Scholarlink

<p>This study introduces a novel non-invasive hemoglobin (Hb) monitoring approach using machine learning models, providing an alternative to traditional, invasive methods commonly used for diagnosing anemia and blood-related disorders. Hemoglobin (Hb) monitoring is vital for the timely diagnosis and management of anemia and other blood-related disorders, yet traditional methods are often invasive and resource-intensive. Accurate monitoring of hemoglobin (Hb) levels is crucial for diagnosing and managing various health conditions. This research explores advanced machine learning techniques—specifically CatBoost, LightGBM, and XGBoost—for non-invasive Hb detection using photoplethysmography (PPG) signals. The dataset includes PPG signals, gender, age, and Hb values collected from 68 subjects aged 18–65&#xa0;years. Comprehensive data preprocessing, sensitivity analysis, and exploratory factor analysis were conducted to understand variable relationships. Models were developed individually and as hybrids with novel applications of Grey Wolf Optimizer (GWO) and Slime Mould Algorithm (SMA) optimization techniques. The study findings highlight that the XGBoost-SMA hybrid model achieved superior predictive accuracy with an <i>R</i><sup>2</sup> value of 0.997232 and a VAF value of 99.72404, demonstrating its potential for real-time Hb monitoring.</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 hemoglobin level estimation using photoplethysmography with performance comparison of advanced ML algorithms

  • 期刊论文

  • 发表日期：

    2025年4月2日

查看全文
  • Yalin Song

摘要

This study introduces a novel non-invasive hemoglobin (Hb) monitoring approach using machine learning models, providing an alternative to traditional, invasive methods commonly used for diagnosing anemia and blood-related disorders. Hemoglobin (Hb) monitoring is vital for the timely diagnosis and management of anemia and other blood-related disorders, yet traditional methods are often invasive and resource-intensive. Accurate monitoring of hemoglobin (Hb) levels is crucial for diagnosing and managing various health conditions. This research explores advanced machine learning techniques—specifically CatBoost, LightGBM, and XGBoost—for non-invasive Hb detection using photoplethysmography (PPG) signals. The dataset includes PPG signals, gender, age, and Hb values collected from 68 subjects aged 18–65 years. Comprehensive data preprocessing, sensitivity analysis, and exploratory factor analysis were conducted to understand variable relationships. Models were developed individually and as hybrids with novel applications of Grey Wolf Optimizer (GWO) and Slime Mould Algorithm (SMA) optimization techniques. The study findings highlight that the XGBoost-SMA hybrid model achieved superior predictive accuracy with an R2 value of 0.997232 and a VAF value of 99.72404, demonstrating its potential for real-time Hb monitoring.