PathEdEx - Uncovering High-explanatory Visual Diagnostics Heuristics Using Digital Pathology and Multiscale Gaze Data

Area of Science:

• Digital pathology
• Computational pathology
• Medical informatics

Background:

• Visual diagnostic heuristics in pathology remain largely unquantified, limiting clinical outcomes and potentially leading to diagnostic errors.
• Existing studies offer only qualitative insights into the visual decision-making processes of pathologists.

Purpose of the Study:

• To introduce PathEdEx, a computational framework for analyzing visual and non-visual diagnostic heuristics in digital pathology.
• To develop web-based interactive atlases for pathology education and datamining diagnostic patterns.

Main Methods:

• PathEdEx integrates whole-slide digital pathology images with multiscale gaze-tracking technology.
• Quantitative analysis of expert and novice zooming/panning dynamics was performed.
• Association rule mining and Markov chain modeling were used to identify diagnostic factors and expertise-based strategy differences.

Main Results:

• The framework successfully mined visual and non-visual diagnostic heuristics from a hematopathology atlas.
• Differences in diagnostic strategies between experts and novices were identified using gaze data translated into semantic labels.
• Quantitative insights into time dynamics of diagnostic operations were obtained.

Conclusions:

• The PathEdEx framework captures and quantifies expert pathology diagnostic heuristics for educational and practical applications.
• This approach has the potential to enhance pathology training and streamline precision diagnostics in precision medicine.