>P</em> = 0.0002), with volume flow measured by both methods also showing robust correlation (<em>r</em> = 0.8621; <em>P</em> < 0.0001). Additionally, agreement was found between mean velocities determined by CFD simulation and those estimated by TCD (<em>P</em> = 0.2446, mean difference −4.2089; limits of agreement -11.5764 to 3.1586). However, agreement between volume flow from CFD simulations and TCD was less consistent (<em>P</em> = 0.0387, mean difference -0.3272, limits of agreement -0.9276 to 0.2731).</div></div><div><h3>Conclusions</h3><div>The computational method used in this study enables the prediction of hemodynamic changes and offers valuable support in tailoring treatment strategies for cerebrovascular stenosis lesions.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"258 ","pages":"Article 108510"},"PeriodicalIF":4.9,"publicationDate":"2024-11-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142643466","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

Localizing the seizure onset zone and predicting the surgery outcomes in patients with drug-resistant epilepsy: A new approach based on the causal network 确定耐药性癫痫患者的发作起始区并预测手术结果：基于因果网络的新方法。

IF 4.9 2区医学 Q1 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Computer methods and programs in biomedicine

Pub Date : 2024-11-08 DOI: 10.1016/j.cmpb.2024.108483

Mingming Chen, Kunlin Guo, Kai Lu, Kunying Meng, Junfeng Lu, Yajing Pang, Lipeng Zhang, Yuxia Hu, Renping Yu, Rui Zhang

Background and Objective:

Accurate localization of the seizure onset zone (SOZ) is crucial for surgical treatment in patients with drug-resistant epilepsy (DRE). However, clinical identification of SOZ often relies on physician experience and has a certain subjectivity. Therefore, it is emergent to develop quantitative computational tools to assist clinicians in identifying SOZ.

Methods:

We conduct a retrospective study on intracranial electroencephalography (iEEG) data from 46 patients with DRE. The interactions between different brain regions are quantified by using the phase transfer entropy (PTE), based on which the causal influence index (CII) is proposed to quantify the degree of influence of nodes on the network. Subsequently, the features extracted by the CII are used to construct a random forest classification model, which the performance in identifying SOZ and the generalizability are validated in patients with successful surgeries. Then, based on the CII features of the clinically labeled SOZ, a logistic regression prediction model is constructed to predict the probability of surgical success. The statistical analysis between patients with successful and failed surgery is conducted with the Mann–Whitney U test. Finally, the consistency between the predicted SOZ and the clinically labeled SOZ is verified across different Engel classes.

Results:

The classification model combining the low-frequency and high-frequency features can achieve an accuracy of 82.18% (sensitivity: 85.01%, specificity: 79.69%) and an area under curve (AUC) of 0.90 in identifying SOZ. Furthermore, the model exhibits strong generalizability in identifying SOZ in patients with MRI lesional and non-lesional, as well as those implanted with electrocorticography (ECOG) and stereotactic EEG (SEEG) electrodes. Moreover, the prediction model could achieve an average accuracy of 79.8% and an AUC of 0.84. Of note, the prediction of surgical success probability is significant between patients with successful and failed surgeries (P

<

0.001). Correspondingly, the highest consistency between model-predicted SOZ and clinically labeled SOZ can be observed in patients with successful surgeries, but this consistency gradually decreases with increasing Engel classes.

Conclusions:

These results demonstrate that the CII may be a potential biomarker for identifying the SOZ in patients with DRE, which may provide a new perspective for the treatment of epilepsy.

