Development of a deep learning-based tool to assist wound classification.

Area of Science:

• Medical imaging
• Artificial intelligence
• Wound care

Background:

• Accurate wound classification is crucial for effective patient management.
• Non-specialized medical personnel often require assistance in identifying complex wound types.
• Existing classification methods may lack the accuracy and efficiency needed in clinical settings.

Purpose of the Study:

• To develop and evaluate a deep learning-based tool for classifying five critical wound conditions.
• To provide an accurate and accessible wound classification solution for medical personnel without specialized wound care training.
• To create a unified deep learning architecture capable of simultaneously classifying multiple wound types.

Main Methods:

• A multi-task deep learning framework utilizing a convolutional neural network (CNN) was developed.
• The model was trained to classify five key wound conditions: deep, infected, arterial, venous, and pressure wounds.
• Performance was evaluated by comparing the model's classification accuracy against human medical personnel using Cohen's kappa coefficients.

Main Results:

• The deep learning model achieved high accuracy in classifying the five specified wound conditions.
• The model's performance was found to be superior or non-inferior to that of all human medical personnel evaluated.
• This represents the first CNN-based model capable of simultaneously classifying deep, infected, arterial, venous, and pressure wounds with notable accuracy.

Conclusions:

• The developed deep learning model offers a reliable and accurate method for wound classification.
• The tool can significantly assist non-specialized medical staff in making informed wound management decisions.
• The compact and high-performing AI model has the potential for integration into mobile applications for widespread clinical use.