Machine-Learning-Enabled Obesity Level Prediction Through Electronic Health Records

Obesity is a critical health condition that severely affects an individual’s quality of life andwell-being.The occurrence of obesity is strongly associated with extreme health conditions,such as cardiac diseases,diabetes,hypertension,and some types of cancer.Therefore,it is vital to avoid obesity and or reverse its occurrence.Incorporating healthy food habits and an active lifestyle can help to prevent obesity.In this regard,artificial intelligence(AI)can play an important role in estimating health conditions and detecting obesity and its types.This study aims to see obesity levels in adults by implementing AIenabled machine learning on a real-life dataset.This dataset is in the form of electronic health records(EHR)containing data on several aspects of daily living,such as dietary habits,physical conditions,and lifestyle variables for various participants with different health conditions(underweight,normal,overweight,and obesity type I,II and III),expressed in terms of a variety of features or parameters,such as physical condition,food intake,lifestyle and mode of transportation.Three classifiers,i.e.,eXtreme gradient boosting classifier(XGB),support vector machine(SVM),and artificial neural network(ANN),are implemented to detect the status of several conditions,including obesity types.The findings indicate that the proposed XGB-based system outperforms the existing obesity level estimation methods,achieving overall performance rates of 98.5%and 99.6%in the scenarios explored.

Medical Knowledge Extraction and Analysis from Electronic Medical Records Using Deep Learning

Objectives Medical knowledge extraction (MKE) plays a key role in natural language processing (NLP) research in electronic medical records (EMR),which are the important digital carriers for recording medical activities of patients.Named entity recognition (NER) and medical relation extraction (MRE) are two basic tasks of MKE.This study aims to improve the recognition accuracy of these two tasks by exploring deep learning methods.Methods This study discussed and built two application scenes of bidirectional long short-term memory combined conditional random field (BiLSTM-CRF) model for NER and MRE tasks.In the data preprocessing of both tasks,a GloVe word embedding model was used to vectorize words.In the NER task,a sequence labeling strategy was used to classify each word tag by the joint probability distribution through the CRF layer.In the MRE task,the medical entity relation category was predicted by transforming the classification problem of a single entity into a sequence classification problem and linking the feature combinations between entities also through the CRF layer.Results Through the validation on the I2B2 2010 public dataset,the BiLSTM-CRF models built in this study got much better results than the baseline methods in the two tasks,where the F1-measure was up to 0.88 in NER task and 0.78 in MRE task.Moreover,the model converged faster and avoided problems such as overfitting.Conclusion This study proved the good performance of deep learning on medical knowledge extraction.It also verified the feasibility of the BiLSTM-CRF model in different application scenarios,laying the foundation for the subsequent work in the EMR field.

Deep Learning with Heterogeneous Graph Embeddings for Mortality Prediction from Electronic Health Records

Computational prediction of in-hospital mortality in the setting of an intensive care unit can help clinical practitioners to guide care and make early decisions for interventions. As clinical data are complex and varied in their structure and components, continued innovation of modelling strategies is required to identify architectures that can best model outcomes. In this work, we trained a Heterogeneous Graph Model(HGM) on electronic health record(EHR) data and used the resulting embedding vector as additional information added to a Convolutional Neural Network(CNN) model for predicting in-hospital mortality. We show that the additional information provided by including time as a vector in the embedding captured the relationships between medical concepts, lab tests, and diagnoses, which enhanced predictive performance. We found that adding HGM to a CNN model increased the mortality prediction accuracy up to 4%. This framework served as a foundation for future experiments involving different EHR data types on important healthcare prediction tasks.

Applying Deep Learning to Individual and Community Health Monitoring Data:A Survey

In the recent years, deep learning models have addressed many problems in various fields. Meanwhile, technology development has spawned the big data in healthcare rapidly. Nowadays, application of deep learning to solve the problems in healthcare is a hot research direction. This paper introduces the application of deep learning in healthcare extensively. We focus on 7 application areas of deep learning, which are electronic health records （EHR）, electrocardiography （ECG）, electroencephalogram （EEG）, community healthcare, data from wearable devices, drug analysis and genomics analysis. The scope of this paper does not cover medical image processing since other researchers have already substantially reviewed it. In addition, we analyze the merits and drawbacks of the existing works, analyze the existing challenges, and discuss future trends.

