Computational modeling of human reasoning processes for interpretable visual knowledge: a case study with...

Area of Science:

• Medical Imaging Analysis
• Cognitive Science
• Human-Computer Interaction

Background:

• Visual reasoning is crucial for medical diagnoses like radiology and pathology.
• Current computational methods often use black-box algorithms, limiting understanding of visual reasoning.
• Explainable methods are needed to improve diagnostic accuracy and computational tools.

Purpose of the Study:

• To propose a computational method for quantifying and dissecting visual reasoning.
• To develop explainable visual reasoning patterns for comparison across different groups.
• To identify differences in visual reasoning strategies based on expertise levels.

Main Methods:

• Characterizing spatial and temporal features of visual attention.
• Identifying common and contrast visual reasoning patterns.
• Extracting significant gaze activities and comparing patterns among groups.

Main Results:

• The method successfully captures temporal and spatial features of visual attention.
• It can distinguish between different expertise levels in radiographers.
• Extracted patterns reveal key differences in visual reasoning processes between expertise levels.

Conclusions:

• The proposed method offers explainable insights into human visual reasoning in medical contexts.
• It has the potential to enhance diagnostic accuracy and inform the development of computational tools.
• Understanding expertise-related visual strategies can lead to improved medical training and decision support systems.