The versatility of the hands is revealed in its movements, but often not noticed before trauma occurs. Joint range of motion is used as a measure to follow the progress of diseases. A digital workflow for 3D data in medical appliances is envisioned for years.
The aim of this research is to develop a method that reliably and reproducability determine the range of motion of the digits. In current practice, the angles are measured using a goniometer. This method is very imprecise. Three methods to determine the location of joints in 3D hand scans can be distinguished: using heuristics, computer vision, and deep learning. Of those, deep learning is the most flexible, modern and accurate method and is therefore applied. The end result is a matrix containing the range of motion per joint and is applied to anatomically correctly manipulate a 3D model. For ease of manipulation, a physical manipulator is proposed. The results of this novel method show lower interrater differences than measurements with a goniometer.