A multimodal data acquisition and registration system for monitoring skin wounds: Wound modeling and accuracy estimation using phantoms.

As the population of developed countries ages, lifestyle diseases are becoming increasingly prevalent. One of the complications of these diseases is chronic wounds, which affect an increasing number of patients. The healing process of these wounds is lengthy, requiring a monitoring approach that is not only simple and repeatable but also provides accurate indicators of disease. In this paper, we introduce a comprehensive wound monitoring system-WoundScanning, comprising a scanner and an application. The scanner is equipped with three distinct cameras: photo, thermal, and depth. These cameras capture three different data modalities: an RGB (red, green, blue) photo, a thermal image, and a three-dimensional point cloud with an amplitude image. After being processed by the application, the data provides a reproducible, objective, and non-invasive method for determining parameters of the monitored wound, including its 2D surface, 3D surface, and volume. All data modalities are transformed into a single common coordinate system using a new calibration protocol. The average fiducial registration error is 0.216 pixels, demonstrating a high coregistration quality of our approach. Furthermore, our study provides a rigorous numerical verification of geometric parameter uncertainty using operator-independent phantoms, offering a systematic approach to evaluating measurement accuracy often lacking in commercial solutions. For a distance of 25 cm between the device and the wound, there are no statistically significant differences between the real and measured wound characteristics (p>0.05).