Segmenting skin ulcers and measuring the wound area using deep convolutional networks.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Wound Care Technology

Background:

• Chronic skin ulcers in bedridden patients require regular home examination.
• Monitoring wound healing involves assessing depth, location, and size through imaging.
• Manual wound size measurement is inaccurate, time-consuming, and uncomfortable for patients.

Purpose of the Study:

• To introduce the Automatic Skin Ulcer Region Assessment (ASURA) framework.
• To enable accurate wound segmentation and automatic size measurement.
• To overcome limitations of manual wound assessment methods.

Main Methods:

• ASURA employs an encoder/decoder deep neural network for image segmentation.
• The framework identifies measurement rulers/tapes within images.
• Pixel density estimation is integrated into the segmentation process.

Main Results:

• ASURA achieves a Dice score over 90% for wound segmentation, outperforming state-of-the-art by 16%.
• Automatic pixel density estimation has a relative error of 5%.
• Semi-automatic wound area estimation in cm² shows a 14% relative error.

Conclusions:

• ASURA demonstrates suitability for segmenting and automatically measuring skin ulcers.
• The framework offers a more accurate and efficient approach to wound monitoring.
• ASURA has the potential to improve patient care for chronic wounds.