背景和目的：准确定位癫痫发作区（SOZ）对于耐药性癫痫（DRE）患者的手术治疗至关重要。然而，临床上对 SOZ 的识别往往依赖于医生的经验，具有一定的主观性。因此，开发定量计算工具来协助临床医生识别SOZ成为当务之急：我们对 46 名 DRE 患者的颅内脑电图（iEEG）数据进行了回顾性研究。利用相位传递熵（PTE）量化不同脑区之间的相互作用，并在此基础上提出因果影响指数（CII），以量化节点对网络的影响程度。随后，利用 CII 提取的特征构建随机森林分类模型，并在手术成功的患者中验证其识别 SOZ 的性能和普适性。然后，根据临床标记的 SOZ 的 CII 特征，构建逻辑回归预测模型，预测手术成功的概率。通过 Mann-Whitney U 检验对手术成功和失败的患者进行统计分析。最后，在不同的恩格尔等级中验证了预测的 SOZ 与临床标记的 SOZ 之间的一致性：结果：结合低频和高频特征的分类模型在识别 SOZ 方面的准确率为 82.18%（灵敏度：85.01%，特异性：79.69%），曲线下面积（AUC）为 0.90。此外，该模型在识别磁共振成像病变和非病变患者，以及植入皮层电图（ECOG）和立体定向脑电图（SEEG）电极的患者的 SOZ 方面具有很强的普适性。此外，该预测模型的平均准确率为 79.8%，AUC 为 0.84。值得注意的是，在手术成功和手术失败的患者之间，手术成功概率的预测结果具有显著性（结论：CII 预测的手术成功概率与手术失败概率之间具有显著性差异）：这些结果表明，CII可能是识别DRE患者SOZ的潜在生物标志物，为癫痫治疗提供了新的视角。

{"title":"Localizing the seizure onset zone and predicting the surgery outcomes in patients with drug-resistant epilepsy: A new approach based on the causal network","authors":"Mingming Chen, Kunlin Guo, Kai Lu, Kunying Meng, Junfeng Lu, Yajing Pang, Lipeng Zhang, Yuxia Hu, Renping Yu, Rui Zhang","doi":"10.1016/j.cmpb.2024.108483","DOIUrl":"10.1016/j.cmpb.2024.108483","url":null,"abstract":"<div><h3>Background and Objective:</h3><div>Accurate localization of the seizure onset zone (SOZ) is crucial for surgical treatment in patients with drug-resistant epilepsy (DRE). However, clinical identification of SOZ often relies on physician experience and has a certain subjectivity. Therefore, it is emergent to develop quantitative computational tools to assist clinicians in identifying SOZ.</div></div><div><h3>Methods:</h3><div>We conduct a retrospective study on intracranial electroencephalography (iEEG) data from 46 patients with DRE. The interactions between different brain regions are quantified by using the phase transfer entropy (PTE), based on which the causal influence index (CII) is proposed to quantify the degree of influence of nodes on the network. Subsequently, the features extracted by the CII are used to construct a random forest classification model, which the performance in identifying SOZ and the generalizability are validated in patients with successful surgeries. Then, based on the CII features of the clinically labeled SOZ, a logistic regression prediction model is constructed to predict the probability of surgical success. The statistical analysis between patients with successful and failed surgery is conducted with the Mann–Whitney U test. Finally, the consistency between the predicted SOZ and the clinically labeled SOZ is verified across different Engel classes.</div></div><div><h3>Results:</h3><div>The classification model combining the low-frequency and high-frequency features can achieve an accuracy of 82.18% (sensitivity: 85.01%, specificity: 79.69%) and an area under curve (AUC) of 0.90 in identifying SOZ. Furthermore, the model exhibits strong generalizability in identifying SOZ in patients with MRI lesional and non-lesional, as well as those implanted with electrocorticography (ECOG) and stereotactic EEG (SEEG) electrodes. Moreover, the prediction model could achieve an average accuracy of 79.8% and an AUC of 0.84. Of note, the prediction of surgical success probability is significant between patients with successful and failed surgeries (P<span><math><mo><</mo></math></span>0.001). Correspondingly, the highest consistency between model-predicted SOZ and clinically labeled SOZ can be observed in patients with successful surgeries, but this consistency gradually decreases with increasing Engel classes.</div></div><div><h3>Conclusions:</h3><div>These results demonstrate that the CII may be a potential biomarker for identifying the SOZ in patients with DRE, which may provide a new perspective for the treatment of epilepsy.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"258 ","pages":"Article 108483"},"PeriodicalIF":4.9,"publicationDate":"2024-11-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142616231","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 multi-scale, multi-task fusion UNet model for accurate breast tumor segmentation 用于精确乳腺肿瘤分割的多尺度、多任务融合 UNet 模型。

