Toward Objective Wound Edge Classification in Clinical Practice

Area of Science:

• Medical imaging
• Computational pathology
• Wound care technology

Background:

• Chronic wound evaluation relies on subjective wound edge assessment.
• Inter-observer variability is a significant issue, especially with 2D clinical photographs.
• Objective and reproducible methods are needed to enhance wound assessment accuracy.

Purpose of the Study:

• To assess the subjectivity of clinical wound edge assessment.
• To develop and evaluate an automated image-analysis pipeline for wound edge characterization.
• To determine if geometric and visual features can improve agreement in wound edge classification.

Main Methods:

• Retrospective analysis of 1,860 wound images from routine clinical practice.
• Independent annotation of images by four expert clinicians.
• Development of an automated pipeline for wound segmentation, border standardization, and 3D edge profile estimation.
• Testing geometry-derived profiles and incorporating visual descriptors (shape, color, pattern) into a supervised classifier.

Main Results:

• Low inter-clinician agreement confirmed the subjectivity of wound edge classification.
• Geometry-based analysis alone showed poor correspondence with clinical annotations.
• A supervised classifier using both geometric and visual features achieved agreement comparable to or exceeding inter-clinician agreement.

Conclusions:

• Clinical wound edge assessment is influenced by visual cues beyond edge geometry.
• Automated image-based analysis, incorporating diverse features, shows potential for more reproducible wound edge assessment.
• External validation in varied clinical settings is crucial for automated systems.