Artificial Intelligence in Medicine最新文献

{"title":"EEG super-resolution with Laplacian Regularized Coupled Matrix Decomposition: A case study of Autism Spectrum Disorder EEG enhancement.","authors":"Yunbo Tang, Qifeng Lin, Yuanlong Yu, Dan Chen","doi":"10.1016/j.artmed.2025.103284","DOIUrl":"https://doi.org/10.1016/j.artmed.2025.103284","url":null,"abstract":"<p><p>EEG Super-resolution (SR) has attracted increasing attention for neuroscience research when fine-grained spatial information is demanding. However, existing SR methods are subject to the performance bottlenecks due to insufficient high-resolution EEG under the condition of few participants undergoing high-density EEG acquisition and unclear intrinsic spatiotemporal relationship amongst EEG channels on the scalp. To tackle the issues, this study proposes a Laplacian Regularized Coupled Matrix Decomposition (LRCMD) model for EEG SR, which takes in HR EEG of a small amount of initial participants and generates HR EEG from the given LR counterparts of new participants: (1) LRCMD first utilizes a Gaussian kernel function according to the spatial distribution of electrodes conforming to a 3-D scalp model to measure the brain structural connectivity amongst EEG channels, (2) Coupled matrix decomposition model is established to transform the HR EEG and the corresponding LR ones to latent source space with common mapping rule, where brain structural connectivity acts as Laplacian regularization to highlight the core mapping rule, (3) LRCMD applies Alternating Direction Method of Multipliers solver to cope with the decomposition model and derive the mapping matrix along with latent source of HR EEG, which are later leveraged to complete the SR reconstruction of LR EEG from new participants. Experimental results on ASD EEG dataset indicate that (1) LRCMD excels in individual EEG super-resolution reconstruction with normalized mean squared error decreased by 2.14% and the improvements of signal-to-noise ratio, Pearson's correlation coefficient respectively reaching 0.52 dB, 1.17%, and (2) the reconstructed EEG by LRCMD demonstrates superiority to LR alternative in ASD discrimination and functional connectivity analysis of ASD.</p>","PeriodicalId":55458,"journal":{"name":"Artificial Intelligence in Medicine","volume":"170 ","pages":"103284"},"PeriodicalIF":6.2,"publicationDate":"2025-10-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145350272","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}

{"title":"Preprocessing narrative texts in electronic medical records to identify hospital adverse events: A scoping review","authors":"Hamed Jafarpour ,&nbsp;Guosong Wu ,&nbsp;Cheligeer (Ken) Cheligeer ,&nbsp;Jun Yan ,&nbsp;Yuan Xu ,&nbsp;Danielle A. Southern ,&nbsp;Cathy A. Eastwood ,&nbsp;Yong Zeng ,&nbsp;Hude Quan","doi":"10.1016/j.artmed.2025.103281","DOIUrl":"10.1016/j.artmed.2025.103281","url":null,"abstract":"&lt;div&gt;&lt;h3&gt;Background:&lt;/h3&gt;&lt;div&gt;Narrative electronic medical records (EMR), which include textual notes created by clinicians within healthcare environments, represent a significant resource for documenting various facets of patient care. This form of text exhibits distinctive characteristics, such as the occurrence of grammatically incorrect sentences, abbreviations, frequent acronyms, specialized characters with particular meanings, negation expressions, and sporadic misspellings. As a result, a primary goal in processing these textual notes is to implement effective preprocessing techniques that enhance data quality and ensure consistency across all entries. Recent advancements in algorithms and methodologies within the fields of natural language processing (NLP), machine learning (ML), and large language models (LLM) have prompted researchers to leverage narrative EMR for the detection of hospital adverse events (HAE).&lt;/div&gt;&lt;/div&gt;&lt;div&gt;&lt;h3&gt;Methods:&lt;/h3&gt;&lt;div&gt;The scoping review adhered to the PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses Extension for Scoping Reviews) guidelines. A scoping review protocol was developed and utilized to guide the research process, clearly outlining the eligibility criteria, information sources, search strategies, data management, selection process, data collection procedures, data items, outcomes and prioritization, data synthesis, and meta-bias considerations. The search strategy was implemented across nine engineering and medical electronic databases.&lt;/div&gt;&lt;/div&gt;&lt;div&gt;&lt;h3&gt;Results:&lt;/h3&gt;&lt;div&gt;The results have indicated that from a total of 3,264 studies retrieved, 48 unique studies were included in the review. Responses to the research questions were systematically extracted from these studies. The review has identified challenges associated with the preprocessing of narrative texts in EMR for HAE identification. Additionally, three research gaps have been identified: (1) the imperative need for a pipeline to preprocess narrative EMR for the identification of HAE, (2) the necessity for a robust system capable of managing the extensive volume of narrative EMR data, and (3) the requirement for temporal event system, which are essential for effective HAE detection. The study also has underscored the essential role of preprocessing tasks in enhancing the performance of HAE detection. The study has emphasized the importance of extracting N-grams from clinical text, normalizing these N-grams through lemmatization and/or stemming, and establishing semantic feature extraction in preprocessing tasks that significantly affect HAE detection performance. While LLM-based systems naturally incorporate tokenization and normalization processes within their frameworks, it remains crucial to address features that hold semantic relevance to the specific type of HAE during preprocessing.&lt;/div&gt;&lt;/div&gt;&lt;div&gt;&lt;h3&gt;Conclusion:&lt;/h3&gt;&lt;div&gt;This scoping review has provided valuable insights for researchers focused on ","PeriodicalId":55458,"journal":{"name":"Artificial Intelligence in Medicine","volume":"170 ","pages":"Article 103281"},"PeriodicalIF":6.2,"publicationDate":"2025-10-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145267614","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}

