Region-Based Semi-Two-Stream Convolutional Neural Networks for Pressure Ulcer Recognition

Area of Science:

• Medical Imaging
• Artificial Intelligence in Healthcare
• Dermatology

Background:

• Pressure ulcers are a significant global health issue, causing pain, high costs, and affecting bedridden patients.
• Early and accurate diagnosis of pressure ulcers is crucial for effective treatment and international standardization.
• Current diagnostic methods, including invasive techniques and manual assessments using scales like Braden, have limitations.

Purpose of the Study:

• To develop and evaluate a novel image-based diagnostic method for pressure ulcers.
• To introduce a new benchmark dataset of pressure ulcer images for research and development.
• To improve the accuracy and efficiency of pressure ulcer staging using deep learning.

Main Methods:

• Creation of a novel benchmark dataset with pressure ulcer images.
• Development of a semi-two-stream approach combining original and cropped wound images.
• Evaluation of various state-of-the-art Convolutional Neural Network (CNN) architectures on the dataset.

Main Results:

• The proposed semi-two-stream method achieved high performance metrics.
• Experimental results demonstrated a test accuracy of 93%, precision of 93%, recall of 92%, and F1-score of 93%.
• The novel approach showed improved recognition results compared to baseline CNN architectures.

Conclusions:

• The developed semi-two-stream method offers a promising, non-invasive approach for accurate pressure ulcer diagnosis.
• The novel dataset and methodology contribute to advancing automated pressure ulcer staging.
• This AI-driven approach has the potential to enhance clinical decision-making and patient care.