Artificial Intelligence in Medicine最新文献_第3页

B2E-CDG: Conditional diffusion-based for label-free OCT angiography artifact removal and robust vascular reconstruction B2E-CDG：基于条件扩散的无标记OCT血管造影伪影去除和稳健血管重建

IF 6.2 2区医学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Artificial Intelligence in Medicine

Pub Date : 2025-12-30 DOI: 10.1016/j.artmed.2025.103345

Jing Xu , Suzhong Fu , Jiwei Xing , Linyan Xue , Qingliang Zhao

Optical Coherence Tomography Angiography (OCTA) is a revolutionary technology widely used in the diagnosis and management of fundus, skin and cardiovascular diseases. However, unavoidable movements, such as breathing, often introduce motion artifacts into OCTA images, which can significantly degrade image quality, obscure critical vascular details, and reduce the diagnostic reliability of the modality. Although recent advances in learning-based image inpainting methods for OCTA enface images have made notable progress in artifact removal, these methods typically require large amounts of accurately labeled data and the generation of pseudo stripes to construct paired training datasets. Additionally, the abundant structural information and flow intensity signals available in OCTA B-scans are often under-utilized. Here we proposed a novel method:B-scans to Enface Conditional Diffusion Guidance (B2E-CDG), which translates signal-void B-scans into normal B-scans. Moreover, the normal B-scans were introduced in a connection manner and the specified reference B-scans in a gradient-based manner as style feature guidance within a diffusion model. Importantly, conditional guidance facilitates a more controlled and precise generation process for flow signal recovery in B-scans. Notably, our method eliminates the need for labeled datasets and pseudo stripes, due to the repetitive scanning nature of OCTA inherently provides paired signal-void and normal B- scans. Our results demonstrated that B2E-CDG effectively removes motion artifacts and restores vascular and structural details. The proposed method shows superior performance in vascular recovery and artifact removal metrics, thereby improving the clinical utility and diagnostic reliability of OCTA.

光学相干断层血管造影（OCTA）是一项革命性的技术，广泛应用于眼底、皮肤和心血管疾病的诊断和治疗。然而，不可避免的运动，如呼吸，通常会在OCTA图像中引入运动伪影，这会显著降低图像质量，模糊关键血管细节，降低模态诊断的可靠性。尽管最近基于学习的OCTA表面图像的图像绘制方法在去除伪影方面取得了显著进展，但这些方法通常需要大量准确标记的数据和生成伪条纹来构建成对的训练数据集。此外，OCTA b扫描中丰富的结构信息和血流强度信号往往没有得到充分利用。本文提出了一种新的方法：b扫描到面条件扩散引导（B2E-CDG），它将信号空洞的b扫描转换为正常的b扫描。此外，在扩散模型中，以连接方式引入正常b扫描，并以基于梯度的方式引入指定参考b扫描作为风格特征指导。重要的是，条件引导有助于b扫描中流量信号恢复的更可控和精确的生成过程。值得注意的是，我们的方法消除了对标记数据集和伪条纹的需要，因为OCTA的重复扫描本质上提供了成对的信号空洞和正常B扫描。我们的研究结果表明，B2E-CDG有效地去除运动伪影，恢复血管和结构细节。该方法在血管恢复和伪影去除指标方面表现出优异的性能，从而提高了OCTA的临床实用性和诊断可靠性。

{"title":"B2E-CDG: Conditional diffusion-based for label-free OCT angiography artifact removal and robust vascular reconstruction","authors":"Jing Xu , Suzhong Fu , Jiwei Xing , Linyan Xue , Qingliang Zhao","doi":"10.1016/j.artmed.2025.103345","DOIUrl":"10.1016/j.artmed.2025.103345","url":null,"abstract":"<div><div>Optical Coherence Tomography Angiography (OCTA) is a revolutionary technology widely used in the diagnosis and management of fundus, skin and cardiovascular diseases. However, unavoidable movements, such as breathing, often introduce motion artifacts into OCTA images, which can significantly degrade image quality, obscure critical vascular details, and reduce the diagnostic reliability of the modality. Although recent advances in learning-based image inpainting methods for OCTA enface images have made notable progress in artifact removal, these methods typically require large amounts of accurately labeled data and the generation of pseudo stripes to construct paired training datasets. Additionally, the abundant structural information and flow intensity signals available in OCTA B-scans are often under-utilized. Here we proposed a novel method:B-scans to Enface Conditional Diffusion Guidance (B2E-CDG), which translates signal-void B-scans into normal B-scans. Moreover, the normal B-scans were introduced in a connection manner and the specified reference B-scans in a gradient-based manner as style feature guidance within a diffusion model. Importantly, conditional guidance facilitates a more controlled and precise generation process for flow signal recovery in B-scans. Notably, our method eliminates the need for labeled datasets and pseudo stripes, due to the repetitive scanning nature of OCTA inherently provides paired signal-void and normal B- scans. Our results demonstrated that B2E-CDG effectively removes motion artifacts and restores vascular and structural details. The proposed method shows superior performance in vascular recovery and artifact removal metrics, thereby improving the clinical utility and diagnostic reliability of OCTA.</div></div>","PeriodicalId":55458,"journal":{"name":"Artificial Intelligence in Medicine","volume":"173 ","pages":"Article 103345"},"PeriodicalIF":6.2,"publicationDate":"2025-12-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145928172","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

Seamless monitoring of stress levels leveraging a foundational model for time sequences 利用时间序列的基础模型无缝监测压力水平

IF 6.2 2区医学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Artificial Intelligence in Medicine

Pub Date : 2025-12-24 DOI: 10.1016/j.artmed.2025.103336

Davide Gabrielli , Bardh Prenkaj , Paola Velardi

Background:

Accurate and continuous monitoring of physiological stress is crucial, especially for patients with neurodegenerative diseases. Traditional monitoring methods, such as Electrocardiogram (ECG), are often invasive and limited in duration, while data from lightweight wearable devices, though more practical for seamless monitoring, typically suffers from significant quality degradation compared to clinical-grade measurements.

Motivation:

The challenge lies in developing a robust, long-term, and patient-friendly stress monitoring system that overcomes the limitations of conventional approaches and the accuracy compromises of current wearables. Such a system must also provide actionable, interpretable insights for clinicians and adapt to individual patient variability.

Method:

This manuscript introduces a methodology for seamless stress level monitoring by leveraging UniTS, a foundational model for time series. Our approach redefines stress detection as an anomaly detection problem, establishing a personalized baseline for each patient’s physiological behavior. Furthermore, to enhance clinical utility and trust, the system integrates a Large Language Model (LLM) to generate human-readable explanations for detected anomalies.

Results:

The proposed UniTS-based methodology demonstrates superior performance, outperforming 12 top-performing methods on three benchmark datasets. Crucially, it achieves performance comparable to that obtained from more invasive, clinical-grade devices (like ECG) even when utilizing data from lightweight wearable devices, thereby enabling truly seamless monitoring. Furthermore, the system has been successfully tested in a real-world environment, in the context of a project to monitor elderly patients with cognitive disorders in their homes.