{"title":"TSFNet: A Temporal–Spectral Fusion Network for advanced speech emotion recognition in medical applications","authors":"Xinran Li ,&nbsp;Peilin Huang ,&nbsp;Xiaojiang Peng ,&nbsp;Feng Sha ,&nbsp;Xiaomao Fan ,&nbsp;Ye Li","doi":"10.1016/j.artmed.2025.103279","DOIUrl":"10.1016/j.artmed.2025.103279","url":null,"abstract":"<div><div>Speech emotion recognition (SER) is a critical component in enhancing communication systems and human–machine interaction, with significant potential for applications in the medical field. Although existing SER methods that combine temporal and spectral features have achieved notable advancements, they still encounter a big challenge in capturing emotional nuances, which are vital in medical diagnostics and patient care. In this study, we introduce a straightforward yet highly efficient network called TSFNet, which is the Temporal–Spectral Fusion Network via a Large-scale Pre-trained Model. This network is specifically designed to effectively process intricate emotional nuances by seamlessly integrating temporal and spectral information present in speech signals. By leveraging the capabilities of a large-scale pre-trained model, which serves as a powerful plug-and-play component for extracting and learning the temporal characteristics of speech, TSFNet enables a more accurate capture of complex emotional details crucial for medical applications. Extensive experiments are conducted on publicly available datasets, to evaluate the performance of TSFNet. Extensive experiments conducted on six public datasets demonstrate that TSFNet significantly outperforms existing baselines, achieving unweighted accuracies of 95.57% for Savee, 92.67% for Crema-D, 85.71% for IEMOCAP, 100.00% for Tess, 95.86% for Emovo, and 80.43% for Meld. It means that TSFNet has the potential in advancing medical diagnostic tools and patient monitoring systems.</div></div>","PeriodicalId":55458,"journal":{"name":"Artificial Intelligence in Medicine","volume":"170 ","pages":"Article 103279"},"PeriodicalIF":6.2,"publicationDate":"2025-10-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145234227","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}

{"title":"Mixture-attention Siamese transformer for video polyp segmentation","authors":"Geng Chen ,&nbsp;Junqing Yang ,&nbsp;Xiaozhou Pu ,&nbsp;Ge-Peng Ji ,&nbsp;Huan Xiong ,&nbsp;Yongsheng Pan ,&nbsp;Hengfei Cui ,&nbsp;Yong Xia","doi":"10.1016/j.artmed.2025.103278","DOIUrl":"10.1016/j.artmed.2025.103278","url":null,"abstract":"<div><div>Accurate segmentation of polyps from colonoscopy videos is of great significance to polyp treatment and early prevention of colorectal cancer. However, it is challenging due to the difficulties associated with modeling long-range spatio-temporal relationships within a colonoscopy video. In this paper, we address this challenging task with a novel <strong>M</strong>ixture-<strong>A</strong>ttention <strong>S</strong>iamese <strong>T</strong>ransformer (<strong>MAST</strong>), which explicitly models the long-range spatio-temporal relationships with a mixture-attention mechanism for accurate polyp segmentation. Specifically, we first construct a Siamese transformer architecture to jointly encode paired video frames for their feature representations. We then design a mixture-attention module to exploit the intra-frame and inter-frame correlations, enhancing the features with rich spatio-temporal relationships. Finally, the enhanced features are fed to two parallel decoders for predicting the segmentation maps. Extensive experiments on the large-scale SUN-SEG benchmark demonstrate the superior performance of MAST in comparison with the cutting-edge competitors. Our code is publicly available at <span><span>https://github.com/Junqing-Yang/MAST</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":55458,"journal":{"name":"Artificial Intelligence in Medicine","volume":"170 ","pages":"Article 103278"},"PeriodicalIF":6.2,"publicationDate":"2025-10-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145253941","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}

