Figure - available from: Scientific Reports
This content is subject to copyright. Terms and conditions apply.
GAT-ETM model overview. (a) The probabilistic graphical model view of GAT-ETM. (b) The augmentation and merger applied on the taxonomy knowledge graphs. (c) The illustration of the deep learning architecture used to perform variational inference over the GAT-ETM model.
Source publication
The rapid growth of electronic health record (EHR) datasets opens up promising opportunities to understand human diseases in a systematic way. However, effective extraction of clinical knowledge from EHR data has been hindered by the sparse and noisy information. We present Graph ATtention-Embedded Topic Model (GAT-ETM), an end-to-end taxonomy-know...
Citations
... GNN can be applied to graphs, and it is very easy to do prediction task at level of node, edge, and graph. Graphical representation of clinical data that is based on medical knowledge will be of knowledge graph or representing the EHR as graph [18]. So, we use HGNN (Heterogeneous graph neural network) to model heterogeneity attributes like doctor notes, lab result, diagnosis etc in EHR data. ...
... Yuri Ahuja et.al [17] proposed automatic phenotyping model MixEHR- guided that will identify latent phenotyping structure in the data and should improve towards continuous data modalities. Yuesongzou et.al [18] proposed knowledge graph embedded topic modelling an end to end bas ed on multi model to find out the latent disease topics from the EHR data. Emma Rocheteau et.al [19] used LSTM-GNN for patient outcome prediction task to extract temporal features and patient neighbourhood information. ...
... The multidimensional nature of pain (21) makes it a good use case for application of such KGE methods. Other EHRbased use cases include patient stratification and drug identification (22), and disease relation extraction (23). This paper describes the development of KGE models of pain incorporating both pain concepts found within a mental health EHR database, and external knowledge about these concepts from a knowledge base, SNOMED CT (24) (detailed in the Methods section), for use in research on the relationships between mental health, pain, and physical multimorbidities. ...
... Another contribution from the field of data science, specifically about using machine learning and natural language processing techniques to analyze electronic health record (EHR) data is the best paper by Zou et al. [12] who use an end-to-end knowledge-graph-informed topic model. They discuss the challenges of extracting clinical knowledge from EHR data and propose a new method called the Graph ATtention-Embedded Topic Model (GAT-ETM) to address these challenges. ...
Objective: In this synopsis, the editors of the Clinical Information Systems (CIS) section of the IMIA Yearbook of Medical Informatics overview recent research and propose a selection of best papers published in 2022 in the CIS field.
Methods: The editors follow a systematic approach to gather relevant articles and select the best papers for the section. This year, they updated the query to incorporate the topic of telemedicine and removed search terms related to geographic information systems. The revised query resulted in a larger number of identified papers, necessitating the appointment of a third section editor to handle the increased workload. The editors narrowed the initial pool of articles to 15 candidate papers through a multi-stage selection process. At least seven independent reviews were collected for each candidate paper, and a selection meeting with the IMIA Yearbook editorial board led to the final selection of the best papers for the CIS section.
Results: The query was carried out in mid-January 2023 and retrieved a deduplicated result set of 5,206 articles from 1,500 journals. This year, 15 papers were nominated as candidates, and four were finally selected as the best papers in the CIS section.
Including telemedicine in the query resulted in a substantial increase in the number of papers found. The analysis highlights the growing convergence between clinical information systems and telemedicine, with mobile health (mHealth) technologies and data science applications gaining prominence. The selected candidate papers emphasize the practical impact of research efforts, focusing on patient-centric outcomes and benefits, including intelligent mobile health monitoring systems and AI-assisted decision-making in healthcare.
Conclusions: Looking ahead, the field of CIS is expected to continue evolving, driven by advances in telemedicine, mHealth technologies, data science, and AI integration, leading to more efficient, patient-oriented, and intelligent healthcare systems and overall improvement of global healthcare outcomes.
... The majority of these approaches employ variational auto-encoder and amortized inference to reduce input data dimensionality (Rezende et al., 2014;Dieng et al., 2020). Among them, ETM (Dieng et al., 2020) and its extensions (Zou et al., 2022; are neural topic models that use word embeddings and also learn topic embeddings. In our study, the technical framework is based on the ETM approach, which leverages the explanatory power of topic modelling to explain subgroups while also incorporating semantic features through the use of embedding representations. ...
Precision medicine aims to provide personalized healthcare for patients by stratifying them into subgroups based on their health conditions, enabling the development of tailored medical management. Various decision support systems (DSSs) are increasingly developed in this field, where the performance is limited to their capability of handling big amounts of heterogeneous and high-dimensional electronic health records (EHRs). In this paper, we focus on developing a deep learning model for patient stratification that can identify and explain patient subgroups from multimodal EHRs. The primary challenge is to effectively align and unify heterogeneous information from various modalities, which includes both unstructured and structured data. Here, we develop a Contrastive Multimodal learning model for EHR (ConMEHR) based on topic modelling. In ConMEHR, modality-level and topic-level contrastive learning (CL) mechanisms are adopted to obtain a unified representation space and diversify patient subgroups, respectively. The performance of ConMEHR will be evaluated on two real-world EHR datasets and the results show that our model outperforms other baseline methods.
... The multidimensional nature of pain (21) makes it a good use case for application of such KGE methods. Other EHRbased use cases include patient stratification and drug identification (22), and disease relation extraction (23). This paper describes the development of KGE models of pain incorporating both pain concepts found within a mental health EHR database, and external knowledge about these concepts from a knowledge base, SNOMED CT (24) (detailed in the Methods section), for use in research on the relationships between mental health, pain, and physical multimorbidities. ...
Pain is a complex concept that can interconnect with other concepts such as a disorder that might cause pain, a medication that might relieve pain, and so on. To fully understand the context of pain experienced by either an individual or across a population, we may need to examine all concepts related to pain and the relationships between them. This is especially useful when modeling pain that has been recorded in electronic health records. Knowledge graphs represent concepts and their relations by an interlinked network, enabling semantic and context-based reasoning in a computationally tractable form. These graphs can, however, be too large for efficient computation. Knowledge graph embeddings help to resolve this by representing the graphs in a low-dimensional vector space. These embeddings can then be used in various downstream tasks such as classification and link prediction. The various relations associated with pain which are required to construct such a knowledge graph can be obtained from external medical knowledge bases such as SNOMED CT, a hierarchical systematic nomenclature of medical terms. A knowledge graph built in this way could be further enriched with real-world examples of pain and its relations extracted from electronic health records. This paper describes the construction of such knowledge graph embedding models of pain concepts, extracted from the unstructured text of mental health electronic health records, combined with external knowledge created from relations described in SNOMED CT, and their evaluation on a subject-object link prediction task. The performance of the models was compared with other baseline models.