Novelty:

This work presents an advancement in physiological stress monitoring by offering a personalized, explainable, and continuously adaptive system. We extend and fine-tune UniTS to support contextual anomaly detection and LLM-driven explainability, addressing critical gaps in current healthcare monitoring, fostering enhanced clinician control, improved system predictability, and facilitating long-term, real-world applicability for patients with neurodegenerative conditions.

背景：准确和持续的生理应激监测是至关重要的，特别是对神经退行性疾病患者。传统的监测方法，如心电图（ECG），通常是侵入性的，持续时间有限，而来自轻量级可穿戴设备的数据，尽管对于无缝监测更实用，但与临床级测量相比，通常会遭受严重的质量下降。动机：挑战在于开发一个强大的、长期的、对患者友好的压力监测系统，克服传统方法的局限性和当前可穿戴设备的准确性妥协。这样的系统还必须为临床医生提供可操作的、可解释的见解，并适应个体患者的可变性。方法：本文介绍了一种方法，无缝应力水平监测利用单位，一个基本模型的时间序列。我们的方法将压力检测重新定义为异常检测问题，为每位患者的生理行为建立个性化基线。此外，为了提高临床效用和信任，该系统集成了一个大型语言模型（LLM），为检测到的异常生成人类可读的解释。结果：提出的基于单元的方法表现出优异的性能，在三个基准数据集上优于12种表现最佳的方法。至关重要的是，即使在使用轻量级可穿戴设备的数据时，它的性能也可与更具侵入性的临床级设备（如ECG）相媲美，从而实现真正的无缝监控。此外，该系统已经成功地在现实环境中进行了测试，在一个监测家中患有认知障碍的老年患者的项目中。新颖性：这项工作通过提供个性化、可解释和持续自适应的系统，在生理应激监测方面取得了进展。我们扩展和微调单元，以支持上下文异常检测和法学硕士驱动的可解释性，解决当前医疗保健监测中的关键差距，促进增强临床医生控制，提高系统可预测性，并促进神经退行性疾病患者的长期、现实应用。

{"title":"Seamless monitoring of stress levels leveraging a foundational model for time sequences","authors":"Davide Gabrielli , Bardh Prenkaj , Paola Velardi","doi":"10.1016/j.artmed.2025.103336","DOIUrl":"10.1016/j.artmed.2025.103336","url":null,"abstract":"<div><h3>Background:</h3><div>Accurate and continuous monitoring of physiological stress is crucial, especially for patients with neurodegenerative diseases. Traditional monitoring methods, such as Electrocardiogram (ECG), are often invasive and limited in duration, while data from lightweight wearable devices, though more practical for seamless monitoring, typically suffers from significant quality degradation compared to clinical-grade measurements.</div></div><div><h3>Motivation:</h3><div>The challenge lies in developing a robust, long-term, and patient-friendly stress monitoring system that overcomes the limitations of conventional approaches and the accuracy compromises of current wearables. Such a system must also provide actionable, interpretable insights for clinicians and adapt to individual patient variability.</div></div><div><h3>Method:</h3><div>This manuscript introduces a methodology for seamless stress level monitoring by leveraging UniTS, a foundational model for time series. Our approach redefines stress detection as an anomaly detection problem, establishing a personalized baseline for each patient’s physiological behavior. Furthermore, to enhance clinical utility and trust, the system integrates a Large Language Model (LLM) to generate human-readable explanations for detected anomalies.</div></div><div><h3>Results:</h3><div>The proposed UniTS-based methodology demonstrates superior performance, outperforming 12 top-performing methods on three benchmark datasets. Crucially, it achieves performance comparable to that obtained from more invasive, clinical-grade devices (like ECG) even when utilizing data from lightweight wearable devices, thereby enabling truly seamless monitoring. Furthermore, the system has been successfully tested in a real-world environment, in the context of a project to monitor elderly patients with cognitive disorders in their homes.</div></div><div><h3>Novelty:</h3><div>This work presents an advancement in physiological stress monitoring by offering a personalized, explainable, and continuously adaptive system. We extend and fine-tune UniTS to support contextual anomaly detection and LLM-driven explainability, addressing critical gaps in current healthcare monitoring, fostering enhanced clinician control, improved system predictability, and facilitating long-term, real-world applicability for patients with neurodegenerative conditions.</div></div>","PeriodicalId":55458,"journal":{"name":"Artificial Intelligence in Medicine","volume":"173 ","pages":"Article 103336"},"PeriodicalIF":6.2,"publicationDate":"2025-12-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145886452","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

Clinical decision support system for detecting right ventricular dysfunction in acute pulmonary embolism: Explainable a new Binary Rule Search (BRS) ensembles and robustness evaluation 检测急性肺栓塞患者右室功能障碍的临床决策支持系统：一种新的二元规则搜索（BRS）集合和鲁棒性评估的可解释性。

IF 6.2 2区医学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Artificial Intelligence in Medicine

Pub Date : 2025-12-20 DOI: 10.1016/j.artmed.2025.103337

Mehmet Tahir Huyut , Andrei Velichko , Maksim Belyaev , Yuriy Izotov , Mustafa Tosun , Bünyamin Sertoğullarından , Şebnem Karaoğlanoğlu , Abdussamed Yasin Demir , Dmitry Korzun

Background

Right-ventricular dysfunction (RVD) in acute pulmonary embolism (PE) carries excess short-term mortality; fast, transparent risk stratification is needed.

Objective

To develop and evaluate an interpretable Binary Rule Search (BRS) framework for RVD detection on a fully binarized clinical dataset, emphasizing conservative robustness.

Methods

We analyzed a single-center cohort (N = 363; development = 250, external validation = 113) recoded into 0/1 predictors. For subset sizes k = 1–5, a Rust BRS searched bit-mask rules maximizing Matthews correlation coefficient (MCC) on development data and was assessed over 500 stratified 250/113 hold-out repeats. We report MCC with bootstrap 95 % CIs plus sensitivity, specificity, precision, F1, and AUC-ROC. Robustness was summarized by the Stability-Bound Rule Score (SBRS).

Results

Performance increased with k; triplets offered the best parsimony–robustness balance, with modest conservative gains for quintets. The strongest quintet by point estimate was S{Age65_79; ThromMain; ThromBilat; DVTuni; Malig} (MCC = 0.326; 95 % CI 0.152–0.498). A balanced k = 4 rule, S{SexMale; ThromMain; Hypertens; HeartFail}, achieved Sensitivity 0.791, Specificity 0.557, F1 0.630, AUC-ROC 0.674 on external validation. Decision-tree checks yielded high lower-bound performance (e.g., L95 = 0.302 for S{ThromMain; DVTdist; COPD}), whereas LogNNet often matched mean MCC but showed consistently lower L95, indicating greater dispersion. PFI highlighted a thrombus-centric signal (ThromMain/ThromBilat/DVT) with meaningful secondary contributions from heart failure and hypertension.

Conclusions

Interpretable BRS ensembles deliver clinically acceptable, conservatively bounded performance using few binary predictors and transparent logic, supporting clinician-facing decision support in resource-constrained settings. We provide rule masks, uncertainty summaries, and an offline CDSS prototype; prospective multicenter validation is warranted.

