Development of a Deep Learning-Based Model for Pressure Injury Surface Assessment

Area of Science:

• Medical Artificial Intelligence
• Digital Health
• Wound Care Technology

Background:

• Pressure injuries pose a significant challenge in healthcare, necessitating accurate and efficient assessment methods.
• Current methods for pressure injury assessment can be subjective and time-consuming.

Purpose of the Study:

• To develop and evaluate a deep learning-based smart assessment model for pressure injury surface analysis.
• To enhance the precision and efficiency of wound assessment in clinical settings.

Main Methods:

• An exploratory analysis study utilizing a neural network model trained on 1063 pressure injury images from four hospitals.
• Evaluation metrics included mean intersection over union (MIoU), pixel accuracy (PA), and accuracy for segmentation.
• Model performance was validated by comparing automated estimations with manual measurements of wound number, dimensions, and area.

Main Results:

• The model achieved 74% IoU, 88% PA, and 83% accuracy for wound bed segmentation.
• High agreement was observed for wound number (Cohen's kappa: 0.810) and dimensions (correlation coefficients: 0.900 for length, 0.814 for width).
• Excellent correlation was found for regional extent (0.930).

Conclusions:

• The developed deep learning model demonstrates exceptional automated estimation capabilities for pressure injuries.
• This AI-based tool can serve as a crucial aid for informed decision-making in wound assessment.
• The model supports precision nursing, equitable resource use, and facilitates clinical decision-making and resource sharing.