IF 4.9 2区医学 Q1 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Computer methods and programs in biomedicine

Pub Date : 2024-11-07 DOI: 10.1016/j.cmpb.2024.108484

Shuo Dai , Xueyan Liu , Wei Wei , Xiaoping Yin , Lishan Qiao , Jianing Wang , Yu Zhang , Yan Hou

Background and Objective:

Breast cancer is the most common cancer type among women worldwide and a leading cause of female death. Accurately interpreting these complex tumors, involving small size and morphology, requires a significant amount of expertise and time. Developing a breast tumor segmentation model to assist clinicians in treatment, therefore, holds great practical significance.

Methods:

We propose a multi-scale, multi-task model framework named MTF-UNet. Firstly, we differ from the common approach of using different convolution kernel sizes to extract multi-scale features, and instead use the same convolution kernel size with different numbers of convolutions to obtain multi-scale, multi-level features. Additionally, to better integrate features from different levels and sizes, we extract a new multi-branch feature fusion block (ADF). This block differs from using channel and spatial attention to fuse features, but considers fusion weights between various branches. Secondly, we propose to use the number of pixels predicted to be related to tumors and background to assist segmentation, which is different from the conventional approach of using classification tasks to assist segmentation.

Results:

We conducted extensive experiments on our proprietary DCE-MRI dataset, as well as two public datasets (BUSI and Kvasir-SEG). In the aforementioned datasets, our model achieved excellent MIoU scores of 90.4516%, 89.8408%, and 92.8431% on the respective test sets. Furthermore, our ablation study has demonstrated the efficacy of each component and the effective integration of our auxiliary prediction branch into other models.

Conclusion:

Through comprehensive experiments and comparisons with other algorithms, the effectiveness, adaptability, and robustness of our proposed method have been demonstrated. We believe that MTF-UNet has great potential for further development in the field of medical image segmentation. The relevant code and data can be found at https://github.com/LCUDai/MTF-UNet.git.

背景和目的：乳腺癌是全球女性最常见的癌症类型，也是女性死亡的主要原因。准确解读这些体积小、形态复杂的肿瘤需要大量的专业知识和时间。因此，开发一种乳腺肿瘤分割模型来协助临床医生进行治疗具有重要的现实意义：我们提出了一个名为 MTF-UNet 的多尺度、多任务模型框架。首先，我们有别于使用不同卷积核大小提取多尺度特征的常见方法，而是使用相同卷积核大小、不同卷积次数来获得多尺度、多层次特征。此外，为了更好地整合不同级别和规模的特征，我们提取了一个新的多分支特征融合块（ADF）。这个区块不同于使用通道和空间注意力来融合特征，而是考虑了不同分支之间的融合权重。其次，我们建议使用预测与肿瘤和背景相关的像素数量来辅助分割，这与使用分类任务来辅助分割的传统方法不同：我们在专有的 DCE-MRI 数据集以及两个公共数据集（BUSI 和 Kvasir-SEG）上进行了大量实验。在上述数据集中，我们的模型在各自的测试集上分别取得了 90.4516%、89.8408% 和 92.8431% 的优异 MIoU 分数。此外，我们的消融研究还证明了每个组件的功效，以及我们的辅助预测分支与其他模型的有效整合：通过全面的实验以及与其他算法的比较，我们提出的方法的有效性、适应性和稳健性得到了证明。我们相信，MTF-UNet 在医学图像分割领域的进一步发展潜力巨大。相关代码和数据见 https://github.com/LCUDai/MTF-UNet.git。