背景：急性肺栓塞（PE）患者的右心室功能障碍（RVD）会导致较高的短期死亡率；需要快速、透明的风险分层。目的：在完全二值化的临床数据集上开发和评估可解释的RVD检测二值化规则搜索（BRS）框架，强调保守稳健性。方法：我们分析了一个单中心队列（N = 363；发展= 250，外部验证= 113），重新编码为0/1预测因子。对于子集大小k = 1-5， Rust BRS在开发数据上搜索最大化Matthews相关系数（MCC）的位掩码规则，并对500多个分层250/113 hold- hold repeat进行评估。我们报告MCC的自举95% ci加上灵敏度、特异性、精度、F1和AUC-ROC。稳健性通过稳定性界限规则评分（SBRS）来总结。结果：性能随k的增加而增加；三连音提供了最好的节俭和健壮的平衡，五重奏有适度的保守收益。按点估计最强的五重奏是S{Age65_79；ThromMain;ThromBilat;DVTuni;（MCC = 0.326; 95% CI 0.152 ~ 0.498）。平衡k = 4规则，S{SexMale；ThromMain;Hypertens;，外部验证灵敏度0.791，特异性0.557,F1 0.630, AUC-ROC 0.674。决策树检查产生了较高的下限性能（例如，S{ThromMain; DVTdist； COPD}的L95 = 0.302），而LogNNet通常匹配平均MCC，但始终显示较低的L95，表明更大的分散性。PFI突出了一个以血栓为中心的信号（ThromMain/ThromBilat/DVT），有意义的继发性贡献来自心力衰竭和高血压。结论：可解释的BRS集合使用较少的二元预测因子和透明的逻辑提供临床可接受的保守有限性能，支持资源受限环境下面向临床医生的决策支持。我们提供了规则掩码、不确定性摘要和离线CDSS原型；前瞻性多中心验证是必要的。

{"title":"Clinical decision support system for detecting right ventricular dysfunction in acute pulmonary embolism: Explainable a new Binary Rule Search (BRS) ensembles and robustness evaluation","authors":"Mehmet Tahir Huyut , Andrei Velichko , Maksim Belyaev , Yuriy Izotov , Mustafa Tosun , Bünyamin Sertoğullarından , Şebnem Karaoğlanoğlu , Abdussamed Yasin Demir , Dmitry Korzun","doi":"10.1016/j.artmed.2025.103337","DOIUrl":"10.1016/j.artmed.2025.103337","url":null,"abstract":"<div><h3>Background</h3><div>Right-ventricular dysfunction (RVD) in acute pulmonary embolism (PE) carries excess short-term mortality; fast, transparent risk stratification is needed.</div></div><div><h3>Objective</h3><div>To develop and evaluate an interpretable Binary Rule Search (BRS) framework for RVD detection on a fully binarized clinical dataset, emphasizing conservative robustness.</div></div><div><h3>Methods</h3><div>We analyzed a single-center cohort (<em>N</em> = 363; development = 250, external validation = 113) recoded into 0/1 predictors. For subset sizes <em>k</em> = 1–5, a Rust BRS searched bit-mask rules maximizing Matthews correlation coefficient (MCC) on development data and was assessed over 500 stratified 250/113 hold-out repeats. We report MCC with bootstrap 95 % CIs plus sensitivity, specificity, precision, F1, and AUC-ROC. Robustness was summarized by the Stability-Bound Rule Score (SBRS).</div></div><div><h3>Results</h3><div>Performance increased with k; triplets offered the best parsimony–robustness balance, with modest conservative gains for quintets. The strongest quintet by point estimate was S{Age65_79; ThromMain; ThromBilat; DVTuni; Malig} (MCC = 0.326; 95 % CI 0.152–0.498). A balanced k = 4 rule, S{SexMale; ThromMain; Hypertens; HeartFail}, achieved Sensitivity 0.791, Specificity 0.557, F1 0.630, AUC-ROC 0.674 on external validation. Decision-tree checks yielded high lower-bound performance (e.g., L95 = 0.302 for S{ThromMain; DVTdist; COPD}), whereas LogNNet often matched mean MCC but showed consistently lower L95, indicating greater dispersion. PFI highlighted a thrombus-centric signal (ThromMain/ThromBilat/DVT) with meaningful secondary contributions from heart failure and hypertension.</div></div><div><h3>Conclusions</h3><div>Interpretable BRS ensembles deliver clinically acceptable, conservatively bounded performance using few binary predictors and transparent logic, supporting clinician-facing decision support in resource-constrained settings. We provide rule masks, uncertainty summaries, and an offline CDSS prototype; prospective multicenter validation is warranted.</div></div>","PeriodicalId":55458,"journal":{"name":"Artificial Intelligence in Medicine","volume":"172 ","pages":"Article 103337"},"PeriodicalIF":6.2,"publicationDate":"2025-12-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145821986","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

Multi-scale feature enhancement in multi-task learning for medical image analysis 医学图像分析多任务学习中的多尺度特征增强

IF 6.2 2区医学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Artificial Intelligence in Medicine

Pub Date : 2025-12-19 DOI: 10.1016/j.artmed.2025.103338

Phuoc-Nguyen Bui , Duc-Tai Le , Junghyun Bum , Jong Chul Han , Van-Nguyen Pham , Hyunseung Choo

Traditional deep learning approaches in medical image analysis usually focus on either segmentation or classification, which limits their ability to exploit shared information between these interrelated tasks. Recent multi-task learning (MTL) methods aim to address this limitation by combining both tasks within a single model through shared feature representations. However, existing MTL models often fall short of optimal performance in both tasks, as they struggle to simultaneously capture the local contextual information essential for segmentation and the global one needed for classification. In this paper, we propose a simple yet effective UNet-based MTL model, where features extracted by the encoder are used to predict classification labels, while the decoder produces the segmentation mask. The model leverages an advanced encoder incorporating a novel ResFormer block that integrates local context from convolutional feature extraction with long-range dependencies modeled by the Transformer. This design captures broader contextual relationships and fine-grained details, improving classification and segmentation accuracy. To enhance classification performance, multi-scale features from different encoder levels are combined to leverage the hierarchical representation of the input image. For segmentation, the features passed to the decoder via skip connections are refined using a novel dilated feature enhancement (DFE) module, which captures information at different scales through three parallel convolution branches with varying dilation rates. This allows the decoder to detect lesions of varying sizes with greater accuracy. Experimental results across multiple medical datasets confirm the superior performance of our model in both segmentation and classification tasks, compared to state-of-the-art single-task and multi-task learning methods. These findings highlight the potential of our approach to advance disease diagnosis and treatment through improved medical image analysis. The code will be available at https://github.com/nguyenpbui/ResFormer.

