Related Experiment Videos
Evaluation of pulse-detection algorithms by computer simulation of hormone secretion
V Guardabasso1, G De Nicolao, M Rocchetti
1Biomathematics and Biostatistics Unit, Istituto di Ricerche Farmacologiche Mario Negri, Milan, Italy.
The American Journal of Physiology
|December 1, 1988
Summary
This study introduces a method for creating synthetic hormone data to test pulse detection algorithms. The findings highlight the importance of sampling frequency and algorithm settings for accurate hormone pulse analysis.
Area of Science:
- Endocrinology and Hormone Secretion Analysis
- Computational Biology and Algorithm Evaluation
Background:
- Accurate detection of hormone pulses is crucial for understanding physiological processes.
- Existing computerized algorithms for pulse detection require objective evaluation of their statistical error rates.
Purpose of the Study:
- To develop a versatile method for generating synthetic hormonal time series with known pulse locations.
- To enable objective evaluation of false-negative (F-) and false-positive (F+) error rates of pulse-detection algorithms.
Main Methods:
- Synthetic data generation by simulating hormone release as Poisson distributed pulses.
- Convolution of pulses to model hormone accumulation and clearance, with added experimental error.
- Emulation of physiological patterns for hormones like luteinizing hormone (LH), growth hormone (GH), and thyrotropin (TSH) by adjusting simulation parameters.
Main Results:
- Demonstrated that 'observable frequency' of hormone pulses is lower in sampled series than the true frequency.
- Analysis of LH secretion simulations using the DETECT program showed that minimizing false-positive (F+) rates can increase false-negative (F-) rates.
- Established methods for evaluating pulse-detection algorithms and presenting results.
Conclusions:
- Optimal selection of sampling frequency and algorithm probability levels is necessary for acceptable F+ and F- error rates.
- The synthetic data generation method provides a robust tool for assessing the performance of endocrine pulse-detection algorithms across various secretion patterns.