Related Experiment Videos
Comparative assessments of objective peak-detection algorithms. II. Studies in men
R J Urban1, D L Kaiser, E van Cauter
1Department of Internal Medicine, University of Virginia School of Medicine, Charlottesville 22908.
The American Journal of Physiology
|January 1, 1988
Summary
Eight luteinizing hormone (LH) pulse-detection algorithms were evaluated. The Ultra, Cluster, and Detect programs demonstrated superior accuracy and reliability in identifying LH pulses, even with varying sampling intensities.
Area of Science:
- Endocrinology
- Biomedical Signal Processing
- Computational Biology
Background:
- Accurate detection of luteinizing hormone (LH) pulses is crucial for understanding reproductive physiology.
- Existing computerized algorithms for pulse detection vary in performance, necessitating comparative analysis.
- The impact of sampling intensity on LH pulse detection accuracy remains a significant concern.
Purpose of the Study:
- To compare the performance of eight commercially available computerized pulse-detection algorithms.
- To evaluate algorithm robustness against varying levels of signal-free noise and simulated sampling intensities.
- To determine the reliability of algorithms in detecting physiological LH pulses in healthy men.
Main Methods:
- Algorithms were tested using signal-free noise (4-36% variance) and physiological LH time series data from healthy men.
- Data were collected via immunoassay of blood samples taken every 5 minutes for 24 hours.
- Tests were conducted at a presumptive 1% false-positive rate, with adjustments for varying sampling intensities.
Main Results:
- Santen and Bardin programs showed elevated false-positive rates with increased variation.
- Regional Dual-Threshold program maintained a 1% false-positive rate, except with high variance.
- Ultra, Cluster, and Detect programs approximated a 1% false-positive rate across tested variances.
- All algorithms were sensitive to sampling intensity, affecting LH pulse frequency estimates.
- Ultra, Cluster, and Detect programs yielded statistically similar, reliable LH pulse frequency estimates.
- These three programs identified the same peaks with at least 72% concordance.
Conclusions:
- The Ultra, Cluster, and Detect algorithms demonstrate superior performance and reliability for LH pulse detection.
- These algorithms are less susceptible to sampling intensity variations compared to others.
- Their congruence in identifying physiological LH pulses suggests high accuracy and clinical utility.