Measuring Patient Similarities via a Deep Architecture with Medical Concept Embedding

Zhu, Zhen; Yin, Changchang; Qian, Buyue; Cheng, Yu; Wei, Jishang; Wang, Fei

doi:10.1109/icdm.2016.0086

Cited by 100 publications

(55 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…As temporal EHR information is often informative, modeling it with CNNs requires considering how to capture temporality. For example, in 12 , 74 , an additional convolutional operation was conducted over the temporal dimension. In 103 , a hybrid convolutional recurrent neural network for joint feature extraction and temporal summarization was used.…”

Section: Resultsmentioning

confidence: 99%

Opportunities and challenges in developing deep learning models using electronic health records data: a systematic review

Xiao¹,

Choi

Sun

2018

Journal of the American Medical Informatics Association

614

346

View full text Add to dashboard Cite

ObjectiveTo conduct a systematic review of deep learning models for electronic health record (EHR) data, and illustrate various deep learning architectures for analyzing different data sources and their target applications. We also highlight ongoing research and identify open challenges in building deep learning models of EHRs.Design/methodWe searched PubMed and Google Scholar for papers on deep learning studies using EHR data published between January 1, 2010, and January 31, 2018. We summarize them according to these axes: types of analytics tasks, types of deep learning model architectures, special challenges arising from health data and tasks and their potential solutions, as well as evaluation strategies.ResultsWe surveyed and analyzed multiple aspects of the 98 articles we found and identified the following analytics tasks: disease detection/classification, sequential prediction of clinical events, concept embedding, data augmentation, and EHR data privacy. We then studied how deep architectures were applied to these tasks. We also discussed some special challenges arising from modeling EHR data and reviewed a few popular approaches. Finally, we summarized how performance evaluations were conducted for each task.DiscussionDespite the early success in using deep learning for health analytics applications, there still exist a number of issues to be addressed. We discuss them in detail including data and label availability, the interpretability and transparency of the model, and ease of deployment.

show abstract

Section: Resultsmentioning

confidence: 99%

Opportunities and challenges in developing deep learning models using electronic health records data: a systematic review

Xiao¹,

Choi

Sun

2018

Journal of the American Medical Informatics Association

614

346

View full text Add to dashboard Cite

show abstract

“…This form of unsupervised clustering has been documented [62][63][64] specificity. An example of this is the identification of signalling pathways specific to immune cell types including T cells (Figure 3c, Supplementary Table 3).…”

Section: Discussionmentioning

confidence: 99%

A systems genomics approach to uncover patient-specific pathogenic pathways and proteins in a complex disease

Brooks

Módos

Sudhakar

et al. 2019

Preprint

View full text Add to dashboard Cite

We describe a novel precision medicine workflow, the integrated single nucleotide polymorphism network platform (iSNP), designed to identify the exact mechanisms of how SNPs affect cellular regulatory networks, and how SNP co-occurrences contribute to disease pathogenesis in ulcerative colitis (UC). Using SNP profiles of 377 UC patients, we mapped the regulatory effects of the SNPs to a human signalling network containing protein-protein, miRNA-mRNA and transcription factor binding interactions. Unsupervised clustering algorithms grouped these patient-specific networks into four distinct clusters based on two large disease hubs, NFKB1 and PKCB. Pathway analysis identified the epigenetic modification as common and the T-cell specific responses as differing signalling pathways in the clusters. By integrating individual transcriptomes in active and quiescent disease setting to the patient networks, we validated the impact of non-coding SNPs. The iSNP approach identified regulatory effects of disease-associated non-coding SNPs, and identified how pathogenesis pathways are activated via different genetic modifications.

show abstract

“…al. [22] measured patient similarities from electronic medical records. Found the correct similarities and classification could ensure a cohort study and treatment.…”

Section: Word Embeddingmentioning

confidence: 99%

Towards early and automatic detection of Urinary Infection during pregnancy

Escobedo¹,

Hirales-Carbajal²

2019

EAI Endorsed Trans Perv Health Tech

View full text Add to dashboard Cite

INTRODUCTION: Worldwide Fetal-Maternal morbidity and mortality is frightfully high. Most of these diseases occur in developing countries. One of the main reasons for this problem, after gestational hypertension and complications in childbirth, is infections. Urinary Tract Infections (UTI) during pregnancy is one of the main causes for fetal-maternal morbidity and mortality in Mexico. Among others, the pervasiveness and heterogeneity of data in Electronic Medical Records (EMR) complicates early diagnosis and treatment of UTI. OBJECTIVES: Our goal is extract empirical knowledge, in the form of association rules, that generalize symptomatology and treatment of UTI patients with positive and nagative diagnosis. METHODS: In this study, we developed a criterion to extract words and expressions that uniquely characterize each patient class. We extracted association rules from EMRs and evaluated its level of correspondence between the rules and the extracted word sets. RESULTS: By defining a bound on word frequency usage and evaluating the positive to negative word ratio we were able to identify word sets that uniquely characterize each patient class. A bound of 47 enabled extraction of 25 unique words and expressions for each patient class. Further, approximately 17% and 27% of association rules drew terms from each word set correspondingly. CONCLUSION: This work seeks to promote the creation of more effective criterions to extract features, from EMRs, that improve characterization of patients and that ultimately lead to a more accurate diagnosis of UTIs.

show abstract

Measuring Patient Similarities via a Deep Architecture with Medical Concept Embedding

Cited by 100 publications

References 22 publications

Opportunities and challenges in developing deep learning models using electronic health records data: a systematic review

Opportunities and challenges in developing deep learning models using electronic health records data: a systematic review

A systems genomics approach to uncover patient-specific pathogenic pathways and proteins in a complex disease

Towards early and automatic detection of Urinary Infection during pregnancy

Contact Info

Product

Resources

About