{"title":"A multi-scale, multi-task fusion UNet model for accurate breast tumor segmentation","authors":"Shuo Dai , Xueyan Liu , Wei Wei , Xiaoping Yin , Lishan Qiao , Jianing Wang , Yu Zhang , Yan Hou","doi":"10.1016/j.cmpb.2024.108484","DOIUrl":"10.1016/j.cmpb.2024.108484","url":null,"abstract":"<div><h3>Background and Objective:</h3><div>Breast cancer is the most common cancer type among women worldwide and a leading cause of female death. Accurately interpreting these complex tumors, involving small size and morphology, requires a significant amount of expertise and time. Developing a breast tumor segmentation model to assist clinicians in treatment, therefore, holds great practical significance.</div></div><div><h3>Methods:</h3><div>We propose a multi-scale, multi-task model framework named MTF-UNet. Firstly, we differ from the common approach of using different convolution kernel sizes to extract multi-scale features, and instead use the same convolution kernel size with different numbers of convolutions to obtain multi-scale, multi-level features. Additionally, to better integrate features from different levels and sizes, we extract a new multi-branch feature fusion block (ADF). This block differs from using channel and spatial attention to fuse features, but considers fusion weights between various branches. Secondly, we propose to use the number of pixels predicted to be related to tumors and background to assist segmentation, which is different from the conventional approach of using classification tasks to assist segmentation.</div></div><div><h3>Results:</h3><div>We conducted extensive experiments on our proprietary DCE-MRI dataset, as well as two public datasets (BUSI and Kvasir-SEG). In the aforementioned datasets, our model achieved excellent MIoU scores of 90.4516%, 89.8408%, and 92.8431% on the respective test sets. Furthermore, our ablation study has demonstrated the efficacy of each component and the effective integration of our auxiliary prediction branch into other models.</div></div><div><h3>Conclusion:</h3><div>Through comprehensive experiments and comparisons with other algorithms, the effectiveness, adaptability, and robustness of our proposed method have been demonstrated. We believe that MTF-UNet has great potential for further development in the field of medical image segmentation. The relevant code and data can be found at <span><span>https://github.com/LCUDai/MTF-UNet.git</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"258 ","pages":"Article 108484"},"PeriodicalIF":4.9,"publicationDate":"2024-11-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142616210","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

Effect of myogenic tone on agonist-mediated vasoconstriction in isolated arteries: A computational study 肌源性张力对激动剂介导的离体动脉血管收缩的影响：计算研究

IF 4.9 2区医学 Q1 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Computer methods and programs in biomedicine

Pub Date : 2024-11-06 DOI: 10.1016/j.cmpb.2024.108495

Ranjan K. Pradhan

<div><h3>Background and objective</h3><div>Vasoconstriction of the resistance artery is mainly determined by an integrated action of multiple local stimuli acting on the vascular smooth muscle cells, which include neuronal delivery of <em>α</em>-adrenoceptor agonists and intraluminal pressure. The contractile activity of the arterial wall has been extensively studied <em>ex vivo</em> using isolated arterial preparations and myography techniques. However, agonist-mediated vasoconstriction response is often confounded by local effects of other stimuli (e.g., pressure) and, it remained unclear whether the pressure-induced myogenic response has any implication on the efficacy of agonist-mediated vasoconstriction during blood flow regulation in tissues. A quantitative understanding of the influence of each stimulus is necessary to understand the interaction between multiple regulatory mechanisms, which is required to ensure timely oxygen delivery to meet tissue needs.</div></div><div><h3>Methods</h3><div>We developed a simple empirical model of isolated vessel vasoreactivity that includes passive vessel wall mechanics and a lumped representation of active smooth muscle activation as a function of agonist concentration and pressure. Pressure myograph data in dog renal arterioles and rat femoral arterioles, isovolumic myograph data in rat femoral arteries, and vasoactive data in rat skeletal muscle arterioles were analyzed using the model. The effect of physiological pressure changes on the sensitivities of vascular segments to adrenergic agonists phenylephrine and norepinephrine was evaluated.</div></div><div><h3>Results</h3><div>Model-based analysis of isolated vasoreactivity data, obtained due to changes in pressure and vasoconstricting agonists revealed that the strength of myogenic response of a resistance vessel has a strong influence on the sensitivity and dynamics of agonist response. An increase in intraluminal pressure was found to reduce the magnitude of agonist-mediated tone by lowering the sensitivity of the vessel segment to agonist. The passive mechanical properties of arterial wall considearably influence the agonist-mediated contraction in isolated arteries. These results demonstrate how passive vessel wall mechanics may dominate the vasoactive responses of the common myogenic and adrenergic pathways of smooth muscle contraction in blood flow regulation, supporting a long standing notion that there exists segment-specific vasoregulation in microvascular networks of various tissues.</div></div><div><h3>Conclusion</h3><div>The present model provides a simple and powerful tool for quantifying <em>ex vivo</em> vasoreactivity of asolated arteries to qualitatively study the interaction between myogenic and <em>α</em>-adrenergic control of vascular tone in isolated vessels. Analysis of pressure myography data and isovolumic myography data in different sizes of vessels and tissues, in response to norepinephrine and phenylephrine revealed the im

