International Journal of Medical Informatics最新文献_第7页

Unsupervised tooth segmentation from three dimensional scans of the dental arch using domain adaptation of synthetic data 基于合成数据域自适应的牙弓三维扫描无监督牙齿分割。

IF 3.7 2区医学 Q2 COMPUTER SCIENCE, INFORMATION SYSTEMS

International Journal of Medical Informatics

Pub Date : 2024-12-19 DOI: 10.1016/j.ijmedinf.2024.105769

Md Sahadul Hasan Arian , Faisal Ahmed Sifat , Saif Ahmed , Nabeel Mohammed , Taseef Hasan Farook

Background

The automated segmentation of individual teeth from 3D models of the human dental arch is challenging due to variations in tooth alignment, arch form and overall maxillofacial anatomy. Domain adaptation is a specialised technique in deep learning which allows models to adapt to data from different domains, such as varying tooth and dental arch forms, without requiring human annotations.

Purpose

This study aimed to segment individual teeth from various dental arch morphologies in 3D intraoral scans using domain adaptation.

Materials and Methods

Twenty scanned dental arches from various age groups and developmental stages were used to generate 20 simplified synthetic variants of the scans. These synthetic variants, along with 16 natural scanned dental arches, were used to train the deep learning models. Domain adaptation was employed using Gradient Reversal Layer and Siamese Network techniques. The PointNet and PointNet++ model backbones were trained to align the latent space distribution of real and synthetic domains. Validations were performed on four unseen natural scanned arches, with and without domain adaptation enabled, to evaluate whether a 3D deep neural network can be trained without any human-annotated 3D models.

Results

PointNet and PointNet++ models demonstrated a mean intersection-over-union between 0.34 and 0.36 mIoU without domain adaptation enabled and 0.80 and 0.95 mIoU, respectively with domain adaptation enabled when assessing natural scanned dental arches.

Conclusion

Domain adaptation techniques can enable training a segmentation deep learning model using synthetically generated 3D jaw scans without requiring human operators annotating the training data.

背景：由于牙齿排列、牙弓形状和整体颌面解剖结构的变化，从人类牙弓三维模型中自动分割单个牙齿是具有挑战性的。领域适应是深度学习中的一种专门技术，它允许模型适应来自不同领域的数据，例如不同的牙齿和牙弓形式，而不需要人工注释。目的：本研究旨在利用区域适应技术在三维口腔内扫描中对不同牙弓形态的单个牙齿进行分割。材料和方法：使用20个不同年龄组和发育阶段的扫描牙弓生成20个简化的合成扫描变体。这些合成变体，以及16个自然扫描的牙弓，被用来训练深度学习模型。采用梯度反转层和暹罗网络技术进行领域自适应。训练PointNet和PointNet++模型主干，对准实域和合成域的潜在空间分布。在四个看不见的自然扫描拱门上进行验证，启用和不启用域适应，以评估3D深度神经网络是否可以在没有任何人类注释的3D模型的情况下进行训练。结果：在评估自然扫描牙弓时，PointNet和PointNet++模型在未启用域自适应和启用域自适应情况下的平均交叉超结合度分别为0.34和0.36 mIoU和0.80和0.95 mIoU。结论：领域自适应技术可以在不需要人工操作员注释训练数据的情况下，使用合成的3D颌骨扫描来训练分割深度学习模型。

{"title":"Unsupervised tooth segmentation from three dimensional scans of the dental arch using domain adaptation of synthetic data","authors":"Md Sahadul Hasan Arian , Faisal Ahmed Sifat , Saif Ahmed , Nabeel Mohammed , Taseef Hasan Farook","doi":"10.1016/j.ijmedinf.2024.105769","DOIUrl":"10.1016/j.ijmedinf.2024.105769","url":null,"abstract":"<div><h3>Background</h3><div>The automated segmentation of individual teeth from 3D models of the human dental arch is challenging due to variations in tooth alignment, arch form and overall maxillofacial anatomy. Domain adaptation is a specialised technique in deep learning which allows models to adapt to data from different domains, such as varying tooth and dental arch forms, without requiring human annotations.</div></div><div><h3>Purpose</h3><div>This study aimed to segment individual teeth from various dental arch morphologies in 3D intraoral scans using domain adaptation.</div></div><div><h3>Materials and Methods</h3><div>Twenty scanned dental arches from various age groups and developmental stages were used to generate 20 simplified synthetic variants of the scans. These synthetic variants, along with 16 natural scanned dental arches, were used to train the deep learning models. Domain adaptation was employed using Gradient Reversal Layer and Siamese Network techniques. The PointNet and PointNet++ model backbones were trained to align the latent space distribution of real and synthetic domains. Validations were performed on four unseen natural scanned arches, with and without domain adaptation enabled, to evaluate whether a 3D deep neural network can be trained without any human-annotated 3D models.</div></div><div><h3>Results</h3><div>PointNet and PointNet++ models demonstrated a mean intersection-over-union between 0.34 and 0.36 mIoU without domain adaptation enabled and 0.80 and 0.95 mIoU, respectively with domain adaptation enabled when assessing natural scanned dental arches.</div></div><div><h3>Conclusion</h3><div>Domain adaptation techniques can enable training a segmentation deep learning model using synthetically generated 3D jaw scans without requiring human operators annotating the training data<strong>.</strong></div></div>","PeriodicalId":54950,"journal":{"name":"International Journal of Medical Informatics","volume":"195 ","pages":"Article 105769"},"PeriodicalIF":3.7,"publicationDate":"2024-12-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142900507","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

OptimCLM: Optimizing clinical language models for predicting patient outcomes via knowledge distillation, pruning and quantization OptimCLM：优化临床语言模型，通过知识蒸馏，修剪和量化来预测患者结果。

IF 3.7 2区医学 Q2 COMPUTER SCIENCE, INFORMATION SYSTEMS

International Journal of Medical Informatics

Pub Date : 2024-12-18 DOI: 10.1016/j.ijmedinf.2024.105764

Mohammad Junayed Hasan , Fuad Rahman , Nabeel Mohammed

Background

Clinical Language Models (CLMs) possess the potential to reform traditional healthcare systems by aiding in clinical decision making and optimal resource utilization. They can enhance patient outcomes and help healthcare management through predictive clinical tasks. However, their real-world deployment is limited due to high computational cost at inference, in terms of both time and space complexity.

Objective

This study aims to develop and optimize an efficient framework that compresses CLMs without significant performance loss, reducing inference time and disk-space, and enabling real-world clinical applications.

Methods

We introduce OptimCLM, a framework for optimizing CLMs with ensemble learning, knowledge distillation (KD), pruning and quantization. Based on domain-knowledge and performance, we select and combine domain-adaptive CLMs DischargeBERT and COReBERT as the teacher ensemble model. We transfer the teacher's knowledge to two smaller generalist models, BERT-PKD and TinyBERT, and apply black-box KD, post-training unstructured pruning and post-training 8-bit model quantization to them. In an admission-to-discharge setting, we evaluate the framework on four clinical outcome prediction tasks (length of stay prediction, mortality prediction, diagnosis prediction and procedure prediction) using admission notes from the MIMIC-III clinical database.

