Validating Wound Severity Assessment via Region-Anchored Convolutional Neural Network Model for Mobile Image-Based...

Area of Science:

• Medical Technology
• Computer Science
• Biomedical Imaging

Background:

• Accurate wound size evaluation is critical for effective chronic wound management.
• Traditional methods for wound measurement can be subjective and time-consuming.
• Digital technologies offer potential for objective and efficient wound assessment.

Purpose of the Study:

• To introduce a novel method for wound size evaluation using mobile device-captured photographs.
• To apply deep learning (DL) and computer vision techniques for enhanced wound assessment.
• To assist healthcare professionals in decision-making for chronic wound treatment.

Main Methods:

• Utilizing mobile device photographs for wound assessment.
• Employing superpixel techniques to determine the wound's region of interest (RoI).
• Implementing a Region Anchored CNN framework for wound detection, differentiation, and tissue classification, using Resnet50.

Main Results:

• The DL method achieved high accuracy (0.85%) using Resnet50 for wound size detection.
• The Tissue Classification CNN demonstrated a Median Deviation Error of 2.91.
• The overall methodology showed a precision range of 0.96%, indicating effectiveness in real-world scenarios.

Conclusions:

• The proposed DL and computer vision approach offers an effective solution for objective wound size measurement.
• This technology has the potential to significantly improve patient care and therapeutic outcomes in chronic wound management.
• Mobile-based digital wound assessment can enhance clinical decision-making and treatment planning.