Deep Learning based Skin-layer Segmentation for Characterizing Cutaneous Wounds from Optical Coherence Tomography...

Area of Science:

• Biomedical Imaging
• Dermatology
• Wound Healing Research

Background:

• Current wound care relies on surface-level visual assessment, neglecting critical subsurface information.
• Subsurface data is essential for accurate wound healing progression monitoring and severity assessment.
• Optical coherence tomography (OCT) offers a non-invasive method to capture this subsurface data.

Purpose of the Study:

• To demonstrate OCT's efficacy for objective, non-invasive assessment of wound severity and healing progress.
• To quantify wound healing dynamics, including epithelial and regenerated tissue thickness, using OCT.
• To develop automated analysis methods for OCT datasets in wound healing studies.

Main Methods:

• Utilized Optical Coherence Tomography (OCT) for in vivo imaging of small animal wound models.
• Employed a U-Net-based model for automated segmentation of skin layers in OCT images.
• Quantified wound closure dynamics by measuring epithelial and regenerated tissue thickness.
• Validated OCT findings against gold-standard histology and expert pathologist input.

Main Results:

• Achieved an average Intersection over Union (IOU) of 0.9234 for skin layer segmentation using the U-Net model.
• Demonstrated OCT's capability to objectively assess wound healing progression in vivo.
• Corroborated OCT results with histology and pathologist evaluations, confirming accuracy.

Conclusions:

• OCT is an effective tool for non-invasive, objective monitoring of wound healing progression.
• Automated OCT analysis, using models like U-Net, can accurately quantify wound healing dynamics.
• This approach enhances wound care by providing critical subsurface insights without disrupting the healing process.