Efficient detection of wound-bed and peripheral skin with statistical colour models

Area of Science:

• Biomedical Engineering
• Medical Imaging
• Computer Vision

Background:

• Pressure ulcers are localized tissue damage requiring precise evaluation for effective treatment.
• Accurate wound assessment is critical for successful clinical management.
• Existing diagnostic methods can be subjective and time-consuming.

Purpose of the Study:

• To develop and validate a computer-vision approach for automated wound-area detection.
• To utilize statistical color models for precise segmentation of pressure ulcer tissues.
• To improve the reliability and efficiency of wound diagnosis.

Main Methods:

• A computer-vision technique based on statistical color models was employed.
• Color histogram models were created for four distinct tissue types using a training set of 113 wound images.
• Bayesian estimation and back-projections were used to classify pixels based on tissue type.

Main Results:

• The developed model achieved high performance rates in segmenting wound areas, including healing tissues.
• Validated on 322 wound images, the model demonstrated robustness with high mean performance metrics (AUC: .9426, accuracy: .8777, F-score: 0.7389, Cohen's kappa: .6585).
• The system effectively distinguishes between different tissue types within pressure ulcers.

Conclusions:

• The proposed computer-vision approach offers a robust and accurate method for pressure ulcer area detection.
• This automated segmentation technique supports reliable diagnosis and informed treatment decisions.
• The findings highlight the potential of computer vision in enhancing wound evaluation and patient care.