{"title":"SCLResNet and DSAF: A self-supervised contrastive learning and deep self-attention fusion-based multimodal network for predicting central lymph node metastasis in papillary thyroid carcinoma","authors":"Shidi Miao ,&nbsp;Yuyang Jiang ,&nbsp;Wenjuan Huang ,&nbsp;Yuxin Jiang ,&nbsp;Mengzhuo Sun ,&nbsp;Mingxuan Wang ,&nbsp;Hongzhuo Qi ,&nbsp;Ao Li ,&nbsp;Zengyao Liu ,&nbsp;Qiujun Wang ,&nbsp;Ruitao Wang ,&nbsp;Xuemei Ding","doi":"10.1016/j.artmed.2025.103280","DOIUrl":"10.1016/j.artmed.2025.103280","url":null,"abstract":"<div><div>Accurate prediction of central lymph node metastasis (CLNM) in papillary thyroid carcinoma (PTC) is crucial to avoid unnecessary invasive procedures, yet existing models often fall short. We constructed the SCLResNet101 model based on a contrastive learning framework to extract network features of tumor ultrasound (US). SeResnet101 was used to extract network features of peri-vascular adipose tissue (PVAT) from the computed tomography (CT) of C6 (the arterial and venous layers beneath the thyroid). Univariate and multivariate analyses were performed using binary logistic regression to select clinical features. Finally, we constructed a Deep Self-Attention Fusion (DSAF) network to integrate features from these three modalities for CLNM prediction. Univariate and multivariate analyses revealed that Gender, Age, Size of US, and Extrathyroidal Extension (ETE) were independent risk factors for CLNM. In the internal test cohort (I-T), the area under the curve (AUC) of model was 0.863 (95 % CI: 0.779–0.932). In the external test cohort (E-T), the AUC was 0.839 (95 % CI: 0.755–0.905). Compared to all radiologists, the model significantly reduced both false-positive and false-negative rates in both the I-T and E-T. This study incorporates PVAT, which significantly enhances the performance of the multimodal deep learning model and may assist surgeons in making more informed and precise surgical decisions in the treatment of PTC.</div></div>","PeriodicalId":55458,"journal":{"name":"Artificial Intelligence in Medicine","volume":"170 ","pages":"Article 103280"},"PeriodicalIF":6.2,"publicationDate":"2025-10-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145253944","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}

{"title":"BPINet: Synchronous blood pressure estimation and user authentication based on ECG and PPG signal with multi-task learning","authors":"Xianliang Jiang ,&nbsp;Dingxin Yu ,&nbsp;Guang Jin ,&nbsp;Fei Lei ,&nbsp;Weihao Zhang ,&nbsp;Xinyan Zhou","doi":"10.1016/j.artmed.2025.103277","DOIUrl":"10.1016/j.artmed.2025.103277","url":null,"abstract":"<div><div>As a vital indicator of health, blood pressure is particularly important for elderly individuals with chronic illnesses who live alone. Daily monitoring is essential to prevent hypertension and related complications. Nevertheless, most current home blood pressure monitors depend on cuff-based techniques, which can produce inaccurate results through improper use or cuff placement. Moreover, these devices generally cannot identify the specific user being measured, which hinders the development of personalized long-term health monitoring reports. In this paper, we propose BPINet, a multi-task model based on the Multi-gate Mixture-of-Experts (MMoE) framework that utilizes CNN-BiLSTM to extract features from ECG/PPG signals for simultaneous blood pressure estimation and user authentication (identity recognition). We also compile a dataset of ECG/PPG signals from multiple families, along with their blood pressure measurements, and incorporate it with the University of Queensland Vital Signs Dataset (UQVS) to evaluate the performance of BPINet. On the UQVS dataset, BPINet achieves a 97.54% user identity recognition accuracy. For systolic blood pressure (SBP) estimation, BPINet yields an MAE ± STD of 3.317 ± 5.771 mmHg. For diastolic blood pressure (DBP) estimation, the corresponding values are 2.444 ± 4.147 mmHg. On our customized dataset, BPINet achieves a 94.30% user identity recognition accuracy. For SBP estimation, it yields an MAE ± STD of 2.940 ± 4.753 mmHg. These results meet both the British Hypertension Society (BHS) Grade A standard and the Association for the Advancement of Medical Instrumentation (AAMI) standard. BPINet not only performs blood pressure estimation effectively but also enables simultaneous user identity recognition, facilitating the creation of personalized health records. The experimental results demonstrate the clinical feasibility and effectiveness of our proposed scheme.</div></div>","PeriodicalId":55458,"journal":{"name":"Artificial Intelligence in Medicine","volume":"170 ","pages":"Article 103277"},"PeriodicalIF":6.2,"publicationDate":"2025-09-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145202105","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}