背景和目的：阻力动脉的血管收缩主要是由作用于血管平滑肌细胞的多种局部刺激综合决定的，其中包括神经元传递的α肾上腺素受体激动剂和腔内压力。利用离体动脉制备物和肌电图技术对动脉壁的收缩活动进行了广泛的体外研究。然而，激动剂介导的血管收缩反应经常受到其他刺激（如压力）的局部影响，而且压力诱导的肌源性反应是否对激动剂介导的血管收缩在组织血流调节过程中的功效有任何影响仍不清楚。要了解多种调节机制之间的相互作用，就必须对每种刺激的影响进行定量了解，以确保及时输送氧气满足组织需要：我们建立了一个简单的孤立血管血管活性经验模型，该模型包括被动血管壁力学以及作为激动剂浓度和压力函数的主动平滑肌活化的综合表示。利用该模型分析了狗肾动脉和大鼠股动脉的压力肌电图数据、大鼠股动脉的等容肌电图数据以及大鼠骨骼肌动脉的血管活性数据。评估了生理压力变化对血管节段对肾上腺素能激动剂苯肾上腺素和去甲肾上腺素敏感性的影响：对压力和血管收缩激动剂变化引起的离体血管反应数据进行的基于模型的分析表明，阻力血管肌源性反应的强度对激动剂反应的敏感性和动态有很大影响。研究发现，腔内压力的增加会降低血管段对激动剂的敏感性，从而降低激动剂介导的张力。在离体动脉中，动脉壁的被动机械特性对激动剂介导的收缩有很大影响。这些结果表明，在血流调节过程中，血管壁的被动机械特性可能会主导平滑肌收缩的常见肌源性和肾上腺素能途径的血管活性反应，从而支持了一个长期存在的观点，即在各种组织的微血管网络中存在分段特异性血管调节：本模型提供了一种简单而强大的工具，可用于量化离体动脉的体外血管活性，从而定性研究离体血管中肌源性和α-肾上腺素能控制血管张力之间的相互作用。通过分析不同大小血管和组织在去甲肾上腺素和苯肾上腺素作用下的压力肌动图数据和等容肌动图数据，发现了被动血管力学在动脉血管运动和单根血管基础血管运动张力建立中的重要性。本研究将有助于量化肌源性张力在多大程度上可能影响激动剂介导的血管收缩以及激动剂对组织血流调节过程中微血管网络压力介导的肌源性反应的影响。

