Image analysis of representative food structures: application of the bootstrap method

Area of Science:

• Food science
• Materials science
• Image analysis

Background:

• Quantitative image analysis relies on accurate sampling of microstructure.
• Estimating appropriate sampling area size is essential for reliable results.
• Current methods may not adequately account for image heterogeneity.

Purpose of the Study:

• To propose and validate the bootstrap method for estimating optimal sampling area size in quantitative image analysis.
• To establish a relationship between image characteristics and required sampling area.
• To provide a tool for determining sampling parameters for food microstructure analysis.

Main Methods:

• Application of the bootstrap method to estimate sampling area size.
• Analysis of simulated (computer-generated circles) and real (apple tissue) structures.
• Calculation of coefficient of variation (CV) and standard error (SE) of bootstrap estimates (CV(Bn), SE(Bn)).

Main Results:

• Increasing sampling area size decreased CV(Bn) and SE(Bn) for both simulated and real structures.
• A linear relationship was observed between image variation (CV(image)) and bootstrap variation (CV(Bn)).
• Higher image heterogeneity necessitates larger sampling areas or increased image acquisition.

Conclusions:

• The bootstrap method provides a robust approach to estimate the required sampling area size in quantitative image analysis.
• Sampling area requirements are directly influenced by the heterogeneity of the microstructure.
• This method aids in optimizing image analysis protocols for accurate food microstructure characterization.