The effectiveness of multi resolution image segmentation for measuring spatial heterogeneity in mixed population biofilms

Multi resolution image clustering and segmentation tools has been developed to measure the distance between clusters of homogeneous microbial populations within two-dimensional sections of biofilms visualized by confocal laser scanning microscopy. The concept underlying multi resolution image segmentation is that the number of clusters are larger for higher resolution image and smaller for lower resolution image. This hierarchical structure analysis can be used to simplify the problem in well-mixed populations. The algorithm combines fuzzy C-means, SOM and LVQ neural networks to segment and identify clusters. The outcome of the segmentation is quantified by the number of clusters of each kind of microorganism within sections of the biofilm, and the centroid distances between the identified clusters. Experimental evaluations of the algorithm showed its effectiveness in analyzing biofilm mixed populations.