In everyday hospital practice, the documentation of wounds is standard. A study has now been published in the Journal of Digital Imaging (2021) that validated a software-based method for the automatic segmentation and measurement of wounds on photos by Mask R-CNN. The study involved Dr. Michael Müller (CEO, mbits) and Hannah Syrek (former Head of Research & Development, mbits), among other researchers.

Manual wound documentation, which has predominated in hospitals to date, is time-consuming and often not very accurate. For this reason, a digital solution was developed whose performance was to be tested by the study. For this purpose, a data set was manually segmented by medical experts at an interval of one month at two different points in time. This dataset, consisting of 35 wound photographs, was also segmented automatically in parallel, using Mask R-CNN. The resulting segmentation results were compared and statistically evaluated.

In both evaluation rounds, no statistically significant differences could be found in the manual segmentation by the experts as well as in the automatic segmentation. However, slight differences in the quality of the segmentation were found between the different experts, whereas Mask R-CNN continuously provided an identical segmentation.

This leads to the conclusion that the use of the software-based method can accelerate the documentation process and optimize the consistency of the measured values while maintaining quality and precision.

The full study can be found here.