Scholarlink

<p>Background and objective: Electronic Health Records (EHR) are primarily used for administrative purposes in the healthcare sector. However, an insightful interpretation and utilization of these records can be greatly utilized to create a comprehensive clinical conceptual view. The objective of this paper is to present an architecture for extracting context features to design a clinical support system using transfer learning for Natural Language Processing (NLP) over biomedical data. These factors can be utilized to interpret the patients’ health state and extract necessary knowledge from the system to provide precise point-of-care clinical services. Methods: The intelligent support system is trained to provide a supporting summary to the medical practitioners on the basis of the respective symptoms and International Classification of Diseases – 9 (ICD-9) codes of the patient. The summary comprises the accurate phenotyping, line of treatment, possible side effects, special care to take, etc. for the patient with the respective medical situation using the patient cohort. We have used the BioALBERT model in this work, which has been trained over biomedical corpora. It has been fine-tuned to extract context from an EHR system that can be used for clinical purposes. The proposed approach requires less computational memory and utilizes improved parameter-sharing mechanisms. It implements the word-piece embeddings through sentence-piece tokenization for the fine-tuning of contextual summary generation. Result: The experimental analysis of the implemented system is consistent with the theoretical analysis. The MIMIC-III data has been utilized for fine tuning of the proposed model and evaluate the competence and effectiveness of the proposed work. The proposed model makes use of a CBOW technique and spontaneous parameter sharing approach that enhanced the performance of BioALBERT and reduced its physical memory requirement by 5.34%. The training process of BioALBERT has also become faster by nearly 4% because of the proposed changes. The designed system has successfully generated the context-based knowledge summaries on the basis of the ICD-9 codes given as input. Conclusion: A context-aware method for obtaining valuable context from EHRs is presented in this paper. In broad range of scenarios including emergency cases, clinicians might use this contextual information to obtain an elementary indication of the treatment path.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Context aware knowledge discovery for clinical support using transfer learning

  • 期刊论文

  • 发表日期：

    2026年7月6日

查看全文
  • Gaurav Paliwal,
  • Aaquil Bunglowala

摘要

Background and objective: Electronic Health Records (EHR) are primarily used for administrative purposes in the healthcare sector. However, an insightful interpretation and utilization of these records can be greatly utilized to create a comprehensive clinical conceptual view. The objective of this paper is to present an architecture for extracting context features to design a clinical support system using transfer learning for Natural Language Processing (NLP) over biomedical data. These factors can be utilized to interpret the patients’ health state and extract necessary knowledge from the system to provide precise point-of-care clinical services. Methods: The intelligent support system is trained to provide a supporting summary to the medical practitioners on the basis of the respective symptoms and International Classification of Diseases – 9 (ICD-9) codes of the patient. The summary comprises the accurate phenotyping, line of treatment, possible side effects, special care to take, etc. for the patient with the respective medical situation using the patient cohort. We have used the BioALBERT model in this work, which has been trained over biomedical corpora. It has been fine-tuned to extract context from an EHR system that can be used for clinical purposes. The proposed approach requires less computational memory and utilizes improved parameter-sharing mechanisms. It implements the word-piece embeddings through sentence-piece tokenization for the fine-tuning of contextual summary generation. Result: The experimental analysis of the implemented system is consistent with the theoretical analysis. The MIMIC-III data has been utilized for fine tuning of the proposed model and evaluate the competence and effectiveness of the proposed work. The proposed model makes use of a CBOW technique and spontaneous parameter sharing approach that enhanced the performance of BioALBERT and reduced its physical memory requirement by 5.34%. The training process of BioALBERT has also become faster by nearly 4% because of the proposed changes. The designed system has successfully generated the context-based knowledge summaries on the basis of the ICD-9 codes given as input. Conclusion: A context-aware method for obtaining valuable context from EHRs is presented in this paper. In broad range of scenarios including emergency cases, clinicians might use this contextual information to obtain an elementary indication of the treatment path.