Using AI and Precision Nutrition to Support Brain Health during Aging

Artificial intelligence, often referred to as AI, is a branch of computer science focused on developing systems that exhibit intelligent behavior. Broadly speaking, AI researchers aim to develop technologies that can think and act in a way that mimics human cognition and decision-making [1]. The foundations of AI can be traced back to early philosophical inquiries into the nature of intelligence and thinking. However, AI is generally considered to have emerged as a formal field of study in the 1940s and 1950s. Pioneering computer scientists at the time theorized that it might be possible to extend basic computer programming concepts using logic and reasoning to develop machines capable of “thinking” like humans. Over time, the definition and goals of AI have evolved. Some theorists argued for a narrower focus on developing computing systems able to efficiently solve problems, while others aimed for a closer replication of human intelligence. Today, AI encompasses a diverse set of techniques used to enable intelligent behavior in machines. Core disciplines that contribute to modern AI research include computer science, mathematics, statistics, linguistics, psychology and cognitive science, and neuroscience. Significant AI approaches used today involve statistical classification models, machine learning, and natural language processing. Classification methods are widely applicable to problems in various domains like healthcare, such as informing diagnostic or treatment decisions based on patterns in data. Dean and Goldreich, 1998, define ML as an approach through which a computer has to learn a model by itself from the data provided but no specification on the sort of model is provided to the computer. They can then predict values for things that are different from the values used in training the models. NLP looks at two interrelated concerns, the task of training computers to understand human languages and the fact that since natural languages are so complex, they lend themselves very well to serving a number of very useful goals when used by computers.

Clinical Big Data and Deep Learning:Applications,Challenges,and Future Outlooks

The explosion of digital healthcare data has led to a surge of data-driven medical research based on machine learning. In recent years, as a powerful technique for big data, deep learning has gained a central position in machine learning circles for its great advantages in feature representation and pattern recognition. This article presents a comprehensive overview of studies that employ deep learning methods to deal with clinical data. Firstly,based on the analysis of the characteristics of clinical data, various types of clinical data(e.g., medical images,clinical notes, lab results, vital signs, and demographic informatics) are discussed and details provided of some public clinical datasets. Secondly, a brief review of common deep learning models and their characteristics is conducted. Then, considering the wide range of clinical research and the diversity of data types, several deep learning applications for clinical data are illustrated: auxiliary diagnosis, prognosis, early warning, and other tasks.Although there are challenges involved in applying deep learning techniques to clinical data, it is still worthwhile to look forward to a promising future for deep learning applications in clinical big data in the direction of precision medicine.

基于字形特征的血管外科命名实体识别

电子病历(EMR)作为医疗信息化建设的核心,蕴含着众多有价值的医疗实体,对电子病历进行命名实体识别有助于推进医学研究。为解决血管外科电子病历研究数据匮乏、实体复杂识别困难等问题,基于某三甲医院血管外科的真实临床数据,构建一个小规模的专科数据集作为实验数据集,并提出一种基于字形特征的命名实体识别模型。首先,采用掩码校正的来自Transformer的双向编码器表示(MacBERT)生成动态字向量,引入汉字四角码与汉字五笔两个维度的字形信息;然后,将文本表示传入双向门控循环单元(BiGRU)与门控空洞卷积神经网络(DGCNN)进行特征提取,并对输出结果进行拼接;最后,通过多头自注意力机制捕捉序列内部元素间的关系,利用条件随机场(CRF)进行标签解码。实验结果表明,所提模型在自建血管外科数据集上的精确率、召回率、F1值分别为96.45%、97.77%、97.10%,均优于对比模型,具有更好的实体识别性能。

中文电子病历信息提取方法研究综述

电子病历里承载的大量医疗信息能够帮助医生更好地了解患者的情况,辅助医生进行临床诊断。作为中文电子病历信息提取的2大核心任务,命名实体识别和实体关系抽取的目标是识别出电子病历文本中的医学实体并提取出各个实体间的医学关系。首先,系统阐述了中文电子病历的研究现状,指出命名实体识别和实体关系抽取2大任务在中文电子病历信息提取中所发挥的重要作用。随后,介绍了面向中文电子病历信息提取的命名实体识别和关系抽取算法的最新研究成果,并分析了每个阶段各个模型的优缺点。最后,讨论了中文电子病历现阶段所存在的问题并对未来的研究趋势进行展望。

基于CTSA-Net的急性肾损伤风险预测研究

针对过去对急性肾损伤(acute kidney injury,AKI)患者的识别存在临床时间序列数据未被充分利用、提前预测窗口较短及缺少连续预测研究等不足,本研究提出了一种卷积神经网络和两阶段交叉注意力的混合网络模型(CTSA-Net),实现对1期及以上AKI的每小时连续预测。CTSA-Net的注意力支路、CNN支路及特征融合模块可增强对时间序列数据的全局表示以及局部细节的感知能力,从而提高对AKI的连续预测性能。在AKI发生时、发生前24、48及72 h四个预测时间点,模型预测AKI的受试者工作特征曲线下面积分别为0.946、0.907、0.895和0.879,准确率-召回率曲线下面积分别为0.979、0.960、0.949和0.939。实验结果表明,CTSA-Net模型在多个预测时间点进行AKI预测的性能较好,可用于患者的实时监测,辅助医生进行临床决策。