{"title":"Effect of myogenic tone on agonist-mediated vasoconstriction in isolated arteries: A computational study","authors":"Ranjan K. Pradhan","doi":"10.1016/j.cmpb.2024.108495","DOIUrl":"10.1016/j.cmpb.2024.108495","url":null,"abstract":"<div><h3>Background and objective</h3><div>Vasoconstriction of the resistance artery is mainly determined by an integrated action of multiple local stimuli acting on the vascular smooth muscle cells, which include neuronal delivery of <em>α</em>-adrenoceptor agonists and intraluminal pressure. The contractile activity of the arterial wall has been extensively studied <em>ex vivo</em> using isolated arterial preparations and myography techniques. However, agonist-mediated vasoconstriction response is often confounded by local effects of other stimuli (e.g., pressure) and, it remained unclear whether the pressure-induced myogenic response has any implication on the efficacy of agonist-mediated vasoconstriction during blood flow regulation in tissues. A quantitative understanding of the influence of each stimulus is necessary to understand the interaction between multiple regulatory mechanisms, which is required to ensure timely oxygen delivery to meet tissue needs.</div></div><div><h3>Methods</h3><div>We developed a simple empirical model of isolated vessel vasoreactivity that includes passive vessel wall mechanics and a lumped representation of active smooth muscle activation as a function of agonist concentration and pressure. Pressure myograph data in dog renal arterioles and rat femoral arterioles, isovolumic myograph data in rat femoral arteries, and vasoactive data in rat skeletal muscle arterioles were analyzed using the model. The effect of physiological pressure changes on the sensitivities of vascular segments to adrenergic agonists phenylephrine and norepinephrine was evaluated.</div></div><div><h3>Results</h3><div>Model-based analysis of isolated vasoreactivity data, obtained due to changes in pressure and vasoconstricting agonists revealed that the strength of myogenic response of a resistance vessel has a strong influence on the sensitivity and dynamics of agonist response. An increase in intraluminal pressure was found to reduce the magnitude of agonist-mediated tone by lowering the sensitivity of the vessel segment to agonist. The passive mechanical properties of arterial wall considearably influence the agonist-mediated contraction in isolated arteries. These results demonstrate how passive vessel wall mechanics may dominate the vasoactive responses of the common myogenic and adrenergic pathways of smooth muscle contraction in blood flow regulation, supporting a long standing notion that there exists segment-specific vasoregulation in microvascular networks of various tissues.</div></div><div><h3>Conclusion</h3><div>The present model provides a simple and powerful tool for quantifying <em>ex vivo</em> vasoreactivity of asolated arteries to qualitatively study the interaction between myogenic and <em>α</em>-adrenergic control of vascular tone in isolated vessels. Analysis of pressure myography data and isovolumic myography data in different sizes of vessels and tissues, in response to norepinephrine and phenylephrine revealed the im","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"258 ","pages":"Article 108495"},"PeriodicalIF":4.9,"publicationDate":"2024-11-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142616211","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

Segmentation of infarct lesions and prognosis prediction for acute ischemic stroke using non-contrast CT scans 使用非对比 CT 扫描对急性缺血性脑卒中的梗死病灶进行分割并预测预后。

IF 4.9 2区医学 Q1 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Computer methods and programs in biomedicine

Pub Date : 2024-11-05 DOI: 10.1016/j.cmpb.2024.108488

Xuechun Wang , Yuting Meng , Zhijian Dong , Zehong Cao , Yichu He , Tianyang Sun , Qing Zhou , Guozhong Niu , Zhongxiang Ding , Feng Shi , Dinggang Shen

Background and Purpose

Ischemic stroke is the most common type of stroke and the second leading cause of global mortality. Prompt and accurate diagnosis is crucial for effective treatment. Non-contrast CT (NCCT) scans are commonly employed as the first-line imaging modality to identify the infarct lesion and affected brain areas, as well as to make prognostic predictions to guide the subsequent treatment planning. However, visual evaluation of infarct lesions in NCCT scans can be subjective and inconsistent due to reliance on expert experience.

Methods

In this study, we propose an automatic method using VB-Net with dual-channel inputs to segment acute infarct lesions (AIL) on NCCT scans and extract affected ASPECTS (Alberta Stroke Program Early CT Score) regions. Secondly, we establish a prediction model to distinguish reperfused patients from non-reperfused patients after treatment, based on multi-dimensional radiological features of baseline NCCT and stroke onset time. Thirdly, we create a prediction model estimating the infarct volume after a period of time, by combining NCCT infarct volume, radiological features, and surgical decision.

Results

The median Dice coefficient of the AIL segmentation network is 0.76. Based on this, the patient triage model has an AUC of 0.837 (95 % confidence interval [CI]: 0.734–0.941), sensitivity of 0.833 (95 % CI: 0.626–0.953). The predicted follow-up infarct volume correlates strongly with the DWI ground truth, with a Pearson correlation coefficient of 0.931.

