Linear and nonlinear techniques for the deconvolution of hormone time-series

Area of Science:

• Endocrinology and Hormone Dynamics
• Biomedical Signal Processing
• Mathematical Modeling in Biology

Background:

• Pulsatile hormone secretion is crucial for physiological regulation but challenging to measure directly.
• Traditional methods rely on peripheral plasma hormone concentrations, which are indirect measures.
• Accurate reconstruction of instantaneous secretion rates from concentration data is essential for understanding endocrine function.

Purpose of the Study:

• To compare various deconvolution techniques for reconstructing pulsatile hormone secretion rates.
• To address the ill-conditioning and computational challenges inherent in deconvolution problems.
• To develop and evaluate a novel nonlinear Maximum A Posteriori (MAP) estimator for improved accuracy.

Main Methods:

• Deconvolution of hormone time-series data to estimate instantaneous secretion rates.
• Comparison of linear techniques: least squares, Maximum A Posteriori (MAP) estimation, and Wiener filtering.
• Development and application of a new nonlinear MAP estimator accounting for non-Gaussian signal distributions.

Main Results:

• Linear deconvolution methods (least squares, MAP, Wiener filtering) were evaluated.
• The newly developed nonlinear MAP estimator demonstrated superior performance in reconstructing hormone secretion rates.
• Algorithm performance was validated using both simulated time-series and real Luteinizing Hormone (LH) data.

Conclusions:

• Deconvolution is a powerful tool for analyzing pulsatile hormone secretion.
• The proposed nonlinear MAP estimator offers a significant advancement for accurately determining instantaneous hormone release.
• This approach enhances the study of endocrine dynamics, exemplified by Luteinizing Hormone (LH) secretion patterns.