Automatic wound detection and size estimation using deep learning algorithms.

Area of Science:

• * Biomedical image analysis
• * Computational biology
• * Deep learning applications in research

Background:

• * Traditional wound size evaluation is manual, time-consuming, and subjective.
• * Accurate wound size tracking is crucial for diagnosis and treatment efficacy.
• * High-throughput studies require efficient and objective wound assessment methods.

Purpose of the Study:

• * To develop a deep learning-based image analysis pipeline for automated wound size assessment.
• * To extract key metrics like wound location, image crops, and size over time.
• * To address challenges of non-uniform images and small datasets in wound healing research.

Main Methods:

• * A deep learning pipeline was developed to analyze non-uniform wound images.
• * The system leverages a ring-shaped splint in mouse images for wound size prediction.
• * Applied to a challenging, small dataset (256 images) not originally intended for quantification.

Main Results:

• * The pipeline achieved high-fidelity results on unseen data with minimal human intervention.
• * Demonstrated preservation of information for predicting wound closure, despite inter-observer variability.
• * Accurately estimated wound sizes even when reference objects were missing in over 50% of images.

Conclusions:

• * The developed deep learning pipeline offers an efficient and accurate solution for wound size tracking.
• * It overcomes limitations of manual assessment, improving objectivity and throughput.
• * Shows promise for application in translational wound healing studies with challenging image data.