Results

The OptimCLM framework achieved up to 22.88× compression ratio and 28.7× inference speedup, with less than 5% and 2% loss in macro-averaged AUROC for TinyBERT and BERT-PKD, respectively. The teacher model outperformed five state-of-the-art models on all tasks. The optimized BERT-PKD model also outperformed them in most tasks.

Conclusion

Our findings suggest that domain-specific fine-tuning with ensemble learning and KD is more effective than domain-specific pre-training for domain-knowledge transfer and text classification tasks. Thus, this work demonstrates the feasibility and potential of deploying optimized CLMs in healthcare settings and developing them with less computational resources.

背景：临床语言模型（CLMs）通过帮助临床决策和优化资源利用，具有改革传统医疗保健系统的潜力。它们可以提高患者的治疗效果，并通过预测性临床任务帮助医疗保健管理。然而，由于在时间和空间复杂性方面的推断计算成本高，它们在现实世界中的部署受到限制。目的：本研究旨在开发和优化一个有效的框架，在不显著性能损失的情况下压缩clm，减少推理时间和磁盘空间，并使现实世界的临床应用成为可能。方法：介绍了一个基于集成学习、知识蒸馏（KD）、剪枝和量化的clm优化框架OptimCLM。基于领域知识和性能，我们选择并组合了领域自适应CLMs DischargeBERT和COReBERT作为教师集成模型。我们将教师的知识转移到两个较小的通才模型BERT-PKD和TinyBERT中，并对它们应用黑箱KD、训练后非结构化修剪和训练后8位模型量化。在入院-出院设置中，我们使用来自MIMIC-III临床数据库的入院记录评估了四个临床结果预测任务（住院时间预测、死亡率预测、诊断预测和程序预测）的框架。结果：OptimCLM框架实现了22.88倍的压缩比和28.7倍的推理加速，TinyBERT和BERT-PKD的宏观平均AUROC损失分别小于5%和2%。教师模型在所有任务上都优于五个最先进的模型。优化后的BERT-PKD模型在大多数任务上也优于它们。结论：我们的研究结果表明，在领域知识转移和文本分类任务中，基于集成学习和KD的领域特定微调比领域特定预训练更有效。因此，这项工作证明了在医疗保健环境中部署优化的clm并使用更少的计算资源开发它们的可行性和潜力。

{"title":"OptimCLM: Optimizing clinical language models for predicting patient outcomes via knowledge distillation, pruning and quantization","authors":"Mohammad Junayed Hasan , Fuad Rahman , Nabeel Mohammed","doi":"10.1016/j.ijmedinf.2024.105764","DOIUrl":"10.1016/j.ijmedinf.2024.105764","url":null,"abstract":"<div><h3>Background</h3><div>Clinical Language Models (CLMs) possess the potential to reform traditional healthcare systems by aiding in clinical decision making and optimal resource utilization. They can enhance patient outcomes and help healthcare management through predictive clinical tasks. However, their real-world deployment is limited due to high computational cost at inference, in terms of both time and space complexity.</div></div><div><h3>Objective</h3><div>This study aims to develop and optimize an efficient framework that compresses CLMs without significant performance loss, reducing inference time and disk-space, and enabling real-world clinical applications.</div></div><div><h3>Methods</h3><div>We introduce OptimCLM, a framework for optimizing CLMs with ensemble learning, knowledge distillation (KD), pruning and quantization. Based on domain-knowledge and performance, we select and combine domain-adaptive CLMs DischargeBERT and COReBERT as the teacher ensemble model. We transfer the teacher's knowledge to two smaller generalist models, BERT-PKD and TinyBERT, and apply black-box KD, post-training unstructured pruning and post-training 8-bit model quantization to them. In an admission-to-discharge setting, we evaluate the framework on four clinical outcome prediction tasks (length of stay prediction, mortality prediction, diagnosis prediction and procedure prediction) using admission notes from the MIMIC-III clinical database.</div></div><div><h3>Results</h3><div>The OptimCLM framework achieved up to <strong>22.88</strong>× compression ratio and <strong>28.7</strong>× inference speedup, with less than <strong>5%</strong> and <strong>2%</strong> loss in macro-averaged AUROC for TinyBERT and BERT-PKD, respectively. The teacher model outperformed five state-of-the-art models on all tasks. The optimized BERT-PKD model also outperformed them in most tasks.</div></div><div><h3>Conclusion</h3><div>Our findings suggest that domain-specific fine-tuning with ensemble learning and KD is more effective than domain-specific pre-training for domain-knowledge transfer and text classification tasks. Thus, this work demonstrates the feasibility and potential of deploying optimized CLMs in healthcare settings and developing them with less computational resources.</div></div>","PeriodicalId":54950,"journal":{"name":"International Journal of Medical Informatics","volume":"195 ","pages":"Article 105764"},"PeriodicalIF":3.7,"publicationDate":"2024-12-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142873522","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 meta-analysis of AI and machine learning in project management: Optimizing vaccine development for emerging viral threats in biotechnology 项目管理中人工智能和机器学习的荟萃分析：优化生物技术中新出现的病毒威胁的疫苗开发。

IF 3.7 2区医学 Q2 COMPUTER SCIENCE, INFORMATION SYSTEMS

International Journal of Medical Informatics

Pub Date : 2024-12-18 DOI: 10.1016/j.ijmedinf.2024.105768

Jatin Vaghasiya , Mahim Khan , Tarak Milan Bakhda

Objectives

Artificial Intelligence (AI) and Machine Learning (ML) have emerged as transformative technologies across various industries, including healthcare, biotechnology, and vaccine development. These technologies offer immense potential to improve project management efficiency, decision-making, and resource utilization, especially in complex tasks such as vaccine development and healthcare innovations.

Methods

A systematic meta-analysis was conducted by reviewing studies from databases like PubMed, IEEE Xplore, Scopus, Web of Science, EMBASE, and Google Scholar until September 2024. The analysis focused on the application of AI and ML in project management for vaccine development, biotechnology, and broader healthcare innovations using the PICO framework to guide study selection and inclusion. Statistical analyses were performed using Review Manager 5.4 and Comprehensive Meta-Analysis (CMA) software.

Results

The meta-analysis reviewed 44 studies examining the integration of Artificial Intelligence (AI) and Machine Learning (ML) in healthcare, biotechnology, and vaccine development project management. Results demonstrated significant improvements in efficiency, resource allocation, decision-making, and risk management. AI/ML applications notably accelerated vaccine development, from candidate identification to clinical trial optimization, and improved predictive modeling for efficacy and safety. Subgroup analysis revealed variations in effectiveness across healthcare sectors, with the highest pooled effect sizes observed in infectious disease control (1.2; 95 % CI: 0.85–1.50) compared to medical imaging (0.85; 95 % CI: 0.75–0.95). Studies employing AI techniques demonstrated a pooled effect size of 0.83 (95 % CI: 0.78–1.08). Despite the observed high heterogeneity (I² = 99.04 %) and moderate-to-high risks of bias, sensitivity analyses confirmed the robustness of the findings. Overall, AI/ML integration offers transformative potential to enhance project management and vaccine development, driving innovation and efficiency in these critical fields.

Conclusion

AI and ML technologies show significant potential to transform project management practices in healthcare, biotechnology, and vaccine development by enhancing efficiency, predictive analytics, and decision-making capabilities. Their integration paves the way for more innovative, data-driven solutions that can adapt to evolving challenges in these fields.

