Hyperspectral Image Data for Diagnosis of Endometriosis

This research focuses on the question whether the use of hyperspectral imaging sensors, in comparison to conventional RGB sensors, offers advantages to assist surgeons during minimally invasive procedures and to increase therapy quality by preventing clinical recurrences in laparoscopic endometriosis diagnostics. Since hyperspectral data for medical applications are not publicly available, the acquisition of suitable data is essential for the realization of this research. Therefore, the method of data collection is addressed in this work. Additionally, results based on the current data collection using Principal Component Analysis (PCA) for dimensionality reduction and Self-Organizing-Map (SOM) for clustering are presented. In this way, spectral differences can be extracted from the data that appear to be visually similar.