Wound Boundary Segmentation


About 2% of the population in most countries suffer from chronic wounds at one stage of their life. Chronic wounds heal very slowly (if at all) and – dependent on the underlying condition that is causing this delay – there are a large amount of possible treatment options available for each individual.

The most objective method of determining weather a wound is responding to a given treatment is to measure its physical dimensions at regular intervalls.

Measurement of wound dimensions (area, volume, circumference), however, depends on the ability of a human operator to outline the wound’s perimeter with great precision (repeatability).

An automatic or semi-automatic process could significantly enhance the precision of many non-contact wound measurement methods.


To produce an accurate and precise wound delineation algorithm


* to investigate existing delineation methods and to identify the most promising ones * to determine the precision of human operators * to design a range of new segmentation algorithems and to compare their performance against those of exisitng ones and against human operators.


Two semi-automated algorithms were produced and tested. Both constitute an improvement over exisitng algorithms in all cases and an improvment over human performance in most cases. Under extreme conditions (e.g. slough or coagulated blood spilling out of the wound) human operators perform better.

The successfully defended Phd thesis is available “here”:http://www.comp.glam.ac.uk/pages/staff/pplassma/MedImaging/Projects/Wounds/Segmentation/Tim%27s%20home%20page/ThesisOL/Title.htm, a summary report published in the IEEE Transactions on Medical Imaging in PDF format can be downloaded “from here”:http://www.comp.glam.ac.uk/pages/staff/pplassma/MedImaging/Projects/Wounds/Segmentation/Tim%27s%20home%20page/19mi12-jones.pdf.