Scholarlink

<p>Significant research works have been carried out in the area of non-invasive hemoglobin estimation in recent times. The current study compares the estimation of hemoglobin from two kinds of photoplethysmogram (PPG) signals with the results obtained from the clinical laboratory tests. They are t-PPG and i-PPG signals. The former is collected using a device that utilizes the MAX30102 sensor, and the latter is collected using a smartphone. For this purpose, twenty-one volunteers aged between 20 and 54 years participated in this study. The data was collected twice on the same day with 2 hours interval. Among the volunteers five of them provided their data once. The PPG signals collected were filtered and processed to remove the noise. The features based on optical attenuation were extracted from the signals. These along with the demographic information were fed into the machine learning algorithms. The Hb values were estimated using decision trees, boosted trees, bootstrap, NTanH and XGBoost. The XGBoost model outperformed the others. The results were obtained with the R-squared values of 0.9862, 0.9812 and RMSE of 0.0489 g/dL, 0.0329 g/dL for t-PPG signals and i-PPG, respectively.</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

Non-invasive estimation of hemoglobin using machine learning algorithms from i-PPG and PPG signals

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年6月30日

查看全文
  • A. S. Kaviya Dharshini,
  • J. B. Jeeva

摘要

Significant research works have been carried out in the area of non-invasive hemoglobin estimation in recent times. The current study compares the estimation of hemoglobin from two kinds of photoplethysmogram (PPG) signals with the results obtained from the clinical laboratory tests. They are t-PPG and i-PPG signals. The former is collected using a device that utilizes the MAX30102 sensor, and the latter is collected using a smartphone. For this purpose, twenty-one volunteers aged between 20 and 54 years participated in this study. The data was collected twice on the same day with 2 hours interval. Among the volunteers five of them provided their data once. The PPG signals collected were filtered and processed to remove the noise. The features based on optical attenuation were extracted from the signals. These along with the demographic information were fed into the machine learning algorithms. The Hb values were estimated using decision trees, boosted trees, bootstrap, NTanH and XGBoost. The XGBoost model outperformed the others. The results were obtained with the R-squared values of 0.9862, 0.9812 and RMSE of 0.0489 g/dL, 0.0329 g/dL for t-PPG signals and i-PPG, respectively.