Axonal transport velocity estimation from kymographs based on curvilinear feature extraction and spline fitting

Axonal transport velocities are obtained from spatio-temporal maps called kymographs developed from time-lapse confocal microscopy movies of neurons. The kymographs of axonal transport of C.elegans worms are much noisier due to in vivo nature of imaging. Existing methodologies for velocity measurement include laborious manual delineation of axonal movement ridges on the kymographs and thereby determining particle velocities from the slopes of ridges marked. Manual kymograph analysis is not only time consuming but also prone to human errors in marking the ridges. An automated algorithm to extract all the ridges and determine the velocities without significant manual efforts is highly preferred. Not many methods are currently available for such biological studies. We present an almost automated method based on information fusion using LDA classifier, morphological image processing and spline fitting for determining axonal transport velocities. Experimental analysis of 50 kymographs shows considerable reduction of 89% in time taken with manual intervention of 10.83%. Comparitive study with the results of two of the previous literatures shows that our algorithm performs better.