Scholarlink

<p>To improve the long-term monitoring of patients receiving catheterized urination support, it is necessary to develop an automated tool that can monitor variations in urine color and void patterns during hospitalization. In this study, a novel intravenous (IV) pole-integrated urination-status monitoring technique was developed to detect the color and volume of in-bag liquids using a deep learning technique and to detect urinary disease symptoms, and performed a proof-of-concept simulation study using various simulated urine samples. In experiments, the error rates of in-bag liquid volume prediction were 5.14 ± 3.72%, 2.93 ± 5.70%, 2.48 ± 5.57%, and 2.00 ± 4.93%, at normal, hematuria, bilirubinuria, and purple urinary bag syndrome, respectively. The range of the average error rate between the threshold of the bag-flush request alarm and the model prediction was 0.71–1.08%. During the long-term testing over 24&#xa0;h, the prototype IV pole classified the types of urinary disease symptoms with 100% accuracy and estimated the total volume of void with an error rate of 14.47 ± 6.29%, 8.75 ± 4.61%, 15.43 ± 8.23%, 14.22 ± 8.13%, and 11.86 ± 4.73% at normal, polyuria, oliguria, anuria, and nocturnal polyuria, respectively. Based on these results, we conclude that the proposed IV pole-integrated urinary monitoring technique has the potential to be used as a tool for real-time, simplified urination-status monitoring of patients with catheterized urination support, and for improving the safety of patients with renal and urological diseases. Nevertheless, further clinical evaluations using actual urine samples are required in future studies.</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

Intravenous pole-integrated automated urinary status-monitoring technique using image-based artificial intelligence: a simulation study

  • 期刊论文

  • 发表日期：

    2025年11月3日

查看全文
  • Min Jae Kim,
  • Gun Ho Kim,
  • Subrata Bhattacharjee,
  • Hongje Lee,
  • Kyoung Won Nam

摘要

To improve the long-term monitoring of patients receiving catheterized urination support, it is necessary to develop an automated tool that can monitor variations in urine color and void patterns during hospitalization. In this study, a novel intravenous (IV) pole-integrated urination-status monitoring technique was developed to detect the color and volume of in-bag liquids using a deep learning technique and to detect urinary disease symptoms, and performed a proof-of-concept simulation study using various simulated urine samples. In experiments, the error rates of in-bag liquid volume prediction were 5.14 ± 3.72%, 2.93 ± 5.70%, 2.48 ± 5.57%, and 2.00 ± 4.93%, at normal, hematuria, bilirubinuria, and purple urinary bag syndrome, respectively. The range of the average error rate between the threshold of the bag-flush request alarm and the model prediction was 0.71–1.08%. During the long-term testing over 24 h, the prototype IV pole classified the types of urinary disease symptoms with 100% accuracy and estimated the total volume of void with an error rate of 14.47 ± 6.29%, 8.75 ± 4.61%, 15.43 ± 8.23%, 14.22 ± 8.13%, and 11.86 ± 4.73% at normal, polyuria, oliguria, anuria, and nocturnal polyuria, respectively. Based on these results, we conclude that the proposed IV pole-integrated urinary monitoring technique has the potential to be used as a tool for real-time, simplified urination-status monitoring of patients with catheterized urination support, and for improving the safety of patients with renal and urological diseases. Nevertheless, further clinical evaluations using actual urine samples are required in future studies.