目标：人工智能（AI）和机器学习（ML）已经成为各个行业的变革性技术，包括医疗保健、生物技术和疫苗开发。这些技术为提高项目管理效率、决策和资源利用提供了巨大的潜力，特别是在疫苗开发和卫生保健创新等复杂任务中。方法：对截至2024年9月PubMed、IEEE explore、Scopus、Web of Science、EMBASE、谷歌Scholar等数据库的研究进行系统荟萃分析。该分析侧重于使用PICO框架指导研究选择和纳入的人工智能和机器学习在疫苗开发、生物技术和更广泛的医疗保健创新项目管理中的应用。采用Review Manager 5.4和综合meta分析（CMA）软件进行统计分析。结果：荟萃分析回顾了44项研究，研究了人工智能（AI）和机器学习（ML）在医疗保健、生物技术和疫苗开发项目管理中的整合。结果显示在效率、资源分配、决策和风险管理方面有显著的改善。人工智能/机器学习应用显著加速了疫苗开发，从候选疫苗识别到临床试验优化，并改进了有效性和安全性的预测模型。亚组分析揭示了各个医疗保健部门的有效性差异，在传染病控制中观察到的综合效应量最高(1.2；95% CI: 0.85-1.50)与医学影像学(0.85；95% ci: 0.75-0.95)。采用人工智能技术的研究显示，合并效应值为0.83 （95% CI: 0.78-1.08）。尽管观察到高异质性（I2 = 99.04%）和中高偏倚风险，敏感性分析证实了研究结果的稳健性。总体而言，AI/ML集成提供了变革性潜力，可以加强项目管理和疫苗开发，推动这些关键领域的创新和效率。结论：人工智能和机器学习技术通过提高效率、预测分析和决策能力，显示出巨大的潜力，可以改变医疗保健、生物技术和疫苗开发领域的项目管理实践。它们的整合为更多创新的、数据驱动的解决方案铺平了道路，这些解决方案可以适应这些领域不断变化的挑战。

{"title":"A meta-analysis of AI and machine learning in project management: Optimizing vaccine development for emerging viral threats in biotechnology","authors":"Jatin Vaghasiya , Mahim Khan , Tarak Milan Bakhda","doi":"10.1016/j.ijmedinf.2024.105768","DOIUrl":"10.1016/j.ijmedinf.2024.105768","url":null,"abstract":"<div><h3>Objectives</h3><div>Artificial Intelligence (AI) and Machine Learning (ML) have emerged as transformative technologies across various industries, including healthcare, biotechnology, and vaccine development. These technologies offer immense potential to improve project management efficiency, decision-making, and resource utilization, especially in complex tasks such as vaccine development and healthcare innovations.</div></div><div><h3>Methods</h3><div>A systematic <em>meta</em>-analysis was conducted by reviewing studies from databases like PubMed, IEEE Xplore, Scopus, Web of Science, EMBASE, and Google Scholar until September 2024. The analysis focused on the application of AI and ML in project management for vaccine development, biotechnology, and broader healthcare innovations using the PICO framework to guide study selection and inclusion. Statistical analyses were performed using Review Manager 5.4 and Comprehensive Meta-Analysis (CMA) software.</div></div><div><h3>Results</h3><div>The <em>meta</em>-analysis reviewed 44 studies examining the integration of Artificial Intelligence (AI) and Machine Learning (ML) in healthcare, biotechnology, and vaccine development project management. Results demonstrated significant improvements in efficiency, resource allocation, decision-making, and risk management. AI/ML applications notably accelerated vaccine development, from candidate identification to clinical trial optimization, and improved predictive modeling for efficacy and safety. Subgroup analysis revealed variations in effectiveness across healthcare sectors, with the highest pooled effect sizes observed in infectious disease control (1.2; 95 % CI: 0.85–1.50) compared to medical imaging (0.85; 95 % CI: 0.75–0.95). Studies employing AI techniques demonstrated a pooled effect size of 0.83 (95 % CI: 0.78–1.08). Despite the observed high heterogeneity (I<sup>2</sup> = 99.04 %) and moderate-to-high risks of bias, sensitivity analyses confirmed the robustness of the findings. Overall, AI/ML integration offers transformative potential to enhance project management and vaccine development, driving innovation and efficiency in these critical fields.</div></div><div><h3>Conclusion</h3><div>AI and ML technologies show significant potential to transform project management practices in healthcare, biotechnology, and vaccine development by enhancing efficiency, predictive analytics, and decision-making capabilities. Their integration paves the way for more innovative, data-driven solutions that can adapt to evolving challenges in these fields.</div></div>","PeriodicalId":54950,"journal":{"name":"International Journal of Medical Informatics","volume":"195 ","pages":"Article 105768"},"PeriodicalIF":3.7,"publicationDate":"2024-12-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142873440","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

Real-time assistance in suicide prevention helplines using a deep learning-based recommender system: A randomized controlled trial 基于深度学习的推荐系统在自杀预防热线中的实时帮助：一项随机对照试验。

IF 3.7 2区医学 Q2 COMPUTER SCIENCE, INFORMATION SYSTEMS

International Journal of Medical Informatics

Pub Date : 2024-12-17 DOI: 10.1016/j.ijmedinf.2024.105760

Salim Salmi , Saskia Mérelle , Nikki van Eijk , Renske Gilissen , Rob van der Mei , Sandjai Bhulai

Objective

To evaluate the effectiveness and usability of an AI-assisted tool in providing real-time assistance to counselors during suicide prevention helpline conversations.

Methods

In this RCT, the intervention group used an AI-assisted tool, which generated suggestions based on sentence embeddings (i.e. BERT) from previous successful counseling sessions. Cosine similarity was used to present the top 5 chat situation to the counsellors. The control group did not have access to the tool (care as usual). Both groups completed a questionnaire assessing their self-efficacy at the end of each shift. Counselors' usage of the tool was evaluated by measuring frequency, duration and content of interactions.

Results

In total, 48 counselors participated in the experiment: 27 counselors in the experimental condition and 21 counselors in the control condition. Together they rated 188 shifts. No significant difference in self-efficacy was observed between the two groups (p=0.36). However, counselors that used the AI-assisted tool had marginally lower response time and used the tool more often during conversations that had a longer duration. A deeper analysis of usage showed that the tool was frequently used in inappropriate situations, e.g. after the counselor had already provided a response to the help-seeker, defeating the purpose of the information. When the tool was employed appropriately (64 conversations), it provided usable information in 53 conversations (83%). However, counselors used the tool less frequently at optimal moments, indicating their potential lack of proficiency with using AI-assisted tools during helpline conversations or initial trust issues with the system.

Conclusion

The study demonstrates benefits and pitfalls of integrating AI-assisted tools in suicide prevention for improving counselor support. Despite the lack of significant impact on self-efficacy, the support tool provided usable suggestions and the frequent use during long conversations suggests counsellors may wish to use the tool in complex or challenging interactions.

