Automatic Recovery of Z-Jumps for Neuronal Morphology Reconstruction
Institute of Automation, Chinese Academy of Sciences, China
Tufts University, Genesys Research Institute, USA
Many researchers share neuronal morphology files along with their publications or contribute the files as independent raw data to public repositories such as the NeuroMorpho project. These morphology files not only serve as important supplements for the original scientific contributions, but also are considered to be extraordinary sources for realistic neuronal morphology simulation. Since reconstructed model neurons are mostly traced manually from brain slices under a light microscopy using a computer tracing system such as Neurolucida system, topological errors are often created during the time-consuming reconstruction. In most cases, the error is a big jump along Z coordinates called ‘Z-Jump’ [Brown et al. 2011], which always happens between the initial point of one branch and the node point that connects two or more branches (an example is given in Figure 1). For instance, the data file corresponding to Neuromorpho.org ID:NMO_01056 appears normal at the view of X and Y coordinates (Figure 1(a), but many Z-Jumps become visible (marked as the yellow lines in Figure 1(b).) at the view of Y and Z coordinates. Obviously, such kind of morphology files cannot accurately represent the morphological properties of original neurons, and hence cannot directly be used for realistic neuronal morphology simulation. A correction of the error is necessary to recover the reconstruction files to normal morphologies of original neurons. For this need, we developed an automatic Z-Jump recover tool.
Since different researchers may be used to tracing neuronal branches with different distances between points in the morphology structure, a fixed threshold value to judge whether there is a Z-jump may not be appropriate. We proposed that the threshold value should be dynamic according to different reconstruction files. We rank the distances between two neighborhood sampled nodes. We assume 97.5% of the distances are in proper ranges (Let dn be the distance at the rank point of 97.5%). For the rest of the distances, if it equals to or is greater than 5 times of dn, then it is considered to be irregular distances (the red points in Figure 2(a) present the irregular distances in NMO_01056, and 14 distance values out of 3215 are irregular distances). For these irregular distances, we move the child node to the Z coordinate of the parent node, and the later connections along the same branch of the child node are moved accordingly. By using this method, the problem of Z-Jumps can be avoided (Figure 1(c) provides a neuronal morphology structure that is corrected from Figure 1(b) by the Z-Jumps recover tool developed in this study. Figure 2(b) and Figure 2(c) present the structure before and after recovery from another angle). We compare the identified Z-Jumps in NMO_01056 with the ones identified by the StdSwc tools developed by the NeuroMorpho project, 2 extra expert confirmed errors were identified (12 irregular distances were identified by StdSwc). If we further restrict the threshold to 3*dn, 10 extra Z-Jumps (with human judges) were identified.
Through automatic detections using the standards proposed above, we found that around 26.14% of the 10004 neuronal morphology files in Neuromorpho.org present the problem of Z-Jumps to varying degrees (covering 18 out of 22 species, except for C. Elegans, Turtle, Drosophila and Frog). All of the identified 2615 files were improved and reproduced using our developed method.
Brown, K.M., Barrionuevo, G., Canty, A.J., De Paola, V., Hirsch, J.A., Jefferis, G.S., Lu, J., Snippe, M., Sugihara, I., Ascoli, G.A. (2011). The DIADEM Data Sets: Representative Light Microscopy Images of Neuronal Morphology to Advance Automation of Digital Reconstructions. Neuroinformatics 9(2-3):143–157.
Neuronal morphology reconstruction,
recovery of morphology
Neuroinformatics 2014, Leiden, Netherlands, 25 Aug - 27 Aug, 2014.
Demo, to be considered for oral presentation
(2014). Automatic Recovery of Z-Jumps for Neuronal Morphology Reconstruction.
31 Mar 2014;
04 Jun 2014.
Dr. Yi Zeng, Institute of Automation, Chinese Academy of Sciences, Beijing, China, email@example.com
Dr. Yun Wang, Tufts University, Genesys Research Institute, Boston, USA, firstname.lastname@example.org