医学图像分析中的传统深度学习方法通常侧重于分割或分类，这限制了它们利用这些相互关联任务之间共享信息的能力。最近的多任务学习（MTL）方法旨在通过共享特征表示将两个任务组合在一个模型中来解决这一限制。然而，现有的MTL模型在这两项任务中往往达不到最佳性能，因为它们难以同时捕获分割所需的局部上下文信息和分类所需的全局上下文信息。在本文中，我们提出了一个简单而有效的基于unet的MTL模型，其中编码器提取的特征用于预测分类标签，而解码器产生分割掩码。该模型利用了一个先进的编码器，该编码器结合了一个新颖的ResFormer块，该块将卷积特征提取的本地上下文与Transformer建模的远程依赖关系集成在一起。这种设计捕获了更广泛的上下文关系和细粒度的细节，提高了分类和分割的准确性。为了提高分类性能，将来自不同编码器级别的多尺度特征结合起来，以利用输入图像的分层表示。对于分割，通过跳过连接传递给解码器的特征使用新的扩展特征增强（DFE）模块进行细化，该模块通过三个具有不同扩展率的并行卷积分支捕获不同尺度的信息。这使得解码器能够以更高的精度检测不同大小的病变。与最先进的单任务和多任务学习方法相比，跨多个医疗数据集的实验结果证实了我们的模型在分割和分类任务方面的优越性能。这些发现突出了我们的方法通过改进医学图像分析来推进疾病诊断和治疗的潜力。代码可在https://github.com/nguyenpbui/ResFormer上获得。

{"title":"Multi-scale feature enhancement in multi-task learning for medical image analysis","authors":"Phuoc-Nguyen Bui , Duc-Tai Le , Junghyun Bum , Jong Chul Han , Van-Nguyen Pham , Hyunseung Choo","doi":"10.1016/j.artmed.2025.103338","DOIUrl":"10.1016/j.artmed.2025.103338","url":null,"abstract":"<div><div>Traditional deep learning approaches in medical image analysis usually focus on either segmentation or classification, which limits their ability to exploit shared information between these interrelated tasks. Recent multi-task learning (MTL) methods aim to address this limitation by combining both tasks within a single model through shared feature representations. However, existing MTL models often fall short of optimal performance in both tasks, as they struggle to simultaneously capture the local contextual information essential for segmentation and the global one needed for classification. In this paper, we propose a simple yet effective UNet-based MTL model, where features extracted by the encoder are used to predict classification labels, while the decoder produces the segmentation mask. The model leverages an advanced encoder incorporating a novel ResFormer block that integrates local context from convolutional feature extraction with long-range dependencies modeled by the Transformer. This design captures broader contextual relationships and fine-grained details, improving classification and segmentation accuracy. To enhance classification performance, multi-scale features from different encoder levels are combined to leverage the hierarchical representation of the input image. For segmentation, the features passed to the decoder via skip connections are refined using a novel dilated feature enhancement (DFE) module, which captures information at different scales through three parallel convolution branches with varying dilation rates. This allows the decoder to detect lesions of varying sizes with greater accuracy. Experimental results across multiple medical datasets confirm the superior performance of our model in both segmentation and classification tasks, compared to state-of-the-art single-task and multi-task learning methods. These findings highlight the potential of our approach to advance disease diagnosis and treatment through improved medical image analysis. The code will be available at <span><span>https://github.com/nguyenpbui/ResFormer</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":55458,"journal":{"name":"Artificial Intelligence in Medicine","volume":"173 ","pages":"Article 103338"},"PeriodicalIF":6.2,"publicationDate":"2025-12-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145852434","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Intelligent sentiment analysis with Arabic patient feedback on healthcare services in King Hussein Cancer Center 侯赛因国王癌症中心阿拉伯病人对医疗服务反馈的智能情感分析

IF 6.2 2区医学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Artificial Intelligence in Medicine

Pub Date : 2025-12-17 DOI: 10.1016/j.artmed.2025.103334

Hanan Saleet , Rana Husni Al Mahmoud , Hamzeh Abuasba , Dana Nashawati , Yasmeen Saidan

The growing digitization of healthcare services has led to an abundance of textual patient feedback, offering a unique opportunity to assess healthcare quality through sentiment analysis. While most existing research focuses on English-language data, non-English contexts — especially Arabic dialects — remain underexplored. This study introduces JADKHCC, a novel corpus specifically designed for sentiment analysis of patient feedback written in the Jordanian Arabic dialect, collected from King Hussein Cancer Center (KHCC). The corpus is manually annotated using a comprehensive methodology to capture sentiments on both three- and five-point Likert scales. This study analyzes 15,812 Jordanian Arabic Dialect comments by employing various pre-processing techniques and feature vectors, including BERT-base-Arabic, Word2Vec, and FastText. Furthermore, the study considers a wide range of deep learning classifiers alongside balancing techniques to address data imbalance. The results demonstrate the superior performance of the CNN with BERT representation model compared to the BiLSTM, LSTM, RNN, and RNNLSTM models. The findings indicate an F1-score of approximately 96%, suggesting the potential for predicting patient sentiment from textual feedback. By automating feedback analysis, this approach enables KHCC to detect dissatisfaction, identify unmet needs, and act on key concerns promptly. It reduces the burden of manual review and supports data-driven service improvements—advancing KHCC’s mission to deliver responsive, high-quality, patient-centered cancer care.

医疗保健服务的日益数字化导致了大量的文本患者反馈，为通过情感分析评估医疗保健质量提供了独特的机会。虽然大多数现有的研究都集中在英语数据上，但非英语语境——尤其是阿拉伯语方言——仍未得到充分探索。本研究介绍了一种新的语料库JADKHCC，该语料库专门用于对以约旦阿拉伯语方言撰写的患者反馈进行情感分析，该语料库来自侯赛因国王癌症中心（KHCC）。语料库使用综合方法手动注释，以捕获三分和五点李克特量表上的情绪。本研究采用BERT-base-Arabic、Word2Vec和FastText等多种预处理技术和特征向量，对15812条约旦阿拉伯语方言评论进行了分析。此外，该研究还考虑了广泛的深度学习分类器以及平衡技术来解决数据不平衡问题。结果表明，与BiLSTM、LSTM、RNN和RNNLSTM模型相比，使用BERT表示模型的CNN具有更好的性能。研究结果表明f1得分约为96%，表明从文本反馈预测患者情绪的潜力。通过自动化反馈分析，该方法使KHCC能够检测不满，识别未满足的需求，并及时对关键问题采取行动。它减少了人工审查的负担，并支持数据驱动的服务改进，推进了KHCC提供响应迅速、高质量、以患者为中心的癌症治疗的使命。