目的：评估人工智能辅助工具在自杀预防热线对话期间为咨询师提供实时帮助的有效性和可用性。方法：在本随机对照试验中，干预组使用人工智能辅助工具，该工具根据之前成功的咨询会话中的句子嵌入（即BERT）生成建议。余弦相似度用于向辅导员呈现前5名的聊天情况。对照组没有使用工具（照例小心）。在每班结束时，两组人都完成了一份评估自我效能的问卷。通过测量互动的频率、持续时间和内容来评估辅导员对该工具的使用情况。结果：共有48名辅导员参与实验，其中实验组27名，对照组21名。他们总共评估了188个班次。两组患者自我效能感差异无统计学意义（p=0.36）。然而，使用人工智能辅助工具的咨询师的响应时间略短，并且在持续时间较长的对话中更频繁地使用该工具。对使用情况的深入分析表明，该工具经常在不适当的情况下使用，例如，在咨询师已经向寻求帮助的人提供了回应之后，违背了信息的目的。当适当地使用该工具（64次对话）时，它在53次对话（83%）中提供了可用的信息。然而，咨询师在最佳时刻使用该工具的频率较低，这表明他们在帮助热线对话或与系统的初始信任问题中可能缺乏使用人工智能辅助工具的熟练程度。结论：该研究展示了将人工智能辅助工具整合到自杀预防中以改善咨询师支持的好处和缺陷。尽管对自我效能没有显著的影响，但支持工具提供了有用的建议，并且在长时间的对话中频繁使用表明咨询师可能希望在复杂或具有挑战性的互动中使用该工具。

{"title":"Real-time assistance in suicide prevention helplines using a deep learning-based recommender system: A randomized controlled trial","authors":"Salim Salmi , Saskia Mérelle , Nikki van Eijk , Renske Gilissen , Rob van der Mei , Sandjai Bhulai","doi":"10.1016/j.ijmedinf.2024.105760","DOIUrl":"10.1016/j.ijmedinf.2024.105760","url":null,"abstract":"<div><h3>Objective</h3><div>To evaluate the effectiveness and usability of an AI-assisted tool in providing real-time assistance to counselors during suicide prevention helpline conversations.</div></div><div><h3>Methods</h3><div>In this RCT, the intervention group used an AI-assisted tool, which generated suggestions based on sentence embeddings (i.e. BERT) from previous successful counseling sessions. Cosine similarity was used to present the top 5 chat situation to the counsellors. The control group did not have access to the tool (care as usual). Both groups completed a questionnaire assessing their self-efficacy at the end of each shift. Counselors' usage of the tool was evaluated by measuring frequency, duration and content of interactions.</div></div><div><h3>Results</h3><div>In total, 48 counselors participated in the experiment: 27 counselors in the experimental condition and 21 counselors in the control condition. Together they rated 188 shifts. No significant difference in self-efficacy was observed between the two groups (p=0.36). However, counselors that used the AI-assisted tool had marginally lower response time and used the tool more often during conversations that had a longer duration. A deeper analysis of usage showed that the tool was frequently used in inappropriate situations, e.g. after the counselor had already provided a response to the help-seeker, defeating the purpose of the information. When the tool was employed appropriately (64 conversations), it provided usable information in 53 conversations (83%). However, counselors used the tool less frequently at optimal moments, indicating their potential lack of proficiency with using AI-assisted tools during helpline conversations or initial trust issues with the system.</div></div><div><h3>Conclusion</h3><div>The study demonstrates benefits and pitfalls of integrating AI-assisted tools in suicide prevention for improving counselor support. Despite the lack of significant impact on self-efficacy, the support tool provided usable suggestions and the frequent use during long conversations suggests counsellors may wish to use the tool in complex or challenging interactions.</div></div>","PeriodicalId":54950,"journal":{"name":"International Journal of Medical Informatics","volume":"195 ","pages":"Article 105760"},"PeriodicalIF":3.7,"publicationDate":"2024-12-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142873487","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

Synthetic data generation in healthcare: A scoping review of reviews on domains, motivations, and future applications 医疗保健中的合成数据生成：对领域、动机和未来应用程序的审查进行范围界定。

IF 3.7 2区医学 Q2 COMPUTER SCIENCE, INFORMATION SYSTEMS

International Journal of Medical Informatics

Pub Date : 2024-12-17 DOI: 10.1016/j.ijmedinf.2024.105763

Miguel Rujas, Rodrigo Martín Gómez del Moral Herranz, Giuseppe Fico , Beatriz Merino-Barbancho

Background

The development of Artificial Intelligence in the healthcare sector is generating a great impact. However, one of the primary challenges for the implementation of this technology is the access to high-quality data due to issues in data collection and regulatory constraints, for which synthetic data is an emerging alternative. While previous research has reviewed synthetic data generation techniques, there is limited focus on their applications and the motivations driving their synthesis. A comprehensive review is needed to expand the potential of synthetic data into less explored healthcare areas.

Objective

This review aims to identify the healthcare domains where synthetic data are currently generated, the motivations behind their creation, their future uses, limitations, and types of data.

Materials and methods

Following the PRISMA-ScR framework, this review analysed literature from the last 10 years within PubMed, Scopus, and Web of Science. Reviews containing information on synthetic data generation in healthcare were screened and analysed. Key healthcare domains, motivations, future uses, and gaps in the literature were identified through a structured data extraction process.

Results

Of the 346 reviews identified, 42 were included for data extraction. Thirteen main domains were identified, with Oncology, Neurology, and Cardiology being the most frequently mentioned. Five primary motivations for synthetic data generation and three major categories of future applications were highlighted. Additionally, unstructured data, particularly images, were found to be the predominant type of synthetic data generated.

Discussion and conclusion

Synthetic data are currently being generated across diverse healthcare domains, showcasing their adaptability and potential. Despite their early stage, synthetic data technologies hold significant promise for future applications. Expanding their use into new domains and less common data types (e.g., video and text) could further enhance their impact. Future work should focus on developing evaluation benchmarks and standardized generative models tailored to specific healthcare domains.

背景：人工智能在医疗领域的发展正在产生巨大的影响。然而，实施该技术的主要挑战之一是，由于数据收集和监管限制方面的问题，无法获得高质量的数据，而合成数据是一种新兴的替代方案。虽然以前的研究已经审查了合成数据生成技术，但对其应用和驱动其合成的动机的关注有限。需要进行全面审查，以扩大合成数据在较少探索的医疗保健领域的潜力。目的：本综述旨在确定当前生成合成数据的医疗保健领域、创建合成数据背后的动机、它们的未来用途、限制和数据类型。材料和方法：遵循PRISMA-ScR框架，本综述分析了PubMed、Scopus和Web of Science中过去10年的文献。审查和分析了关于保健领域合成数据生成的信息。通过结构化数据提取过程确定了关键的医疗保健领域、动机、未来用途和文献中的差距。结果：在确定的346篇综述中，有42篇纳入了数据提取。确定了13个主要领域，其中肿瘤学，神经学和心脏病学是最常被提及的。强调了合成数据生成的五个主要动机和未来应用的三个主要类别。此外，非结构化数据，特别是图像，被发现是生成的合成数据的主要类型。讨论和结论：目前正在不同的医疗保健领域生成合成数据，展示了它们的适应性和潜力。尽管合成数据技术处于早期阶段，但它在未来的应用中具有重要的前景。将它们的使用扩大到新的领域和不太常见的数据类型（例如视频和文本）可以进一步加强它们的影响。未来的工作应侧重于开发针对特定医疗保健领域的评估基准和标准化生成模型。

