During dialysis, up to 25% of patients may experience intradialysis hypotension (IDH). If hypotension develops during dialysis, doctors usually monitor the patient’s blood pressure and symptoms. If the dialysis machine settings need to be adjusted, this usually takes a few minutes. Dealing with the situation more slowly may exacerbate the patient’s discomfort. Therefore, we built an innovative deep learning model using the gated recurrent unit (GRU) neural network to analyze time-series medical record data to predict the occurrence and blood pressure values of dialysis hypotension. In addition, SHApley Additive exPlanations (SHAP) is used to analyze and predict the cause and inform doctors that the parameters of the dialysis machine should be adjusted to prevent the occurrence of dialysis hypotension. The excellent performance of this model can lead to safer and simpler hemodialysis treatment for patients.

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

Expert-Like Automatic Tuning of Dialysis Parameters for Dialysis Hypotension Prediction and Prevention

  • An-Chao Tsai,
  • Tzu-Yang Chiu,
  • Hsuan-Ming Lin,
  • Jhing-Fa Wang

摘要

During dialysis, up to 25% of patients may experience intradialysis hypotension (IDH). If hypotension develops during dialysis, doctors usually monitor the patient’s blood pressure and symptoms. If the dialysis machine settings need to be adjusted, this usually takes a few minutes. Dealing with the situation more slowly may exacerbate the patient’s discomfort. Therefore, we built an innovative deep learning model using the gated recurrent unit (GRU) neural network to analyze time-series medical record data to predict the occurrence and blood pressure values of dialysis hypotension. In addition, SHApley Additive exPlanations (SHAP) is used to analyze and predict the cause and inform doctors that the parameters of the dialysis machine should be adjusted to prevent the occurrence of dialysis hypotension. The excellent performance of this model can lead to safer and simpler hemodialysis treatment for patients.