A pressure ulcers assessment system for diagnosis and decision making using convolutional neural networks

Area of Science:

• Medical technology
• Artificial intelligence in healthcare
• Dermatology

Background:

• Pressure ulcers are a significant clinical challenge in healthcare settings.
• Prolonged tissue compression leads to damage and potential infections.
• Accurate diagnosis and timely treatment are crucial for patient outcomes.

Purpose of the Study:

• To develop a deep learning system for pressure ulcer diagnosis.
• To assist healthcare providers in treatment decision-making.
• To provide a reliable tool for first-line caregivers.

Main Methods:

• Retrospective analysis of 327 pressure ulcer images (2016-2019).
• Image classification tasks: erythema/non-erythema and necrosis severity.
• Utilized an Inception-ResNet-v2 Convolutional Neural Network (CNN) model.
• Validated model accuracy with images from 2019-2020.

Main Results:

• CNN model achieved 98.5% accuracy in classifying erythema vs. non-erythema.
• CNN model achieved 97% accuracy in classifying necrotic tissue severity.
• Demonstrated high performance in pressure ulcer assessment.

Conclusions:

• The Inception-ResNet-v2 based CNN model is highly accurate for pressure ulcer classification.
• The developed system shows clinical applicability in real-world healthcare scenarios.
• This AI tool can enhance the management of pressure ulcers.