Application of deep learning to pressure injury staging

Area of Science:

• Medical technology
• Artificial intelligence in healthcare
• Wound care informatics

Background:

• Accurate pressure injury (PI) staging is crucial for patient outcomes.
• Junior and non-specialist nurses often struggle with PI staging due to limited experience.
• Existing methods for PI assessment can be time-consuming and prone to inaccuracies.

Purpose of the Study:

• To develop and evaluate a deep learning-based system for automated PI staging and tissue classification.
• To enhance the accuracy and efficiency of PI assessment in clinical settings.
• To provide a cost-effective tool to support nurses in PI management.

Main Methods:

• Utilized a dataset of 1610 clinical PI cases with expert-verified staging and tissue labels.
• Trained a Mask Region-based Convolutional Neural Network (Mask R-CNN) for object detection and instance segmentation.
• Developed a recognition system for automatic staging and tissue classification from remotely uploaded PI photographs.

Main Results:

• The deep learning model achieved an average precision of 0.603 for PI stage recognition on a test set of 100 samples.
• The model's performance surpassed that of medical personnel, including an enterostomal therapist, in comparative evaluations.
• The system demonstrated evaluation performance comparable to a professionally trained wound care nurse.

Conclusions:

• The developed deep learning system offers a viable solution for accurate and efficient PI staging.
• This low-cost system can assist less experienced nurses in overcoming challenges in PI identification.
• The tool provides valuable auxiliary clinical information, potentially improving patient care and reducing healthcare expenses.