Conclusions

Our proposed pipeline offers qualitative and quantitative assessment of infarct lesions based on NCCT scans, facilitating physicians in patient triage and prognosis prediction.

背景和目的：缺血性中风是最常见的中风类型，也是全球第二大死亡原因。及时准确的诊断对有效治疗至关重要。非对比 CT（NCCT）扫描通常被用作一线成像模式，以确定梗死病灶和受影响的脑区，并预测预后以指导后续治疗计划。然而，由于对专家经验的依赖，NCCT 扫描中对梗死病灶的视觉评估可能存在主观性和不一致性：在这项研究中，我们提出了一种使用 VB-Net 的自动方法，利用双通道输入分割 NCCT 扫描中的急性梗死病灶（AIL），并提取受影响的 ASPECTS（阿尔伯塔省卒中项目早期 CT 评分）区域。其次，我们根据基线 NCCT 和卒中发生时间的多维放射学特征建立了一个预测模型，用于区分治疗后再灌注患者和非再灌注患者。第三，我们结合 NCCT 梗死体积、放射学特征和手术决定，建立了一个预测模型，估计一段时间后的梗死体积：结果：AIL分割网络的中位Dice系数为0.76。在此基础上，患者分流模型的 AUC 为 0.837（95% 置信区间 [CI]：0.734-0.941），灵敏度为 0.833（95% 置信区间 [CI]：0.626-0.953）。预测的随访梗死体积与 DWI 地面真实值密切相关，皮尔逊相关系数为 0.931：我们提出的管道可根据 NCCT 扫描结果对梗死病灶进行定性和定量评估，方便医生对患者进行分流和预后预测。

{"title":"Segmentation of infarct lesions and prognosis prediction for acute ischemic stroke using non-contrast CT scans","authors":"Xuechun Wang , Yuting Meng , Zhijian Dong , Zehong Cao , Yichu He , Tianyang Sun , Qing Zhou , Guozhong Niu , Zhongxiang Ding , Feng Shi , Dinggang Shen","doi":"10.1016/j.cmpb.2024.108488","DOIUrl":"10.1016/j.cmpb.2024.108488","url":null,"abstract":"<div><h3>Background and Purpose</h3><div>Ischemic stroke is the most common type of stroke and the second leading cause of global mortality. Prompt and accurate diagnosis is crucial for effective treatment. Non-contrast CT (NCCT) scans are commonly employed as the first-line imaging modality to identify the infarct lesion and affected brain areas, as well as to make prognostic predictions to guide the subsequent treatment planning. However, visual evaluation of infarct lesions in NCCT scans can be subjective and inconsistent due to reliance on expert experience.</div></div><div><h3>Methods</h3><div>In this study, we propose an automatic method using VB-Net with dual-channel inputs to segment acute infarct lesions (AIL) on NCCT scans and extract affected ASPECTS (Alberta Stroke Program Early CT Score) regions. Secondly, we establish a prediction model to distinguish reperfused patients from non-reperfused patients after treatment, based on multi-dimensional radiological features of baseline NCCT and stroke onset time. Thirdly, we create a prediction model estimating the infarct volume after a period of time, by combining NCCT infarct volume, radiological features, and surgical decision.</div></div><div><h3>Results</h3><div>The median Dice coefficient of the AIL segmentation network is 0.76. Based on this, the patient triage model has an AUC of 0.837 (95 % confidence interval [CI]: 0.734–0.941), sensitivity of 0.833 (95 % CI: 0.626–0.953). The predicted follow-up infarct volume correlates strongly with the DWI ground truth, with a Pearson correlation coefficient of 0.931.</div></div><div><h3>Conclusions</h3><div>Our proposed pipeline offers qualitative and quantitative assessment of infarct lesions based on NCCT scans, facilitating physicians in patient triage and prognosis prediction.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"258 ","pages":"Article 108488"},"PeriodicalIF":4.9,"publicationDate":"2024-11-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142616198","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