{"title":"Synthetic data generation in healthcare: A scoping review of reviews on domains, motivations, and future applications","authors":"Miguel Rujas, Rodrigo Martín Gómez del Moral Herranz, Giuseppe Fico , Beatriz Merino-Barbancho","doi":"10.1016/j.ijmedinf.2024.105763","DOIUrl":"10.1016/j.ijmedinf.2024.105763","url":null,"abstract":"<div><h3>Background</h3><div>The development of Artificial Intelligence in the healthcare sector is generating a great impact. However, one of the primary challenges for the implementation of this technology is the access to high-quality data due to issues in data collection and regulatory constraints, for which synthetic data is an emerging alternative. While previous research has reviewed synthetic data generation techniques, there is limited focus on their applications and the motivations driving their synthesis. A comprehensive review is needed to expand the potential of synthetic data into less explored healthcare areas.</div></div><div><h3>Objective</h3><div>This review aims to identify the healthcare domains where synthetic data are currently generated, the motivations behind their creation, their future uses, limitations, and types of data.</div></div><div><h3>Materials and methods</h3><div>Following the PRISMA-ScR framework, this review analysed literature from the last 10 years within PubMed, Scopus, and Web of Science. Reviews containing information on synthetic data generation in healthcare were screened and analysed. Key healthcare domains, motivations, future uses, and gaps in the literature were identified through a structured data extraction process.</div></div><div><h3>Results</h3><div>Of the 346 reviews identified, 42 were included for data extraction. Thirteen main domains were identified, with Oncology, Neurology, and Cardiology being the most frequently mentioned. Five primary motivations for synthetic data generation and three major categories of future applications were highlighted. Additionally, unstructured data, particularly images, were found to be the predominant type of synthetic data generated.</div></div><div><h3>Discussion and conclusion</h3><div>Synthetic data are currently being generated across diverse healthcare domains, showcasing their adaptability and potential. Despite their early stage, synthetic data technologies hold significant promise for future applications. Expanding their use into new domains and less common data types (e.g., video and text) could further enhance their impact. Future work should focus on developing evaluation benchmarks and standardized generative models tailored to specific healthcare domains.</div></div>","PeriodicalId":54950,"journal":{"name":"International Journal of Medical Informatics","volume":"195 ","pages":"Article 105763"},"PeriodicalIF":3.7,"publicationDate":"2024-12-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142886492","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

Unmasking the chameleons: A benchmark for out-of-distribution detection in medical tabular data 揭露变色龙：医疗表格数据中分布外检测的基准。

IF 3.7 2区医学 Q2 COMPUTER SCIENCE, INFORMATION SYSTEMS

International Journal of Medical Informatics

Pub Date : 2024-12-17 DOI: 10.1016/j.ijmedinf.2024.105762

Mohammad Azizmalayeri , Ameen Abu-Hanna , Giovanni Cinà

Background

Machine Learning (ML) models often struggle to generalize effectively to data that deviates from the training distribution. This raises significant concerns about the reliability of real-world healthcare systems encountering such inputs known as out-of-distribution (OOD) data. These concerns can be addressed by real-time detection of OOD inputs. While numerous OOD detection approaches have been suggested in other fields - especially in computer vision - it remains unclear whether similar methods effectively address challenges posed by medical tabular data.

Objective

To answer this important question, we propose an extensive reproducible benchmark to compare different OOD detection methods in medical tabular data across a comprehensive suite of tests.

Method

To achieve this, we leverage 4 different and large public medical datasets, including eICU and MIMIC-IV, and consider various kinds of OOD cases within these datasets. For example, we examine OODs originating from a statistically different dataset than the training set according to the membership model introduced by Debray et al. [1], as well as OODs obtained by splitting a given dataset based on a value of a distinguishing variable. To identify OOD instances, we explore a range of 10 density-based methods that learn the marginal distribution of the data, alongside 17 post-hoc detectors that are applied on top of prediction models already trained on the data. The prediction models involve three distinct architectures, namely MLP, ResNet, and Transformer.

Main results

In our experiments, when the membership model achieved an AUC of 0.98, which indicated a clear distinction between OOD data and the training set, we observed that the OOD detection methods had achieved AUC values exceeding 0.95 in distinguishing OOD data. In contrast, in the experiments with subtler changes in data distribution such as selecting OOD data based on ethnicity and age characteristics, many OOD detection methods performed similarly to a random classifier with AUC values close to 0.5. This may suggest a correlation between separability, as indicated by the membership model, and OOD detection performance, as indicated by the AUC of the detection model. This warrants future research.

背景：机器学习（ML）模型通常难以有效地泛化到偏离训练分布的数据。这引起了人们对现实世界医疗保健系统遇到这种被称为分布外（OOD）数据输入的可靠性的重大关注。这些问题可以通过实时检测OOD输入来解决。虽然在其他领域已经提出了许多OOD检测方法，特别是在计算机视觉领域，但尚不清楚类似的方法是否能有效地解决医疗表格数据带来的挑战。目的：为了回答这个重要的问题，我们提出了一个广泛的可重复的基准，以比较不同的医学表格数据中的OOD检测方法。方法：为了实现这一目标，我们利用了4个不同的大型公共医疗数据集，包括eICU和MIMIC-IV，并考虑了这些数据集中的各种OOD病例。例如，我们根据Debray et al.[1]引入的隶属度模型，检查来自统计上不同于训练集的数据集的ood，以及根据区分变量的值分割给定数据集获得的ood。为了识别OOD实例，我们探索了10种基于密度的方法来学习数据的边际分布，以及17种应用于已经在数据上训练过的预测模型之上的事后检测器。预测模型涉及三种不同的体系结构，即MLP、ResNet和Transformer。主要结果：在我们的实验中，当隶属度模型的AUC达到0.98，表明OOD数据与训练集之间有明显的区别时，我们观察到OOD检测方法在区分OOD数据方面的AUC值已经超过0.95。相比之下，在数据分布有细微变化的实验中，如基于种族和年龄特征选择OOD数据，许多OOD检测方法的表现与AUC值接近0.5的随机分类器相似。这可能表明可分离性（如成员模型所示）与OOD检测性能（如检测模型的AUC所示）之间存在相关性。这值得未来的研究。

{"title":"Unmasking the chameleons: A benchmark for out-of-distribution detection in medical tabular data","authors":"Mohammad Azizmalayeri , Ameen Abu-Hanna , Giovanni Cinà","doi":"10.1016/j.ijmedinf.2024.105762","DOIUrl":"10.1016/j.ijmedinf.2024.105762","url":null,"abstract":"<div><h3>Background</h3><div>Machine Learning (ML) models often struggle to generalize effectively to data that deviates from the training distribution. This raises significant concerns about the reliability of real-world healthcare systems encountering such inputs known as out-of-distribution (OOD) data. These concerns can be addressed by real-time detection of OOD inputs. While numerous OOD detection approaches have been suggested in other fields - especially in computer vision - it remains unclear whether similar methods effectively address challenges posed by medical tabular data.</div></div><div><h3>Objective</h3><div>To answer this important question, we propose an extensive reproducible benchmark to compare different OOD detection methods in medical tabular data across a comprehensive suite of tests.</div></div><div><h3>Method</h3><div>To achieve this, we leverage 4 different and large public medical datasets, including eICU and MIMIC-IV, and consider various kinds of OOD cases within these datasets. For example, we examine OODs originating from a statistically different dataset than the training set according to the membership model introduced by Debray et al. <span><span>[1]</span></span>, as well as OODs obtained by splitting a given dataset based on a value of a distinguishing variable. To identify OOD instances, we explore a range of 10 density-based methods that learn the marginal distribution of the data, alongside 17 post-hoc detectors that are applied on top of prediction models already trained on the data. The prediction models involve three distinct architectures, namely MLP, ResNet, and Transformer.</div></div><div><h3>Main results</h3><div>In our experiments, when the membership model achieved an AUC of 0.98, which indicated a clear distinction between OOD data and the training set, we observed that the OOD detection methods had achieved AUC values exceeding 0.95 in distinguishing OOD data. In contrast, in the experiments with subtler changes in data distribution such as selecting OOD data based on ethnicity and age characteristics, many OOD detection methods performed similarly to a random classifier with AUC values close to 0.5. This may suggest a correlation between separability, as indicated by the membership model, and OOD detection performance, as indicated by the AUC of the detection model. This warrants future research.</div></div>","PeriodicalId":54950,"journal":{"name":"International Journal of Medical Informatics","volume":"195 ","pages":"Article 105762"},"PeriodicalIF":3.7,"publicationDate":"2024-12-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142873437","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