{"title":"Intelligent sentiment analysis with Arabic patient feedback on healthcare services in King Hussein Cancer Center","authors":"Hanan Saleet , Rana Husni Al Mahmoud , Hamzeh Abuasba , Dana Nashawati , Yasmeen Saidan","doi":"10.1016/j.artmed.2025.103334","DOIUrl":"10.1016/j.artmed.2025.103334","url":null,"abstract":"<div><div>The growing digitization of healthcare services has led to an abundance of textual patient feedback, offering a unique opportunity to assess healthcare quality through sentiment analysis. While most existing research focuses on English-language data, non-English contexts — especially Arabic dialects — remain underexplored. This study introduces JADKHCC, a novel corpus specifically designed for sentiment analysis of patient feedback written in the Jordanian Arabic dialect, collected from King Hussein Cancer Center (KHCC). The corpus is manually annotated using a comprehensive methodology to capture sentiments on both three- and five-point Likert scales. This study analyzes 15,812 Jordanian Arabic Dialect comments by employing various pre-processing techniques and feature vectors, including BERT-base-Arabic, Word2Vec, and FastText. Furthermore, the study considers a wide range of deep learning classifiers alongside balancing techniques to address data imbalance. The results demonstrate the superior performance of the CNN with BERT representation model compared to the BiLSTM, LSTM, RNN, and RNNLSTM models. The findings indicate an F1-score of approximately 96%, suggesting the potential for predicting patient sentiment from textual feedback. By automating feedback analysis, this approach enables KHCC to detect dissatisfaction, identify unmet needs, and act on key concerns promptly. It reduces the burden of manual review and supports data-driven service improvements—advancing KHCC’s mission to deliver responsive, high-quality, patient-centered cancer care.</div></div>","PeriodicalId":55458,"journal":{"name":"Artificial Intelligence in Medicine","volume":"172 ","pages":"Article 103334"},"PeriodicalIF":6.2,"publicationDate":"2025-12-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145790622","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

Regret theory-based clinical efficacy evaluation method with three-way decision 基于后悔理论的三向决策临床疗效评价方法

IF 6.2 2区医学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Artificial Intelligence in Medicine

Pub Date : 2025-12-10 DOI: 10.1016/j.artmed.2025.103333

Jin Ye , Bingzhen Sun , Jianming Zhan , Xiaoli Chu

Scientific assessment of clinical efficacy regarding treatment regimens is crucial for ensuring healthcare quality and improving patient prognosis. For most diseases, efficacy evaluation process of treatment regimens is a temporal process. Under the intervention of each treatment regimen, a patient’s efficacy information often alternates among improvement, deterioration, or stability of the condition. This dynamic variation demonstrates that the clinical efficacy evaluation problem is inherently a three-way classification problem involving temporal data. While numerous clinical efficacy evaluation methods have been developed, their practical utility is often constrained by limited interpretability in decision-making and insufficient accuracy—shortcomings that impede their alignment with the evolving demands of precision medicine. To this end, our study uses three-way decision (TWD) theory to propose an interpretable method for evaluating clinical efficacy. First, a multi-granularity unbalanced temporal incomplete hybrid decision system (MGUTIHDS) is introduced to represent the decision information of the problems considered. Second, in order to handle incomplete and hybrid attribute information, a reference sequence is employed to transform the MGUTIHDS into a normalized multi-granularity unbalanced temporal decision system (NMGUTDS). Considering the similarity of the patients’ condition who administered the same treatment regimen, we subsequently segment the NMGUTDS into multiple information systems related to treatment regimens. Based on this, we theoretically construct a data-driven TWD model by exploring the measurement of the loss functions and conditional probabilities. Further, a regret theory-based three-way efficacy evaluation method is devised to address the problems considered. Meanwhile, the associated decision-making framework and algorithmic implementation are presented. At last, an application research, i.e., regimen recommendation based on clinical data of rheumatoid arthritis is carried out. The experimental results demonstrate the performance and superiority of our model and method.

科学评价治疗方案的临床疗效对保证医疗质量和改善患者预后至关重要。对于大多数疾病，治疗方案的疗效评价过程是一个时间过程。在每种治疗方案的干预下，患者的疗效信息经常在病情改善、恶化或稳定之间交替。这种动态变化表明，临床疗效评价问题本质上是一个涉及时间数据的三向分类问题。虽然已经开发了许多临床疗效评估方法，但它们的实际效用往往受到决策可解释性有限和准确性不足的限制，这些缺点阻碍了它们与精密医学不断发展的需求保持一致。为此，本研究运用三向决策（TWD）理论，提出一种可解释的临床疗效评估方法。首先，引入一种多粒度非平衡时间不完全混合决策系统（MGUTIHDS）来表示所考虑问题的决策信息。其次，为了处理不完整和混合属性信息，采用参考序列将MGUTIHDS转换为规范化的多粒度不平衡时间决策系统（NMGUTDS）；考虑到接受相同治疗方案的患者病情的相似性，我们随后将NMGUTDS分割为与治疗方案相关的多个信息系统。在此基础上，通过探索损失函数和条件概率的度量，从理论上构建了数据驱动的TWD模型。在此基础上，提出了基于后悔理论的三方面疗效评价方法。同时，给出了相应的决策框架和算法实现。最后进行了基于类风湿关节炎临床资料的方案推荐应用研究。实验结果证明了该模型和方法的性能和优越性。

{"title":"Regret theory-based clinical efficacy evaluation method with three-way decision","authors":"Jin Ye , Bingzhen Sun , Jianming Zhan , Xiaoli Chu","doi":"10.1016/j.artmed.2025.103333","DOIUrl":"10.1016/j.artmed.2025.103333","url":null,"abstract":"<div><div>Scientific assessment of clinical efficacy regarding treatment regimens is crucial for ensuring healthcare quality and improving patient prognosis. For most diseases, efficacy evaluation process of treatment regimens is a temporal process. Under the intervention of each treatment regimen, a patient’s efficacy information often alternates among improvement, deterioration, or stability of the condition. This dynamic variation demonstrates that the clinical efficacy evaluation problem is inherently a three-way classification problem involving temporal data. While numerous clinical efficacy evaluation methods have been developed, their practical utility is often constrained by limited interpretability in decision-making and insufficient accuracy—shortcomings that impede their alignment with the evolving demands of precision medicine. To this end, our study uses three-way decision (TWD) theory to propose an interpretable method for evaluating clinical efficacy. First, a multi-granularity unbalanced temporal incomplete hybrid decision system (MGUTIHDS) is introduced to represent the decision information of the problems considered. Second, in order to handle incomplete and hybrid attribute information, a reference sequence is employed to transform the MGUTIHDS into a normalized multi-granularity unbalanced temporal decision system (NMGUTDS). Considering the similarity of the patients’ condition who administered the same treatment regimen, we subsequently segment the NMGUTDS into multiple information systems related to treatment regimens. Based on this, we theoretically construct a data-driven TWD model by exploring the measurement of the loss functions and conditional probabilities. Further, a regret theory-based three-way efficacy evaluation method is devised to address the problems considered. Meanwhile, the associated decision-making framework and algorithmic implementation are presented. At last, an application research, i.e., regimen recommendation based on clinical data of rheumatoid arthritis is carried out. The experimental results demonstrate the performance and superiority of our model and method.</div></div>","PeriodicalId":55458,"journal":{"name":"Artificial Intelligence in Medicine","volume":"172 ","pages":"Article 103333"},"PeriodicalIF":6.2,"publicationDate":"2025-12-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145738105","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

The accuracy, validity and reliability of Theia3D markerless motion capture for studying the biomechanics of human movement: A systematic review Theia3D无标记运动捕捉用于人体运动生物力学研究的准确性、有效性和可靠性：系统综述

