Non-Invasive Hemoglobin Estimation Using PPG Signal Features from 68 Subjects

A dataset of photoplethysmography (PPG) signals from 68 subjects used to estimate hemoglobin levels non-invasively. The data includes raw red and infrared light signals, with extracted features like mean, kurtosis, and skewness, plus demographic information. Author Bukao Ni published the dataset on figshare under a CC-BY-4.0 license, with a last update recorded as 2026-04-21.

Use Cases

• Training neural networks for hemoglobin level prediction based on extracted PPG signal features.
• Developing feature extraction pipelines for photoplethysmography data in clinical settings.
• Benchmarking non-invasive diagnostic methods against traditional blood-based measurements.
• Studying the relationship between demographic factors and physiological signals for anemia screening.

Strengths

• Includes data from 68 subjects, providing a basis for initial model development.
• Features a reported Mean Relative Error (MRE) of less than 2.46% for the hemoglobin prediction model described.
• Contains six extracted features per subject (mean, kurtosis, skewness for two signal types) alongside demographic data.

Limitations

• Row count is unknown, which may limit suitability assessment for large-scale modeling.
• Column-level documentation is absent; field semantics must be inferred after download.
• The dataset is small in scale at 139.2 KB, indicating limited scope.

Provenance

Source

figshare

Collection Method

Raw PPG data collected from 68 subjects, with features extracted for machine learning.

Freshness

Last updated 2026-04-21 04:20:06; freshness should be verified.