The determinants of help-seeking behaviors among cancer patients in online health communities: Evidence from China 在线健康社区中癌症患者寻求帮助行为的决定因素：来自中国的证据

IF 3.7 2区医学 Q2 COMPUTER SCIENCE, INFORMATION SYSTEMS

International Journal of Medical Informatics

Pub Date : 2024-12-16 DOI: 10.1016/j.ijmedinf.2024.105767

Xiandong Feng , Yinhuan Hu , Holger Pfaff , Sha Liu , Hui Wang , Zhen Qi

Objective

Although online health communities offer a new approach to patient interaction, the help-seeking behaviors of cancer patients within these platforms remain unexplored. This study aims to identify the determinants influencing online help-seeking behaviors among cancer patients.

Method

Based on motivation theory, we proposed six hypotheses and developed a research model. Data were collected from 1100 cancer patients who sought help in a leading Chinese online cancer community in March, June, and September 2023. We used the fixed-effect negative binomial model to test research hypotheses.

Results

The findings indicated that the time since diagnosis (β = -0.127, P < 0.001) was negatively associated with online help-seeking behaviors among cancer patients. In contrast, social support (β = 0.002, P = 0.003) and disease stigma (β = 0.170, P < 0.001) positively influenced their help-seeking behaviors in online health communities. Furthermore, while male and female cancer patients showed decreased help-seeking behaviors as time since diagnosis increased, the decline was less pronounced for females (β = 0.040, P < 0.001). The positive impact of disease stigma on help-seeking behaviors is stronger for female patients than male patients (β = 0.098, P < 0.001).

Conclusion

This research broadens the understanding of how cancer patients seek help in digital environments and enhances theoretical insights into these behaviors.

目的：尽管在线健康社区为患者互动提供了一种新的途径，但癌症患者在这些平台上的求助行为仍未被探索。本研究旨在找出影响癌症患者网路求助行为的决定因素。方法：基于动机理论，提出六个假设，建立研究模型。数据收集自1100名癌症患者，这些患者于2023年3月、6月和9月在中国一个领先的在线癌症社区寻求帮助。我们使用固定效应负二项模型来检验研究假设。结论：本研究拓宽了对癌症患者如何在数字环境中寻求帮助的理解，并增强了对这些行为的理论见解。

{"title":"The determinants of help-seeking behaviors among cancer patients in online health communities: Evidence from China","authors":"Xiandong Feng , Yinhuan Hu , Holger Pfaff , Sha Liu , Hui Wang , Zhen Qi","doi":"10.1016/j.ijmedinf.2024.105767","DOIUrl":"10.1016/j.ijmedinf.2024.105767","url":null,"abstract":"<div><h3>Objective</h3><div>Although online health communities offer a new approach to patient interaction, the help-seeking behaviors of cancer patients within these platforms remain unexplored. This study aims to identify the determinants influencing online help-seeking behaviors among cancer patients.</div></div><div><h3>Method</h3><div>Based on motivation theory, we proposed six hypotheses and developed a research model. Data were collected from 1100 cancer patients who sought help in a leading Chinese online cancer community in March, June, and September 2023. We used the fixed-effect negative binomial model to test research hypotheses.</div></div><div><h3>Results</h3><div>The findings indicated that the time since diagnosis (β = -0.127, P < 0.001) was negatively associated with online help-seeking behaviors among cancer patients. In contrast, social support (β = 0.002, P = 0.003) and disease stigma (β = 0.170, P < 0.001) positively influenced their help-seeking behaviors in online health communities. Furthermore, while male and female cancer patients showed decreased help-seeking behaviors as time since diagnosis increased, the decline was less pronounced for females (β = 0.040, P < 0.001). The positive impact of disease stigma on help-seeking behaviors is stronger for female patients than male patients (β = 0.098, P < 0.001).</div></div><div><h3>Conclusion</h3><div>This research broadens the understanding of how cancer patients seek help in digital environments and enhances theoretical insights into these behaviors.</div></div>","PeriodicalId":54950,"journal":{"name":"International Journal of Medical Informatics","volume":"195 ","pages":"Article 105767"},"PeriodicalIF":3.7,"publicationDate":"2024-12-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142900127","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

Training machine learning models to detect rare inborn errors of metabolism (IEMs) based on GC–MS urinary metabolomics for diseases screening 训练机器学习模型以检测基于GC-MS尿液代谢组学的罕见先天性代谢错误（IEMs），用于疾病筛查。

IF 3.7 2区医学 Q2 COMPUTER SCIENCE, INFORMATION SYSTEMS

International Journal of Medical Informatics

Pub Date : 2024-12-16 DOI: 10.1016/j.ijmedinf.2024.105765

Haomin Li , Siyuan Gao , Dan Wu , Min Zhu , Zhenzhen Hu , Kexin Fang , Xiuru Chen , Zhou Ni , Jing Li , Beibei Zhao , Xuhui She , Xinwen Huang

Background

Gas chromatography-mass spectrometry (GC–MS) has been shown to be a potentially efficient metabolic profiling platform in urine analysis. However, the widespread use of GC–MS for inborn errors of metabolism (IEM) screening is constrained by the rarity of IEM in population, and the difficult and specialized complexity of the interpretation of GC–MS organic acid profiles.

Methods

Based on 355,197 GC–MS test cases accumulated from 2013 to 2021 in China, a random forest-based machine learning model was proposed, trained, and evaluated. Weighted undersampling or oversampling data processing and staged modeling strategies were used to handle the highly imbalanced data and improve the ability of the model to identify different types of rare IEM cases.

Result

In the first-stage model, which only identified positive cases without discriminating the specific IEM, the screening sensitivity was 0.938 (or 0.991 if abnormal cases were also included). The average sensitivity of the second-stage models that classify 11 particular IEMs is 0.992, with an average specificity and accuracy of 0.944 and 0.969, respectively. The SHAP values visualized for each model explain the basis for the differential diagnosis made by the model.

Conclusion

With sufficient high-quality data, machine learning models can provide high-sensitivity GC–MS interpretation and greatly improve the efficiency and quality of GC–MS based IEM screening.

