Neurite tracing in fluorescence microscopy images using ridge filtering and graph searching: principles and validation

To assist neurobiologists investigating the molecular mechanisms involved in neurite formation and differentiation, we have developed an interactive technique for the tracing and quantification of elongated image structures. The technique is based on an improved steerable filter for computing local ridge strength and orientation. It also uses a graph-searching algorithm with a novel cost function exploiting these image features to obtain globally optimal tracings between user-defined control points. To compare the performance of the technique to that of the currently used approach of fully manual delineation, four observers traced selected neurites in fluorescence microscopy images of cells in culture, using both methods. The results indicated that the proposed technique yields comparable accuracy in measuring neurite length, significantly improved accuracy in neurite centerline extraction, significantly improved reproducibility and reduced user interaction.