{"title":"Evaluating AI-driven characters in extended reality (XR) healthcare simulations: A systematic review","authors":"David Dasa ,&nbsp;Michele Board ,&nbsp;Ursula Rolfe ,&nbsp;Tom Dolby ,&nbsp;Wen Tang","doi":"10.1016/j.artmed.2025.103270","DOIUrl":"10.1016/j.artmed.2025.103270","url":null,"abstract":"<div><div>AI-driven characters in extended reality (XR) healthcare simulations are increasingly used for clinical training, yet their effectiveness, implementation, and quality assurance remain poorly understood.</div><div>We conducted a systematic review of 132 studies published between January 2015 and July 2025, including 11 randomized controlled trials (RCTs), sourced from biomedical, computing, and education databases and targeted proceedings. Most studies used virtual reality (62.1%) and focused on effectiveness (n = 71), with fewer examining implementation (n = 45) or quality assurance (n = 44). Meta-analysis of two RCTs found a large effect on knowledge and decision-making (Hedges’ g = 1.31, 95% CI 0.08–2.54, <span><math><msup><mrow><mi>I</mi></mrow><mrow><mn>2</mn></mrow></msup></math></span> = 85%), while one RCT reported faster task performance with AI-driven characters (g = -0.68, 95% CI -1.32 to -0.04). Certainty of evidence was low due to small samples and high heterogeneity. Implementation success was often associated with phased roll-outs and faculty training, but quality assurance practices (particularly bias audits and transparency measures) were rarely documented.</div><div>The review proposes the DASEX framework to address these gaps and guide future integration of AI-driven characters in XR training.</div></div>","PeriodicalId":55458,"journal":{"name":"Artificial Intelligence in Medicine","volume":"170 ","pages":"Article 103270"},"PeriodicalIF":6.2,"publicationDate":"2025-09-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145304645","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}

{"title":"A survey for large language models in biomedicine","authors":"Chong Wang ,&nbsp;Mengyao Li ,&nbsp;Junjun He ,&nbsp;Zhongruo Wang ,&nbsp;Erfan Darzi ,&nbsp;Zan Chen ,&nbsp;Jin Ye ,&nbsp;Tianbin Li ,&nbsp;Yanzhou Su ,&nbsp;Jing Ke ,&nbsp;Kaili Qu ,&nbsp;Shuxin Li ,&nbsp;Yi Yu ,&nbsp;Pietro Liò ,&nbsp;Tianyun Wang ,&nbsp;Yu Guang Wang ,&nbsp;Yiqing Shen","doi":"10.1016/j.artmed.2025.103268","DOIUrl":"10.1016/j.artmed.2025.103268","url":null,"abstract":"<div><div>Recent breakthroughs in large language models (LLMs) offer unprecedented natural language understanding and generation capabilities. However, existing surveys on LLMs in biomedicine often focus on specific applications or model architectures, lacking a comprehensive analysis that integrates the latest advancements across various biomedical domains. This review, based on an analysis of 484 publications sourced from databases including PubMed, Web of Science, and arXiv, provides an in-depth examination of the current landscape, applications, challenges, and prospects of LLMs in biomedicine, distinguishing itself by focusing on the practical implications of these models in real-world biomedical contexts. Firstly, we explore the capabilities of LLMs in zero-shot learning across a broad spectrum of biomedical tasks, including diagnostic assistance, drug discovery, and personalized medicine, among others, with insights drawn from 137 key studies. Then, we discuss adaptation strategies of LLMs, including fine-tuning methods for both uni-modal and multi-modal LLMs to enhance their performance in specialized biomedical contexts where zero-shot fails to achieve, such as medical question answering and efficient processing of biomedical literature. Finally, we discuss the challenges that LLMs face in the biomedicine domain including data privacy concerns, limited model interpretability, issues with dataset quality, and ethics due to the sensitive nature of biomedical data, the need for highly reliable model outputs, and the ethical implications of deploying AI in healthcare. To address these challenges, we also identify future research directions of LLM in biomedicine including federated learning methods to preserve data privacy and integrating explainable AI methodologies to enhance the transparency of LLMs. As this field of LLM rapidly evolves, continued research and development are essential to fully harness the capabilities of LLMs in biomedicine while ensuring their responsible and effective deployment.</div></div>","PeriodicalId":55458,"journal":{"name":"Artificial Intelligence in Medicine","volume":"170 ","pages":"Article 103268"},"PeriodicalIF":6.2,"publicationDate":"2025-09-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145152027","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}

