Information and observer disagreement in histopathology

Area of Science:

• Pathology
• Information Theory
• Biostatistics

Background:

• Histopathological classification involves ordinal scales for disease grading (e.g., dysplasia, differentiation).
• Inter-observer agreement is crucial but often inversely related to the number of categories used.
• Information theory provides a framework to quantify classification performance.

Purpose of the Study:

• To evaluate the trade-off between inter-observer agreement and information transmission in histopathological classification.
• To propose an optimal approach for classifying histopathological continua.

Main Methods:

• Application of information theory concepts to model ordinal classification.
• Analysis of a working model demonstrating the effect of category reduction on kappa statistics and information transmission.
• Utilizing kappa statistics to measure inter-observer agreement.

Main Results:

• Reducing the number of categories increases inter-observer agreement (kappa statistics).
• Conversely, reducing categories decreases the amount of information transmitted.
• A working model illustrates this inverse relationship.

Conclusions:

• Histopathologists' primary role is transmitting visual pattern information to clinicians.
• More categories, not fewer, should be used to maximize information transfer.
• Suggests using 100-point scales with confidence limits and averaging observer judgments for reference ranges.