Samenvatting

The Nedap Healthcare business unit is developing a mobile wound recognition solution named Vesalius using Kotlin, Python and tools / libraries such as Anaconda and TensorFlow Lite in order to utilize machine learning to determine the boundary of a wound as bitmap based on its picture.This solution is aimed to become a SaaS product within the Ons Wondzorg mobile platform so that healthcare professionals can benefit from the wound assessments made by the Vesalius machine learning algorithm using the mobile solution.
The Vesalius MLA is not ready for commercial use yet due to the fact that it has to be trained further in order for it to give the most accurate predictions possible. In order to provide datasets for the MLA to learn from, a pipeline was constructed by the previous graduation students that allowed for the assessing of wound images on the mobile Kotlin solution using TensorFlow Lite after which the user is able to correct the predicted wound boundary mask and upload it to the MLA back-end running using Anaconda.However, to further the advancement of the Vesalius wound recognition software, the thesis student was asked to expand the user interface and improve the user experience so that healthcare professionals could eventually make use of the Vesalius software and gather user-corrected datasets for the MLA to improve on.Extensive background research has been conducted by asking critical questions to six stakeholders regarding the starting situation and the required improvements. Furthermore, the students’ specialization, human-machine interfacing, has been applied in order to research, design, test and validate the optimal form of UX / UI which resulted in a Hi-Fi Figma prototype that got tested by five survey & prototype participants.
The UX / UI of the mobile solution was expanded so that the solution could create additional masks according to the wound care classification methodology which together with the wound image itself would become the dataset that the Vesalius MLA will learn from and increase its assessment accuracy. The Vesalius Python back-end has been adapted to facilitate the sending of the new dataset.Lastly, changes were made to the UI so that it would conform to the Ons Wondzorg interface theme and iconography.