Theory of sampling and its application in tissue based diagnosis

Area of Science:

• Biomedical Engineering
• Computational Pathology
• Diagnostic Imaging

Background:

• Presents a general theory of sampling applicable to tissue-based diagnosis.
• Defines sampling as reproducible information extraction from limited spaces for broader application.
• Highlights sampling's dual aspects: procedure of sample selection and performance efficiency.

Purpose of the Study:

• To introduce a general theory of sampling for tissue-based diagnosis.
• To differentiate between evaluating diagnostic significance and detection probability of objects.
• To explore the application of sampling in diagnostic processes.

Main Methods:

• Distinguishes two diagnostic procedures: evaluating diagnostic significance and detection probability.
• Describes sampling with or without external knowledge (e.g., object size, spatial distribution).
• Introduces random sampling (based on reference space) and stratified sampling (based on interspatial relationships).

Main Results:

• Random sampling estimates object number/size and features like area fraction and densities.
• Stratified sampling utilizes object knowledge to evaluate spatial features and define active segmentation parameters.
• Kriege's formula is applicable when sample size significantly exceeds object size.

Conclusions:

• The sampling method aids in standardizing immunohistochemically stained slides.
• The approach is implemented in the EAMUS system for diagnostic applications.
• Provides a formula for calculating sampling efficiency and potential error rates.