背景：气相色谱-质谱（GC-MS）已被证明是尿液分析中潜在有效的代谢分析平台。然而，由于IEM在人群中的罕见性，以及GC-MS有机酸谱分析的难度和专业复杂性，GC-MS在先天性代谢错误（IEM）筛查中的广泛应用受到了限制。方法：基于2013年至2021年在中国积累的355197个GC-MS测试用例，提出、训练和评估一个基于随机森林的机器学习模型。采用加权欠采样或过采样数据处理和分阶段建模策略来处理高度不平衡的数据，提高模型识别不同类型罕见IEM案例的能力。结果：在第一阶段模型中，仅识别阳性病例，不区分特异性IEM，筛选敏感性为0.938（如果同时包括异常病例，则为0.991）。第二阶段模型对11种特定IEMs的平均灵敏度为0.992，平均特异性和准确率分别为0.944和0.969。每个模型可视化的SHAP值解释了该模型所作鉴别诊断的基础。结论：有了足够的高质量数据，机器学习模型可以提供高灵敏度的GC-MS解释，大大提高了基于GC-MS的IEM筛选的效率和质量。

{"title":"Training machine learning models to detect rare inborn errors of metabolism (IEMs) based on GC–MS urinary metabolomics for diseases screening","authors":"Haomin Li , Siyuan Gao , Dan Wu , Min Zhu , Zhenzhen Hu , Kexin Fang , Xiuru Chen , Zhou Ni , Jing Li , Beibei Zhao , Xuhui She , Xinwen Huang","doi":"10.1016/j.ijmedinf.2024.105765","DOIUrl":"10.1016/j.ijmedinf.2024.105765","url":null,"abstract":"<div><h3>Background</h3><div>Gas chromatography-mass spectrometry (GC–MS) has been shown to be a potentially efficient metabolic profiling platform in urine analysis. However, the widespread use of GC–MS for inborn errors of metabolism (IEM) screening is constrained by the rarity of IEM in population, and the difficult and specialized complexity of the interpretation of GC–MS organic acid profiles.</div></div><div><h3>Methods</h3><div>Based on 355,197 GC–MS test cases accumulated from 2013 to 2021 in China, a random forest-based machine learning model was proposed, trained, and evaluated. Weighted undersampling or oversampling data processing and staged modeling strategies were used to handle the highly imbalanced data and improve the ability of the model to identify different types of rare IEM cases.</div></div><div><h3>Result</h3><div>In the first-stage model, which only identified positive cases without discriminating the specific IEM, the screening sensitivity was 0.938 (or 0.991 if abnormal cases were also included). The average sensitivity of the second-stage models that classify 11 particular IEMs is 0.992, with an average specificity and accuracy of 0.944 and 0.969, respectively. The SHAP values visualized for each model explain the basis for the differential diagnosis made by the model.</div></div><div><h3>Conclusion</h3><div>With sufficient high-quality data, machine learning models can provide high-sensitivity GC–MS interpretation and greatly improve the efficiency and quality of GC–MS based IEM screening.</div></div>","PeriodicalId":54950,"journal":{"name":"International Journal of Medical Informatics","volume":"195 ","pages":"Article 105765"},"PeriodicalIF":3.7,"publicationDate":"2024-12-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142873433","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

Persuasive strategies in digital interventions to combat internet addiction: A systematic review 说服策略在数字干预对抗网络成瘾：一个系统的回顾。

IF 3.7 2区医学 Q2 COMPUTER SCIENCE, INFORMATION SYSTEMS

International Journal of Medical Informatics

Pub Date : 2024-12-16 DOI: 10.1016/j.ijmedinf.2024.105725

Yansen Theopilus , Abdullah Al Mahmud , Hilary Davis , Johanna Renny Octavia

Background

The internet provides valuable benefits in supporting our lives. However, concerns arise regarding internet addiction, a behavioural disorder due to excessive and uncontrolled internet use that has harmful effects on human health and wellbeing. Studies highlighted the potential of digital behaviour change interventions to address health behaviour problems. However, little is known about how and to what extent persuasive strategies can be utilised in digital interventions to combat internet addiction. Accordingly, this systematic review aims to investigate the design and implementation of persuasive strategies in digital interventions to combat internet addiction, including their contexts, features, and outcomes.

Methods

We searched for peer-reviewed articles from four databases (Scopus, Web of Science, ACM, and PubMed). The Persuasive Systems Design (PSD) model and Behaviour Change Technique (BCT) taxonomy were used to identify persuasive strategies. We included 14 primary studies discussing digital interventions to address the problem and their outcomes.

Results

Four persuasion contexts were identified, including 1) self-management systems to reduce internet use, 2) analytics systems to examine use patterns and provide behavioural suggestions, 3) parental control systems to manage children’s internet use, and 4) unattractive settings to discourage internet use. The promising interventions used the following persuasion route: help the user determine behaviour goals, facilitate actions to accomplish behaviour goals, and reinforce the user to perform behaviour goals. Potential persuasive strategies were also identified, including goal-setting, action planning, task reduction, tunnelling how to perform a behaviour, tailored and personalised suggestions/prompts, reminders, trustworthiness, anticipated regret, and social support strategies.

Conclusion

Our findings shed light on the promising persuasive contexts and strategies to combat internet addiction using digital interventions. We suggest future research and practices to utilise our findings to develop effective digital interventions, especially for combatting internet addiction in vulnerable populations like children or people from developing regions.

背景：互联网为我们的生活提供了宝贵的好处。然而，人们对网络成瘾感到担忧，这是一种由于过度和不受控制地使用互联网而导致的行为障碍，对人类健康和福祉产生有害影响。研究强调了数字行为改变干预措施在解决健康行为问题方面的潜力。然而，关于说服策略如何以及在多大程度上可以用于数字干预来对抗网络成瘾，人们知之甚少。因此，本系统综述旨在研究说服策略在对抗网络成瘾的数字干预中的设计和实施，包括其背景、特征和结果。方法：我们从四个数据库（Scopus、Web of Science、ACM和PubMed）中检索同行评议的文章。说服系统设计（PSD）模型和行为改变技术（BCT）分类法被用于确定说服策略。我们纳入了14项讨论数字干预来解决问题及其结果的初步研究。结果：确定了四种说服环境，包括1)减少互联网使用的自我管理系统，2)检查使用模式并提供行为建议的分析系统，3)管理儿童互联网使用的家长控制系统，以及4)不吸引人的设置以阻止互联网使用。有希望的干预使用以下说服路线：帮助用户确定行为目标，促进行动以实现行为目标，并加强用户执行行为目标。潜在的说服策略也被确定，包括目标设定、行动计划、任务减少、如何执行行为、量身定制和个性化的建议/提示、提醒、可信度、预期后悔和社会支持策略。结论：我们的研究结果揭示了利用数字干预来对抗网络成瘾的有前途的有说服力的背景和策略。我们建议未来的研究和实践利用我们的发现来开发有效的数字干预措施，特别是在儿童或发展中地区的弱势群体中对抗网络成瘾。

