Evolution strategies for optical tomographic characterization of homogeneous media

Area of Science:

• Biomedical optics
• Image reconstruction
• Computational modeling

Background:

• Diffuse optical tomography (DOT) reconstruction is an optimization problem.
• Gradient-based methods often converge to local minima or fail with poor initial guesses.
• Inefficiency of gradient-based schemes for global minimum search in DOT.

Purpose of the Study:

• Introduce evolution-strategy (ES) algorithms for DOT.
• Compare ES algorithms with gradient-based schemes for DOT reconstruction.
• Evaluate ES algorithm performance in finding optical properties.

Main Methods:

• Formulating DOT reconstruction as an optimization problem.
• Implementing and applying evolution-strategy (ES) algorithms.
• Comparing ES algorithms against gradient-based search algorithms.

Main Results:

• ES algorithms demonstrate robustness in finding global minima.
• ES algorithms are less sensitive to initial guesses than gradient-based methods.
• ES algorithms perform effectively in regions with small gradients.

Conclusions:

• Evolution-strategy algorithms are a promising alternative for DOT reconstruction.
• ES algorithms offer improved accuracy and reliability in estimating optical properties.
• ES algorithms mitigate common issues faced by gradient-based DOT reconstruction techniques.