Scholarlink

Background <p>Accurate and rapid assessment of fluid status of maintenance hemodialysis (MHD) patients and maintaining fluid balance is essential to ensure the quality of dialysis treatment. Currently, clinical methods for assessing ultrafiltration volume are still insufficient, and reliable tools that are more accurate and rapid are needed. The objective of this study was to construct a model for predicting ultrafiltration volume (UF) in MHD patients based on artificial neural network (ANN) algorithms, to validate and evaluate this model, and to investigate the impact of body composition prior to dialysis on UF in MHD patients.</p> Methods <p>A total of 319 patients undergoing MHD treatment at our center were enrolled. Basic demographic and clinical characteristics were collected and evaluated using the hemodialysis information system. Body composition was measured on ≥ 3 separate days before dialysis treatment using an Inbody bioimpedance instrument. The target ultrafiltration volume was determined by nephrologists based on the integration of body composition measurements and clinical characteristics, yielding a dataset of 1,205 entries. Heat maps were used to demonstrate the correlation between body composition and UF in MHD patients, and LASSO regression and multifactorial linear regression were used to screen the relevant indicator factors for final inclusion in the model, and Backpropagation Neural Network model (BPNN) was developed using the MATLAB (R2022a) neural network toolbox to establish the projected relationship between UF and pre-dialysis body composition. The effectiveness of the model was assessed based on the coefficient of determination (R<sup>2</sup>) and root mean square error (RMSE) of the calculated regression.</p> Results <p>The artificial neural network model demonstrated an optimal predictive performance metric of R<sup>2</sup> = 0.965 for forecasting ultrafiltration volume in MHD patients. With an average difference of 0.182&#xa0;L between observed and predicted values, and highlighted the significant influence of certain body composition indicators on UF in MHD patients.</p> Conclusion <p>This study effectively demonstrates the predictive role of an artificial neural network model based on pre-dialysis body composition information in estimating ultrafiltration providing a valuable predictive tool to optimize assessment volume for MHD patients, of ultrafiltration volume in MHD patients.</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

Prediction of ultrafiltration volume in maintenance hemodialysis patients using an artificial neural network model based on body composition information

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年11月11日

查看全文
  • Jiaoyan Chen,
  • Jurong Yang,
  • Xianqiong Lu,
  • Jingrong Peng,
  • Liangji He,
  • Wei Tan,
  • Qing Yu,
  • Yunyan Wang

摘要

Background

Accurate and rapid assessment of fluid status of maintenance hemodialysis (MHD) patients and maintaining fluid balance is essential to ensure the quality of dialysis treatment. Currently, clinical methods for assessing ultrafiltration volume are still insufficient, and reliable tools that are more accurate and rapid are needed. The objective of this study was to construct a model for predicting ultrafiltration volume (UF) in MHD patients based on artificial neural network (ANN) algorithms, to validate and evaluate this model, and to investigate the impact of body composition prior to dialysis on UF in MHD patients.

Methods

A total of 319 patients undergoing MHD treatment at our center were enrolled. Basic demographic and clinical characteristics were collected and evaluated using the hemodialysis information system. Body composition was measured on ≥ 3 separate days before dialysis treatment using an Inbody bioimpedance instrument. The target ultrafiltration volume was determined by nephrologists based on the integration of body composition measurements and clinical characteristics, yielding a dataset of 1,205 entries. Heat maps were used to demonstrate the correlation between body composition and UF in MHD patients, and LASSO regression and multifactorial linear regression were used to screen the relevant indicator factors for final inclusion in the model, and Backpropagation Neural Network model (BPNN) was developed using the MATLAB (R2022a) neural network toolbox to establish the projected relationship between UF and pre-dialysis body composition. The effectiveness of the model was assessed based on the coefficient of determination (R2) and root mean square error (RMSE) of the calculated regression.

Results

The artificial neural network model demonstrated an optimal predictive performance metric of R2 = 0.965 for forecasting ultrafiltration volume in MHD patients. With an average difference of 0.182 L between observed and predicted values, and highlighted the significant influence of certain body composition indicators on UF in MHD patients.

Conclusion

This study effectively demonstrates the predictive role of an artificial neural network model based on pre-dialysis body composition information in estimating ultrafiltration providing a valuable predictive tool to optimize assessment volume for MHD patients, of ultrafiltration volume in MHD patients.