{"title":"Persuasive strategies in digital interventions to combat internet addiction: A systematic review","authors":"Yansen Theopilus , Abdullah Al Mahmud , Hilary Davis , Johanna Renny Octavia","doi":"10.1016/j.ijmedinf.2024.105725","DOIUrl":"10.1016/j.ijmedinf.2024.105725","url":null,"abstract":"<div><h3>Background</h3><div>The internet provides valuable benefits in supporting our lives. However, concerns arise regarding internet addiction, a behavioural disorder due to excessive and uncontrolled internet use that has harmful effects on human health and wellbeing. Studies highlighted the potential of digital behaviour change interventions to address health behaviour problems. However, little is known about how and to what extent persuasive strategies can be utilised in digital interventions to combat internet addiction. Accordingly, this systematic review aims to investigate the design and implementation of persuasive strategies in digital interventions to combat internet addiction, including their contexts, features, and outcomes.</div></div><div><h3>Methods</h3><div>We searched for peer-reviewed articles from four databases (Scopus, Web of Science, ACM, and PubMed). The Persuasive Systems Design (PSD) model and Behaviour Change Technique (BCT) taxonomy were used to identify persuasive strategies. We included 14 primary studies discussing digital interventions to address the problem and their outcomes.</div></div><div><h3>Results</h3><div>Four persuasion contexts were identified, including 1) self-management systems to reduce internet use, 2) analytics systems to examine use patterns and provide behavioural suggestions, 3) parental control systems to manage children’s internet use, and 4) unattractive settings to discourage internet use. The promising interventions used the following persuasion route: help the user determine behaviour goals, facilitate actions to accomplish behaviour goals, and reinforce the user to perform behaviour goals. Potential persuasive strategies were also identified, including goal-setting, action planning, task reduction, tunnelling how to perform a behaviour, tailored and personalised suggestions/prompts, reminders, trustworthiness, anticipated regret, and social support strategies.</div></div><div><h3>Conclusion</h3><div>Our findings shed light on the promising persuasive contexts and strategies to combat internet addiction using digital interventions. We suggest future research and practices to utilise our findings to develop effective digital interventions, especially for combatting internet addiction in vulnerable populations like children or people from developing regions.</div></div>","PeriodicalId":54950,"journal":{"name":"International Journal of Medical Informatics","volume":"195 ","pages":"Article 105725"},"PeriodicalIF":3.7,"publicationDate":"2024-12-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142873482","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

Active learning for extracting rare adverse events from electronic health records: A study in pediatric cardiology 从电子健康记录中提取罕见不良事件的主动学习：一项儿科心脏病学研究。

IF 3.7 2区医学 Q2 COMPUTER SCIENCE, INFORMATION SYSTEMS

International Journal of Medical Informatics

Pub Date : 2024-12-12 DOI: 10.1016/j.ijmedinf.2024.105761

Sophie Quennelle , Sophie Malekzadeh-Milani , Nicolas Garcelon , Hassan Faour , Anita Burgun , Carole Faviez , Rosy Tsopra , Damien Bonnet , Antoine Neuraz

Objective

Automate the extraction of adverse events from the text of electronic medical records of patients hospitalized for cardiac catheterization.

Methods

We focused on events related to cardiac catheterization as defined by the NCDR-IMPACT registry. These events were extracted from the Necker Children’s Hospital data warehouse. Electronic health records were pre-screened using regular expressions. The resulting datasets contained numerous false positives sentences that were annotated by a cardiologist using an active learning process. A deep learning text classifier was then trained on this active learning-annotated dataset to accurately identify patients who have suffered a serious adverse event.

Results

The dataset included 2,980 patients. Regular expression based extraction of adverse events related to cardiac catheterization achieved a perfect recall. Due to the rarity of adverse events, the dataset obtained from this initial pre-screening step was imbalanced, containing a significant number of false positives. The active learning annotation enabled the acquisition of a representative dataset suitable for training a deep learning model. The deep learning text-classifier identified patients who underwent adverse events after cardiac catheterization with a recall of 0.78 and a specificity of 0.94.

Conclusion

Our model effectively identified patients who experienced adverse events related to cardiac catheterization using real clinical data. Enabled by an active learning annotation process, it shows promise for large language model applications in clinical research, especially for rare diseases with limited annotated databases. Our model’s strength lies in its development by physicians for physicians, ensuring its relevance and applicability in clinical practice.

目的：从心导管住院患者的电子病历文本中自动提取不良事件。方法：我们关注ndr - impact注册表中定义的与心导管术相关的事件。这些事件是从Necker儿童医院的数据仓库中提取的。使用正则表达式对电子健康记录进行预筛选。结果数据集包含许多假阳性句子，由心脏病专家使用主动学习过程注释。然后在这个主动学习注释数据集上训练深度学习文本分类器，以准确识别遭受严重不良事件的患者。结果：数据集包括2980例患者。基于正则表达式的心导管相关不良事件提取达到了完美的召回率。由于不良事件的罕见性，从最初的预筛选步骤获得的数据集是不平衡的，包含大量的假阳性。主动学习标注使得获取适合训练深度学习模型的代表性数据集成为可能。深度学习文本分类器识别出心导管插入术后发生不良事件的患者，召回率为0.78，特异性为0.94。结论：我们的模型使用真实的临床数据有效地识别了经历心导管相关不良事件的患者。通过主动学习注释过程，它显示了在临床研究中大型语言模型应用的前景，特别是对于具有有限注释数据库的罕见疾病。我们的模型的优势在于它是由医生为医生开发的，保证了它在临床实践中的相关性和适用性。

{"title":"Active learning for extracting rare adverse events from electronic health records: A study in pediatric cardiology","authors":"Sophie Quennelle , Sophie Malekzadeh-Milani , Nicolas Garcelon , Hassan Faour , Anita Burgun , Carole Faviez , Rosy Tsopra , Damien Bonnet , Antoine Neuraz","doi":"10.1016/j.ijmedinf.2024.105761","DOIUrl":"10.1016/j.ijmedinf.2024.105761","url":null,"abstract":"<div><h3>Objective</h3><div>Automate the extraction of adverse events from the text of electronic medical records of patients hospitalized for cardiac catheterization.</div></div><div><h3>Methods</h3><div>We focused on events related to cardiac catheterization as defined by the NCDR-IMPACT registry. These events were extracted from the Necker Children’s Hospital data warehouse. Electronic health records were pre-screened using regular expressions. The resulting datasets contained numerous false positives sentences that were annotated by a cardiologist using an active learning process. A deep learning text classifier was then trained on this active learning-annotated dataset to accurately identify patients who have suffered a serious adverse event.</div></div><div><h3>Results</h3><div>The dataset included 2,980 patients. Regular expression based extraction of adverse events related to cardiac catheterization achieved a perfect recall. Due to the rarity of adverse events, the dataset obtained from this initial pre-screening step was imbalanced, containing a significant number of false positives. The active learning annotation enabled the acquisition of a representative dataset suitable for training a deep learning model. The deep learning text-classifier identified patients who underwent adverse events after cardiac catheterization with a recall of 0.78 and a specificity of 0.94.</div></div><div><h3>Conclusion</h3><div>Our model effectively identified patients who experienced adverse events related to cardiac catheterization using real clinical data. Enabled by an active learning annotation process, it shows promise for large language model applications in clinical research, especially for rare diseases with limited annotated databases. Our model’s strength lies in its development by physicians for physicians, ensuring its relevance and applicability in clinical practice.</div></div>","PeriodicalId":54950,"journal":{"name":"International Journal of Medical Informatics","volume":"195 ","pages":"Article 105761"},"PeriodicalIF":3.7,"publicationDate":"2024-12-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142848434","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