Improving skeleton algorithm for helping Caenorhabditis elegans trackers

Abstract

One of the main problems when monitoring Caenorhabditis elegans nematodes (C. elegans) is tracking their poses by automatic computer vision systems. This is a challenge given the marked flexibility that their bodies present and the different poses that can be performed during their behaviour individually, which become even more complicated when worms aggregate with others while moving. This work proposes a simple solution by combining some computer vision techniques to help to determine certain worm poses and to identify each one during aggregation or in coiled shapes. This new method is based on the distance transformation function to obtain better worm skeletons. Experiments were performed with 205 plates, each with 10, 15, 30, 60 or 100 worms, which totals 100,000 worm poses approximately. A comparison of the proposed method was made to a classic skeletonisation method to find that 2196 problematic poses had improved by between 22% and 1% on average in the pose predictions of each worm.

Caenorhabditis elegans Multi-Tracker Based on a Modified Skeleton Algorithm

“…Tracking of each worm started from image k + 1 to image 30 and from image k − 1 to the first. The skeletonization method proposed in previous work [ 35 ] was used in each image. This method used distance transformation [ 39 ] to obtain possible worm skeletons, and through of an optimization method using different criteria found the best skeleton prediction.…”

“…The skeletonization method proposed in the previous work [ 35 ], unlike classical methods, enables the separation of aggregated worms ( Figure 3 a), creating new paths in the skeleton, and some possible solutions for each worm ( Figure 3 b–e). Maximum and minimum values of the width vector are used in the distance to transform images of each segmentation to find this new skeleton as mentioned in [ 35 ].…”

“…The skeletonization method proposed in the previous work [ 35 ], unlike classical methods, enables the separation of aggregated worms ( Figure 3 a), creating new paths in the skeleton, and some possible solutions for each worm ( Figure 3 b–e). Maximum and minimum values of the width vector are used in the distance to transform images of each segmentation to find this new skeleton as mentioned in [ 35 ]. The possible solution skeletons are obtained from a recursive function, which runs through ends and branch points of the new skeleton that overlap the previous segmentation of the worm’s body (red circle in Figure 3 b–e).…”

“…As its name indicates, it is obtained by dividing the total area of the intersection by the union of the elements [ 41 ]. For the evaluation, the area of the reconstructed bodies of the manually labeled skeletons was used, skeletons using the skeletonization method proposed in [ 35 ] and the classical skeletonization method using the Matlab bwmorph command. …”

“…The aforementioned methods were tested by our team and results indicated that previous segmentation is one of the most significant criteria to help identify worms within an aggregation. This was taken as a starting point to design different trackers, adding and combining numerous criteria, and it was found that the criteria of completeness and color used with the improvement of the form of skeletonizing [ 35 ] can help to identity the worms in the next image. Furthermore, we introduced three new criteria: length, smoothness, and noise, revealing that they can be useful in re-identification.…”