Federated Learning Based on Extremely Sparse Series Clinic Monitoring Data

Decentralized machine learning frameworks,e.g.,federated learning,are emerging to facilitate learning with medical data under privacy protection.It is widely agreed that the establishment of an accurate and robust medical learning model requires a large number of continuous synchronous monitoring data of patients from various types of monitoring facilities.However,the clinic monitoring data are usually sparse and imbalanced with errors and time irregularity,leading to inaccurate risk prediction results.To address this issue,this paper designs a medical data resampling and balancing scheme for federated learning to eliminate model biases caused by sample imbalance and provide accurate disease risk prediction on multi-center medical data.Experimental results on a real-world clinical database MIMIC-Ⅳ demonstrate that the proposed method can improve AUC(the area under the receiver operating characteristic) from 50.1% to 62.8%,with a significant performance improvement of accuracy from 76.8% to 82.2%,compared to a vanilla federated learning artificial neural network(ANN).Moreover,we increase the model’s tolerance for missing data from 20% to 50% compared with a stand-alone baseline model.

基于深度学习与模糊聚类的档案数据分析算法

为了对智能化健康档案平台所采集到的海量数据进行细致分析,文中提出了一种基于深度学习与模糊聚类的档案数据分析算法,在健康数据档案的基础上对健康状况加以预测。该算法基于深度学习模型BiGRU对海量数据进行智能分析,并通过不断迭代学习获得数据中的关键特征信息。同时结合模糊C均值聚类方法对BiGRU所学习到的特征信息进行聚类,从而预测出健康状况。在UK BioBank公开电子医疗记录数据集上进行的实验结果表明,所提算法在测试集上的预测准确率为98.76%、精确率为98.81%、召回率为98.56%、F1值为98.68%,综合性能均优于原始BiGRU、LSTM以及TCN等深度学习算法。

基于多核学习的电子病历多视图聚类研究及应用

电子病历的分析和应用是智慧医疗领域研究的热点,提出一种基于多核学习的电子病历多视图聚类模型,对多核学习挖掘电子病历的非线性相关信息和高阶相关性信息、利用谱聚类对融合的核矩阵(相似性矩阵)进行聚类研究,得到多视图数据空间的聚类结果。实验结果表明,该方法能快速准确地对电子病历中的信息进行聚类,在性能指标上优于其它方法。

Dr.Deep:基于医疗特征上下文学习的患者健康状态可解释评估

深度学习是当前医疗多变量时序数据分析的主流方法.临床辅助决策关乎病人生命健康,因此深度模型需要抽取患者个性化表示,保证较高的分析、预测准确率;同时还需提供足够的可解释性,即能解释模型给出分析、预测结论的依据.而现有工作暂未能匹配医疗领域多变量时间序列数据的特性来进行个性化表示学习,同时源于深度学习的黑盒性质,现有模型大都可解释性不足,难以满足临床应用的需求.在此背景下,提出了基于医疗特征上下文学习的患者健康状态可解释评估方法Dr.Deep,将各变量的时序特征分别编码,利用多头去协同的自注意力机制,学习不同特征之间关联关系;提出了基于压缩激励机制的特征跳连编码,提升模型对最新病情变化的敏感性并针对不同患者情况分析特征重要性.实验表明:Dr.Deep在重症监护患者脓毒症预测、新冠肺炎重症患者出院时间预测等任务中相比业界方法性能提升,且可以针对不同患者的不同指标自适应学习其重要性作为可解释性的关键因素.同时设计并实现了基于医疗多变量时序数据分析的医生临床辅助系统,该系统建立病人的健康表示学习和预后预测模型并可视化患者病情进展以便医生分析.实验代码已开源于https:github.com Accountable-Machine-Intelligence Dr.Deep.所设计的智能医生可视化交互系统已发布于http:47.93.42.104 challenge 100049.

