Simultaneous Segmentation and Classification of Pressure Injury Image Data Using Mask-R-CNN

Area of Science:

• Medical imaging analysis
• Artificial intelligence in healthcare
• Computer vision for clinical applications

Background:

• Pressure injuries (PIs) affect millions, incurring significant healthcare costs and posing diagnostic challenges.
• Accurate staging of PIs is crucial for effective patient management but remains clinically difficult.
• Advancements in object detection and semantic segmentation offer new possibilities for medical image analysis.

Purpose of the Study:

• To develop and evaluate a Mask R-CNN model for automated segmentation and classification of pressure injuries.
• To assess the accuracy of the Mask R-CNN model in distinguishing between stage 1-4 pressure injuries.

Main Methods:

• Utilized the Mask R-CNN deep learning algorithm for image segmentation and classification.
• Trained the model on 969 pressure injury images from the eKare Inc. repository.
• Validated the model's performance on a separate test set of 121 images.

Main Results:

• The Mask R-CNN model achieved an overall classification accuracy of 92.6% and segmentation accuracy of 93.0%.
• High F1 scores were reported for all stages (1-4), ranging from 0.842 to 0.947.
• Dice coefficients for stages 1-4 demonstrated strong segmentation performance, with values from 0.85 to 0.93.

Conclusions:

• The Mask R-CNN model demonstrates high accuracy in pressure injury staging, surpassing average healthcare professional performance.
• This AI-powered tool can be readily integrated into clinical workflows to assist healthcare providers.
• The developed model offers a promising solution for improving the accuracy and efficiency of pressure injury assessment.