As part of the BoneSens research project, which aims to develop a novel digitized spacer system for continuous, longitudinal monitoring of infection status in vivo (multi-parametric data acquisition: temperature measurements; images of the inner joint skin and spectral evaluation of the bacterial load), this study investigates the potential of light scattering as a diagnostic tool using a random forest (RF) approach. Bacterial suspensions of different concentrations (e2, e4, e6, e7, e8 of Staphylococcus Aureus in Dulbecco’s Modified Eagle’s Medium) were illuminated by the radiation of a broadband light source to detect scattering patterns with an angle of 90 degrees. In the next step, calculations of amplitude, standard deviation, and AUC values form the basis for the classification of bacterial loads by an RF algorithm. Based on the graphical results of light scattering, bacterial loads may show differentiation with a lower limit of 104 CFU/ml. The calculated values enable a more sensitive classification of concentrations of 10² CFU/ml with an accuracy of up to 98%. Out of 200 measurements to be analyzed, only four were incorrectly assigned, whereby a differentiation is made between critical (bacterial load is underestimated; 1 out of 4) and non-critical errors (bacterial load is overestimated; 3 out of 4). |
|