基于BERT的中文电子病历命名实体识别

针对中文电子病历命名实体识别过程中实体特征利用率低,语义表示不充分等问题,提出一种基于BERT语言模型的命名实体识别方法。运用Char-CNN学习字符的多种特征,将特征加入BERT预训练生成的词向量中,获得融合领域信息和汉字特征的词向量表示,将词向量输入迭代扩张卷积神经网络中进行特征抽取,引入注意力机制加强实体特征的关注度,通过CRF解码标注命名实体。实验结果表明,该方法在CCKS17中取得91.64%的F1值,识别性能优于现有方法。

用于心衰患者重入院预测的LSTM模型

电子健康记录(EHR)中蕴含着丰富的语义信息,目前对EHR的数据挖掘主要基于传统机器学习方法,涉及大量人工特征工程。但由于EHR数据存在维度大、时间跨度长等特点,传统机器学习方法较难有效地捕捉EHR中的深层语义信息。使用LSTM模型对心力衰竭患者的EHR数据进行建模和训练,以30天内重入院作为预测目标开展研究。实验结果表明,LSTM模型能够有效地捕捉EHR时序数据中的语义信息,与传统机器学习方法相比,在ROC-AUC指标上提升了10.48百分点。

机器学习在电子病历中应用的可视化分析

为探讨机器学习方法在电子病历领域应用的研究现状、研究热点与前沿,以2000~2022年中国知网数据库和Web of Science核心合集数据库中关于机器学习在电子病历中应用的相关文献为数据来源,运用CiteSpace软件绘制国家/地区、作者、机构、关键词共现以及关键词突现5个方面科学知识图谱进行可视化对比分析,以便了解国内外研究的差异,为该领域的研究和发展提供参考。

电子病历关系抽取综述

信息抽取在电子病历上的应用取得丰富的研究成果,使得非结构化的生物医学数据得以利用。关系抽取是信息抽取的重要子任务,是从数据转化为知识的桥梁。根据关系抽取存在的不同问题以及不同解决方案,对关系抽取进行详细分类。整理了电子病历关系抽取领域的相关评测任务和具有代表性的数据集。分阶段对关系抽取在电子病历文本上的应用进展进行综述,重点介绍了深度学习方法在关系抽取上的广泛应用,以及现阶段预训练模型在电子病历关系抽取任务上的进展。对该领域进行展望,提出了未解决的问题以及未来的研究方向。

基于双向长短期记忆网络的医院电子病历数据挖掘

在当前信息化智慧医院系统中,临床病历和门诊病历是追踪诊疗进度的重要依据,挖掘电子病历数据有助于专家对病情的掌控。针对这种情况,基于长短期记忆网络提出电子病历方面级观点分析系统,智能地分析诊断过程的治疗效果。借助现有工具识别电子病历的中文命名实体,过滤与医疗领域相关的方面项;提取每个句子的三种语法特征,将句子和相关特征送入长短期记忆网络进行分析;结合注意力机制对句子的上下文语义进行分析,识别每个方面项的情感极性。在江苏省第二中医院的电子病历数据集上完成了验证实验,结果表明该算法在中文电子病历数据上实现了较好的性能。

基于前馈神经网络电子病历辅助诊断系统的研究

目的:利用前馈神经网络实现电子病历的辅助诊断。方法:通过定义数据元,按照规则从入院记录、首次病程记录中提取临床信息;然后基于前馈神经网络训练学习模型进行辅助诊断。系统从首次病程记录的诊断依据中提取临床信息作为神经网络的输入,以病历的出院诊断(ICD-10疾病编码)作为输出,利用科室的归档病历训练模型。主管医生录入入院记录后,系统自动解析病历,推荐诊断。结果:基于该模型对心血管内科1000份、神经内科914份电子病历进行测试,诊断符合率分别达到医生诊断水平的86%、76%。结论:基于前馈神经网络的电子病历辅助诊断系统性能整体表现良好,是人工智能辅助诊断的有益探索。

基于知识增强的中文电子病历命名实体识别

针对中文电子病历中医疗嵌套实体难以处理的问题,本文基于RoBERTa-wwm-ext-large预训练模型提出一种知识增强的中文电子病历命名实体识别模型ERBEGP.RoBERTa-wwm-ext-large采用的全词掩码策略能够获得词级别的语义表示,更适用于中文文本.首先结合知识图谱,使模型学习到了大量的医疗实体名词,进一步提高模型对电子病历实体识别的准确性.然后通过BiLSTM对电子病历输入序列编码,能够更好捕获病历的中上下语义信息.最后利用全局指针网络模型EGP(efficient GlobalPointer)同时考虑实体的头部和尾部的特征信息来预测嵌套实体,更加有效地解决中文电子病历命名实体识别任务中嵌套实体难以处理的问题.在CBLUE中的4个数据集上本文方法均取得了更好的识别效果,证明了ERBEGP模型的有效性.