IF 6.2 2区医学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Artificial Intelligence in Medicine

Pub Date : 2025-12-04 DOI: 10.1016/j.artmed.2025.103332

Florent Varcin , Mark G. Boocock

Recent advancements in computer vision recognition combined with the use of pose estimation algorithms has led to a rapid increase in the use of 3D video-based markerless (ML) motion capture to study human movement. One such prominent system is Theia3D. To determine the accuracy, validity, and reliability of Theia3D, a systematic literature review was conducted across five electronic databases using the PRISMA (Preferred Reporting Items for Systematic Review and Meta-Analysis) guidelines. Studies were included if they investigated the accuracy, validity, or reliability of Theia3D against a standardised method and reported on at least one biomechanical measure. A modified version of COSMIN (Consensus-based Standards for the Selection of Health Measurement Instruments) and GRADE (Grading of Recommendations Assessment, Development, and Evaluation) were used to evaluate the quality of evidence. Sixteen studies met the inclusion criteria, the majority of which assessed the validity of kinematics during gait or running. Pooled lower limb kinematics showed reasonable accuracy, whilst hip flexion/extension and rotations of the lower limb joints in the transverse plane suggests poor accuracy. Most spatiotemporal gait parameters measured using Theia3D demonstrated excellent validity (Intraclass correlation coefficient (ICC) > 0.9) and inter-session reliability (gait speed, Standard Error of Measurement (SEM) ≤ 0.07 m/s; step/stride length, SEM ≤ 0.06 m; ICC > 0.95). The accuracy, validity, and reliability of Theia3D used in the biomechanical analysis of functional tasks and in different population groups shows promise. However, there is a need for improved methods by which to compare data and a standardisation of biomechanical modelling approaches.

计算机视觉识别的最新进展与姿态估计算法的使用相结合，导致使用基于3D视频的无标记（ML）动作捕捉来研究人体运动的快速增加。其中一个突出的系统是Theia3D。为了确定Theia3D的准确性、有效性和可靠性，使用PRISMA（系统评价和荟萃分析首选报告项目）指南对五个电子数据库进行了系统的文献综述。如果研究针对标准化方法调查了Theia3D的准确性、有效性或可靠性，并报告了至少一项生物力学测量，则纳入研究。采用改良版的COSMIN（基于共识的健康测量工具选择标准）和GRADE（建议评估、发展和评价分级）来评估证据质量。16项研究符合纳入标准，其中大多数评估了步态或跑步时运动学的有效性。汇集的下肢运动学显示出合理的准确性，而髋关节屈伸和下肢关节在横平面上的旋转表明准确性较差。使用Theia3D测量的大多数时空步态参数具有良好的效度（类内相关系数（ICC） > 0.9）和会话间信度（步态速度，测量标准误差(SEM)≤0.07 m/s）；步长/步长，SEM≤0.06 m；ICC > 0.95)。Theia3D用于功能性任务和不同人群的生物力学分析的准确性、有效性和可靠性显示出前景。然而，需要改进的方法来比较数据和标准化的生物力学建模方法。

{"title":"The accuracy, validity and reliability of Theia3D markerless motion capture for studying the biomechanics of human movement: A systematic review","authors":"Florent Varcin , Mark G. Boocock","doi":"10.1016/j.artmed.2025.103332","DOIUrl":"10.1016/j.artmed.2025.103332","url":null,"abstract":"<div><div>Recent advancements in computer vision recognition combined with the use of pose estimation algorithms has led to a rapid increase in the use of 3D video-based markerless (ML) motion capture to study human movement. One such prominent system is Theia3D. To determine the accuracy, validity, and reliability of Theia3D, a systematic literature review was conducted across five electronic databases using the PRISMA (Preferred Reporting Items for Systematic Review and Meta-Analysis) guidelines. Studies were included if they investigated the accuracy, validity, or reliability of Theia3D against a standardised method and reported on at least one biomechanical measure. A modified version of COSMIN (Consensus-based Standards for the Selection of Health Measurement Instruments) and GRADE (Grading of Recommendations Assessment, Development, and Evaluation) were used to evaluate the quality of evidence. Sixteen studies met the inclusion criteria, the majority of which assessed the validity of kinematics during gait or running. Pooled lower limb kinematics showed reasonable accuracy, whilst hip flexion/extension and rotations of the lower limb joints in the transverse plane suggests poor accuracy. Most spatiotemporal gait parameters measured using Theia3D demonstrated excellent validity (Intraclass correlation coefficient (ICC) > 0.9) and inter-session reliability (gait speed, Standard Error of Measurement (SEM) ≤ 0.07 m/s; step/stride length, SEM ≤ 0.06 m; ICC > 0.95). The accuracy, validity, and reliability of Theia3D used in the biomechanical analysis of functional tasks and in different population groups shows promise. However, there is a need for improved methods by which to compare data and a standardisation of biomechanical modelling approaches.</div></div>","PeriodicalId":55458,"journal":{"name":"Artificial Intelligence in Medicine","volume":"173 ","pages":"Article 103332"},"PeriodicalIF":6.2,"publicationDate":"2025-12-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145928174","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

ProtoRadNet: Prototypical patches of Convolutional Features for Radiology Image Classification Network ProtoRadNet：用于放射学图像分类网络的卷积特征原型补丁

IF 6.2 2区医学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Artificial Intelligence in Medicine

Pub Date : 2025-12-03 DOI: 10.1016/j.artmed.2025.103324

Prateek Sarangi , Riya Agarwal , Tanmay Basu

Convolutional Neural Networks (CNNs) have achieved significant success in classifying radiology images; however, their implementation often resembles a ”black box,” limiting medical practitioners’ ability to comprehend and trust the decisions made due to a lack of interpretability. Recent advancements in patch-based prototypical networks have sought to enhance the interpretability of image classification systems. Still, the use of these models, specifically developed for the radiology domain, has been limited. This paper presents ProtoRadNet - Prototypical Patches of Convolutional Features for Radiology Image Classification Network. ProtoRadNet provides explicit visualisations of the prototypes identified within an image during classification tasks, thereby offering transparent reasoning for its decisions and effectively bridging the divide between CNN findings and their practical implications to the domain experts. The primary objective of ProtoRadNet is to identify significant prototypes of convolutional features within individual classes and across all classes, refining the CNN’s training to bolster interpretability rather than relying on all convolutional features indiscriminately. The model achieves localised and global interpretability by integrating inter-class and intra-class prototypes, enhancing overall decision-making processes. This interpretability is particularly noteworthy as it is accomplished using only image-level ground truths, rendering it semantically meaningful for real-world applications, where detailed annotations are frequently unavailable or time-consuming. Empirical evaluation demonstrates that ProtoRadNet surpasses state-of-the-art in most cases. It achieves macro-averaged F1-scores of

92.16 %, 96.14 %

and 29.32% with an improvement of

+ 2.04 %, + 0.73 %

and

+ 0.41 %

respectively than the best competing method on Brain MRI, Chest CT and MIMIC CXR-LT datasets. These results show the value and validity of our ProtoRadNet model.