{"title":"Types, functions, and mechanisms of machine learning for personalizing smoking cessation interventions: A systematic scoping review","authors":"Yu Jie Xavia Ng ,&nbsp;Shing Hui Reina Cheong ,&nbsp;Wen Wei Ang ,&nbsp;Ying Lau ,&nbsp;Siew Tiang Lau","doi":"10.1016/j.artmed.2025.103274","DOIUrl":"10.1016/j.artmed.2025.103274","url":null,"abstract":"<div><h3>Purpose</h3><div>Artificial intelligence can realize personalization. This systematic scoping review provides the types, functions, and mechanisms of machine learning (ML) for personalizing smoking cessation interventions.</div></div><div><h3>Methodology</h3><div>We searched fourteen databases including PubMed, CINAHL, EMBASE, the Cochrane Library, IEEE Xplore, PsycINFO, Scopus, Web of Science, AAAI, ACM Digital Library, ArXIV, Mednar, ProQuest, and <span><span>Science.gov</span><svg><path></path></svg></span>. We selected 98 articles from 4073 records that met the criteria. Two independent reviewers screened and selected the articles. Two reviewers extracted the data using a self-developed data charting form independently.</div></div><div><h3>Results</h3><div>The findings are reported in narrative syntheses, tables, and figures. The types of ML included artificial neural networks, Bayesian algorithms, clustering algorithms, decision tree algorithms, deep learning (DL) algorithms, ensemble algorithms, linear classifiers, others, and unspecified. The most common ML technique used was supervised learning (81 %), and the ML functions included (1) message tailoring (17 %), (2) prediction and detection of smoking events (34 %), (3) social media surveillance (14 %), (4) predictive models (24 %), and (5) biomarker analysis (10 %). The ML mechanisms involved the following sequence: data input, data preprocessing, feature extraction and selection, training and validation, and data output.</div></div><div><h3>Conclusion</h3><div>This review is the first to describe the potential use of ML for personalizing smoking cessation interventions. We provide recommendations for future research by identifying the limitations and gaps in the studies. Future studies should refine, validate, and test ML models using robust experimental methods to conclude their effectiveness.</div></div>","PeriodicalId":55458,"journal":{"name":"Artificial Intelligence in Medicine","volume":"170 ","pages":"Article 103274"},"PeriodicalIF":6.2,"publicationDate":"2025-09-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145158830","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}

{"title":"A novel memory interaction neural network for multi-label drug–drug interaction prediction with neighbor importance sampling","authors":"Jing Wang ,&nbsp;Runzhi Li ,&nbsp;Shuo Zhang ,&nbsp;YunLi Xing ,&nbsp;Siyu Yan ,&nbsp;Lihong Ma","doi":"10.1016/j.artmed.2025.103275","DOIUrl":"10.1016/j.artmed.2025.103275","url":null,"abstract":"<div><div>Co-administration of multiple drugs can frequently cause drug–drug interactions (DDIs), including adverse drug reactions (ADRs) that may increase the likelihood of morbidity and mortality. Identifying potential DDIs presents a significant challenge, due to the complexity of pharmacology. Recent advances in knowledge graphs have contributed to DDI prediction by providing a robust framework for representing various relationships between drugs and other entities, such as proteins, diseases, and drug attributes. However, current network-based models often fail to uncover interaction information among DDI triplets, as they typically encode triplets independently. Additionally, uniform sampling methods may overlook differences in neighboring node properties. In this work, we propose a novel memory interaction neural network for DDI prediction, which integrates drug molecular sequences with semantic information from the drug knowledge graph. Specifically, we introduce a neighbor importance sampling strategy that selectively samples highly connected neighbors, improving computational efficiency and reducing noise. We also design a memory interaction module that utilizes multi-head attention mechanisms and deep neural networks to capture interactions among DDI triplets. Experimental evaluation on KEGG and OGB-biokg datasets demonstrates the superiority of our model compared to classical and state-of-the-art methods in predicting DDIs. Datasets and code for this proposed DDIs prediction model are freely accessible at <span><span>https://github.com/wj1108114106/Multi-label-DDIs</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":55458,"journal":{"name":"Artificial Intelligence in Medicine","volume":"170 ","pages":"Article 103275"},"PeriodicalIF":6.2,"publicationDate":"2025-09-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145121253","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}