A hemoglobin concentration dataset derived from triple-wavelength photoplethysmography for machine learning...

Area of Science:

• Biomedical Engineering
• Medical Devices
• Health Informatics

Background:

• Invasive blood sampling for hemoglobin (Hb) measurement causes discomfort and pain.
• Photoplethysmography (PPG) offers a non-invasive alternative for detecting blood volume changes.
• PPG devices can potentially measure Hb levels without invasive procedures.

Purpose of the Study:

• To introduce a public dataset for developing non-invasive Hb measurement techniques.
• To facilitate research in machine learning applications for Hb estimation.
• To provide a resource for advancing PPG-based diagnostic tools.

Main Methods:

• Collected PPG signals using a MAXM86161 sensor with triple-wavelength light emission.
• Acquired data from 52 subjects (60% female, 40% male) with varying Hb levels.
• Included raw PPG signals, filtered outputs, and peak detection results in the dataset.

Main Results:

• The dataset comprises PPG signals, Hb values, age, and gender information.
• Hb values ranged from 12-20 g/dL for males and 8-18 g/dL for females.
• Data was collected in West Java, Indonesia, from April to May 2025.

Conclusions:

• The public dataset supports research on non-invasive Hb measurement.
• Machine learning models can be developed using this PPG signal data.
• This resource aids in the advancement of non-invasive diagnostic technologies.