卷积神经网络（cnn）在放射学图像分类方面取得了显著成功；然而，它们的实施往往类似于一个“黑盒子”，由于缺乏可解释性，限制了医生理解和信任做出的决定的能力。基于补丁的原型网络的最新进展已经寻求提高图像分类系统的可解释性。尽管如此，这些专门为放射学领域开发的模型的使用仍然受到限制。提出了一种用于放射学图像分类网络的卷积特征原型补丁ProtoRadNet。ProtoRadNet在分类任务中为图像中识别的原型提供了明确的可视化，从而为其决策提供了透明的推理，并有效地弥合了CNN发现与其对领域专家的实际影响之间的鸿沟。ProtoRadNet的主要目标是在单个类和所有类中识别卷积特征的重要原型，改进CNN的训练以增强可解释性，而不是不加区分地依赖所有卷积特征。该模型通过整合类间和类内原型实现了局部和全局可解释性，增强了整体决策过程。这种可解释性特别值得注意，因为它只使用图像级的基础事实来完成，使其在语义上对现实世界的应用程序有意义，在现实世界中，详细的注释经常不可用或耗时。实证评估表明，ProtoRadNet在大多数情况下都超越了最先进的技术。该方法在脑MRI、胸部CT和MIMIC CXR-LT数据集上的宏观平均f1评分分别为92.16%、96.14%和29.32%，分别比最佳竞争方法提高了+2.04%、+0.73%和+0.41%。这些结果表明了ProtoRadNet模型的价值和有效性。

{"title":"ProtoRadNet: Prototypical patches of Convolutional Features for Radiology Image Classification Network","authors":"Prateek Sarangi , Riya Agarwal , Tanmay Basu","doi":"10.1016/j.artmed.2025.103324","DOIUrl":"10.1016/j.artmed.2025.103324","url":null,"abstract":"<div><div>Convolutional Neural Networks (CNNs) have achieved significant success in classifying radiology images; however, their implementation often resembles a ”black box,” limiting medical practitioners’ ability to comprehend and trust the decisions made due to a lack of interpretability. Recent advancements in patch-based prototypical networks have sought to enhance the interpretability of image classification systems. Still, the use of these models, specifically developed for the radiology domain, has been limited. This paper presents ProtoRadNet - Prototypical Patches of Convolutional Features for Radiology Image Classification Network. ProtoRadNet provides explicit visualisations of the prototypes identified within an image during classification tasks, thereby offering transparent reasoning for its decisions and effectively bridging the divide between CNN findings and their practical implications to the domain experts. The primary objective of ProtoRadNet is to identify significant prototypes of convolutional features within individual classes and across all classes, refining the CNN’s training to bolster interpretability rather than relying on all convolutional features indiscriminately. The model achieves localised and global interpretability by integrating inter-class and intra-class prototypes, enhancing overall decision-making processes. This interpretability is particularly noteworthy as it is accomplished using only image-level ground truths, rendering it semantically meaningful for real-world applications, where detailed annotations are frequently unavailable or time-consuming. Empirical evaluation demonstrates that ProtoRadNet surpasses state-of-the-art in most cases. It achieves macro-averaged F1-scores of <span><math><mrow><mn>92</mn><mo>.</mo><mn>16</mn><mtext>%</mtext><mo>,</mo><mn>96</mn><mo>.</mo><mn>14</mn><mtext>%</mtext></mrow></math></span> and 29.32% with an improvement of <span><math><mrow><mo>+</mo><mn>2</mn><mo>.</mo><mn>04</mn><mtext>%</mtext><mo>,</mo><mo>+</mo><mn>0</mn><mo>.</mo><mn>73</mn><mtext>%</mtext></mrow></math></span> and <span><math><mrow><mo>+</mo><mn>0</mn><mo>.</mo><mn>41</mn><mtext>%</mtext></mrow></math></span> respectively than the best competing method on Brain MRI, Chest CT and MIMIC CXR-LT datasets. These results show the value and validity of our ProtoRadNet model.</div></div>","PeriodicalId":55458,"journal":{"name":"Artificial Intelligence in Medicine","volume":"172 ","pages":"Article 103324"},"PeriodicalIF":6.2,"publicationDate":"2025-12-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145685085","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

TransFair: Transferring fairness from ocular disease classification to progression prediction TransFair：将公平性从眼病分类转移到进展预测

IF 6.2 2区医学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Artificial Intelligence in Medicine

Pub Date : 2025-12-01 DOI: 10.1016/j.artmed.2025.103331

Min Shi , Leila Gheisi , Chee-Hung Henry Chu , Raju Gottumukkala , Yan Luo , Mengyu Wang , Xingquan Zhu

The use of artificial intelligence (AI) in automated disease classification significantly reduces healthcare costs and improves the accessibility of services. However, this transformation has given rise to concerns about the fairness of AI, which disproportionately affects certain groups, particularly patients from underprivileged populations. Recently, a number of methods and large-scale datasets have been proposed to address group performance disparities. Although these methods have shown effectiveness in disease classification tasks, they may fall short in ensuring fair prediction of disease progression, mainly because of limited longitudinal data with diverse demographics available for training a robust and equitable prediction model. In this paper, we introduce TransFair to enhance demographic fairness in progression prediction for ocular diseases. TransFair aims to transfer a fairness-enhanced disease classification model to the task of progression prediction with fairness preserved. First, we train a fairness-aware EfficientNet called FairEN using extensive data for ocular disease classification. Subsequently, this fair classification model is adapted to a fair progression prediction model through knowledge distillation, which minimizes the latent feature distances between classification and progression prediction models. We evaluate FairEN and TransFair for fairness-enhanced ocular disease classification and progression prediction using both two-dimensional (2D) and 3D retinal images. Extensive experiments and comparisons show that TransFair enhances group fairness in predicting ocular disease progression.

在自动疾病分类中使用人工智能（AI）可显着降低医疗成本并提高服务的可及性。然而，这种转变引发了人们对人工智能公平性的担忧，人工智能对某些群体的影响不成比例，尤其是来自弱势群体的患者。最近，已经提出了许多方法和大规模数据集来解决群体绩效差异。尽管这些方法在疾病分类任务中显示出有效性，但它们在确保公平预测疾病进展方面可能存在不足，主要是因为可用于训练稳健和公平预测模型的不同人口统计数据的纵向数据有限。在本文中，我们引入TransFair来提高眼科疾病进展预测的人口公平性。TransFair旨在将公平性增强的疾病分类模型转移到保持公平性的进展预测任务中。首先，我们使用大量的眼部疾病分类数据来训练一个具有公平性意识的高效网络FairEN。然后，通过知识蒸馏将该公平分类模型适应为公平递进预测模型，使分类模型和递进预测模型之间的潜在特征距离最小化。我们使用二维（2D）和三维视网膜图像对FairEN和TransFair进行公平性增强的眼部疾病分类和进展预测。广泛的实验和比较表明，TransFair在预测眼部疾病进展方面增强了群体公平性。

