Classification of pressure ulcer tissues with 3D convolutional neural network

Area of Science:

• Medical imaging
• Artificial intelligence
• Wound care technology

Background:

• Accurate classification of pressure ulcer tissues is crucial for effective treatment.
• Current methods for tissue characterization can be subjective and time-consuming.

Purpose of the Study:

• To develop and evaluate a 3D Convolutional Neural Network (CNN) for automated classification and segmentation of granulation, necrotic eschar, and slough tissues in pressure ulcer images.
• To assess the accuracy and robustness of the proposed deep learning framework.

Main Methods:

• Utilized a 3D CNN architecture for image analysis.
• Extracted visual features from original and Gaussian-filtered 3D HSI images within a region of interest.
• Employed first-order models approximating probability distributions using linear combinations of discrete Gaussians (LCDG).
• Trained and tested the framework on 193 pressure ulcer images.

Main Results:

• Achieved a preliminary Dice Similarity Coefficient (DSC) of 92% for tissue classification.
• Obtained a Percentage Area Distance (PAD) of 13% and an Area Under the ROC Curve (AUC) of 95%.
• Demonstrated promising accuracy and robustness in classifying different tissue types.

Conclusions:

• The 3D CNN framework shows significant potential for accurate and objective pressure ulcer tissue classification.
• This AI-driven approach could enhance clinical decision-making and improve patient outcomes in wound management.
• Further validation on larger datasets is warranted to confirm clinical utility.