# Clinical Assistant Diagnosis for Electronic Medical Record Based on Convolutional Neural Network

**Authors:** Zhongliang Yang, Yongfeng Huang, Yiran Jiang, Yuxi Sun, Yu-Jin Zhan, Pengcheng Luo  
**arXiv:** [1804.08261](https://arxiv.org/abs/1804.08261)  
**Date:** 2018-04-23  
**Summary page:** https://www.summarizepaper.com/en/paper/1804.08261v1/

## Key points

- Growing interest in automatically extracting useful information from electronic medical records (EMRs) and conducting disease diagnoses
- Existing systems rely on artificially constructed knowledge bases and rule matching for auxiliary diagnosis
- Authors propose a novel approach based on Convolutional Neural Networks (CNNs) for automatic extraction of high-level semantic information from EMRs and automatic diagnosis
- Experimental results show impressive accuracy of 98.67% and recall of 96.02%
- CNNs eliminate the need for manual rule construction or knowledge base creation, saving time and improving accuracy
- Study contributes to advancing the field by providing a promising solution for extracting valuable information from EMRs and conducting disease diagnoses
- Potential integration into clinical practice to provide clinicians with decision support tools
- Highlights the potential of neural network-based approaches in revolutionizing healthcare through improved diagnostic accuracy and efficiency

## Summary

In the field of clinical decision support (CDS) and neural language processing (NLP), there is a growing interest in automatically extracting useful information from electronic medical records (EMRs) and conducting disease diagnoses. Most existing systems rely on artificially constructed knowledge bases and rule matching for auxiliary diagnosis. However, in this study, the authors propose a novel approach based on Convolutional Neural Networks (CNNs) that can automatically extract high-level semantic information from EMRs and perform automatic diagnosis without the need for artificial construction of rules or knowledge bases. To evaluate their approach, the authors collected 18,590 real-world clinical EMRs to train and test their model. The experimental results demonstrate that their proposed model achieves an impressive accuracy of 98.67% and recall of 96.02%. These findings strongly support the feasibility and effectiveness of using CNNs to automatically learn high-level semantic features from EMRs and assist in diagnosis. This study contributes to advancing the field by providing a promising solution for extracting valuable information from EMRs and conducting disease diagnoses. By leveraging CNNs, which are known for their ability to capture complex patterns in data, the proposed approach eliminates the need for manual rule construction or knowledge base creation. This not only saves time but also improves accuracy by allowing the model to learn directly from the data. The authors' work has significant implications for clinical decision-making processes as it offers a more efficient and accurate method for diagnosing diseases based on EMR analysis. With further development and refinement, this approach could potentially be integrated into clinical practice to provide clinicians with valuable decision support tools. Overall, this study highlights the potential of neural network-based approaches in revolutionizing healthcare by leveraging large-scale EMR data for improved diagnostic accuracy and efficiency.