{"title":"TransFair: Transferring fairness from ocular disease classification to progression prediction","authors":"Min Shi , Leila Gheisi , Chee-Hung Henry Chu , Raju Gottumukkala , Yan Luo , Mengyu Wang , Xingquan Zhu","doi":"10.1016/j.artmed.2025.103331","DOIUrl":"10.1016/j.artmed.2025.103331","url":null,"abstract":"<div><div>The use of artificial intelligence (AI) in automated disease classification significantly reduces healthcare costs and improves the accessibility of services. However, this transformation has given rise to concerns about the fairness of AI, which disproportionately affects certain groups, particularly patients from underprivileged populations. Recently, a number of methods and large-scale datasets have been proposed to address group performance disparities. Although these methods have shown effectiveness in disease classification tasks, they may fall short in ensuring fair prediction of disease progression, mainly because of limited longitudinal data with diverse demographics available for training a robust and equitable prediction model. In this paper, we introduce TransFair to enhance demographic fairness in progression prediction for ocular diseases. TransFair aims to transfer a fairness-enhanced disease classification model to the task of progression prediction with fairness preserved. First, we train a fairness-aware EfficientNet called FairEN using extensive data for ocular disease classification. Subsequently, this fair classification model is adapted to a fair progression prediction model through knowledge distillation, which minimizes the latent feature distances between classification and progression prediction models. We evaluate FairEN and TransFair for fairness-enhanced ocular disease classification and progression prediction using both two-dimensional (2D) and 3D retinal images. Extensive experiments and comparisons show that TransFair enhances group fairness in predicting ocular disease progression.</div></div>","PeriodicalId":55458,"journal":{"name":"Artificial Intelligence in Medicine","volume":"172 ","pages":"Article 103331"},"PeriodicalIF":6.2,"publicationDate":"2025-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145738104","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

Geometric deep learning for local growth prediction on abdominal aortic aneurysm surfaces 基于几何深度学习的腹主动脉瘤表面局部生长预测。

IF 6.2 2区医学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Artificial Intelligence in Medicine

Pub Date : 2025-11-29 DOI: 10.1016/j.artmed.2025.103323

Dieuwertje Alblas , Patryk Rygiel , Julian Suk , Kaj O. Kappe , Marieke Hofman , Christoph Brune , Kak Khee Yeung , Jelmer M. Wolterink

Abdominal aortic aneurysms (AAAs) are progressive focal dilatations of the abdominal aorta. AAAs may rupture, with fatal consequences in

>

80% of cases. Current clinical guidelines recommend elective surgical repair when the maximum AAA diameter exceeds 55 mm in men or 50 mm in women. Patients that do not meet these criteria are periodically monitored, with surveillance intervals based on the maximum AAA diameter. However, this diameter does not take into account the complex relation between the 3D AAA shape and its growth, making standardized intervals potentially unfit. Personalized AAA growth predictions could improve monitoring strategies. We propose to use an SE(3)-symmetric transformer model to predict AAA growth directly on the vascular model surface enriched with local, multi-physical features. In contrast to other works which have parameterized the AAA shape, this representation preserves the vascular surface’s anatomical structure and geometric fidelity. We train our model using a longitudinal dataset of 113 computed tomography angiography (CTA) scans of 24 AAA patients at irregularly sampled intervals. After training, our model predicts AAA growth to the next scan moment with a median diameter error of 1.18 mm. We further demonstrate our model’s utility to identify whether a patient will become eligible for elective repair within two years (acc = 0.93). Finally, we evaluate our model’s generalization on an external validation set consisting of 25 CTAs from 7 AAA patients from a different hospital. Our results show that local directional AAA growth prediction from the vascular surface is feasible and may contribute to personalized surveillance strategies.

腹主动脉瘤（AAAs）是腹主动脉进行性局灶性扩张。AAAs可能破裂，在80%的病例中有致命的后果。目前的临床指南建议，当男性最大AAA直径超过55mm或女性超过50mm时，可选择手术修复。不符合这些标准的患者定期监测，监测间隔基于最大AAA直径。然而，该直径没有考虑到3D AAA形状与其生长之间的复杂关系，使得标准化间隔可能不合适。个性化的AAA增长预测可以改善监测策略。我们建议使用SE(3)对称变压器模型来直接预测具有局部多物理特征的血管模型表面上的AAA生长。与其他将AAA形状参数化的作品相比，这种表现保留了血管表面的解剖结构和几何保真度。我们使用24名AAA患者的113次计算机断层血管造影（CTA）纵向数据集不规则采样间隔来训练我们的模型。经过训练，我们的模型预测到下一个扫描时刻的AAA生长，中值直径误差为1.18 mm。我们进一步证明了我们的模型的效用，以确定患者是否有资格在两年内进行选择性修复（acc = 0.93）。最后，我们在由来自不同医院的7名AAA级患者的25个cta组成的外部验证集上评估了模型的泛化性。我们的研究结果表明，从血管表面预测局部定向AAA生长是可行的，并可能有助于个性化的监测策略。

{"title":"Geometric deep learning for local growth prediction on abdominal aortic aneurysm surfaces","authors":"Dieuwertje Alblas , Patryk Rygiel , Julian Suk , Kaj O. Kappe , Marieke Hofman , Christoph Brune , Kak Khee Yeung , Jelmer M. Wolterink","doi":"10.1016/j.artmed.2025.103323","DOIUrl":"10.1016/j.artmed.2025.103323","url":null,"abstract":"<div><div>Abdominal aortic aneurysms (AAAs) are progressive focal dilatations of the abdominal aorta. AAAs may rupture, with fatal consequences in <span><math><mo>></mo></math></span>80% of cases. Current clinical guidelines recommend elective surgical repair when the maximum AAA diameter exceeds 55 mm in men or 50 mm in women. Patients that do not meet these criteria are periodically monitored, with surveillance intervals based on the maximum AAA diameter. However, this diameter does not take into account the complex relation between the 3D AAA shape and its growth, making standardized intervals potentially unfit. Personalized AAA growth predictions could improve monitoring strategies. We propose to use an SE(3)-symmetric transformer model to predict AAA growth directly on the vascular model surface enriched with local, multi-physical features. In contrast to other works which have parameterized the AAA shape, this representation preserves the vascular surface’s anatomical structure and geometric fidelity. We train our model using a longitudinal dataset of 113 computed tomography angiography (CTA) scans of 24 AAA patients at irregularly sampled intervals. After training, our model predicts AAA growth to the next scan moment with a median diameter error of 1.18 mm. We further demonstrate our model’s utility to identify whether a patient will become eligible for elective repair within two years (acc = 0.93). Finally, we evaluate our model’s generalization on an external validation set consisting of 25 CTAs from 7 AAA patients from a different hospital. Our results show that local directional AAA growth prediction from the vascular surface is feasible and may contribute to personalized surveillance strategies.</div></div>","PeriodicalId":55458,"journal":{"name":"Artificial Intelligence in Medicine","volume":"172 ","pages":"Article 103323"},"PeriodicalIF":6.2,"publicationDate":"2025-11-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145662935","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