Automatic reconstruction of 3D neuron structures using a graph-augmented deformable model

Abstract

Motivation: Digital reconstruction of 3D neuron structures is an important step toward reverse engineering the wiring and functions of a brain. However, despite a number of existing studies, this task is still challenging, especially when a 3D microscopic image has low single-to-noise ratio and discontinued segments of neurite patterns.Results: We developed a graph-augmented deformable model (GD) to reconstruct (trace) the 3D structure of a neuron when it has a broken structure and/or fuzzy boundary. We formulated a variational problem using the geodesic shortest path, which is defined as a combination of Euclidean distance, exponent of inverse intensity of pixels along the path and closeness to local centers of image intensity distribution. We solved it in two steps. We first used a shortest path graph algorithm to guarantee that we find the global optimal solution of this step. Then we optimized a discrete deformable curve model to achieve visually more satisfactory reconstructions. Within our framework, it is also easy to define an optional prior curve that reflects the domain knowledge of a user. We investigated the performance of our method using a number of challenging 3D neuronal image datasets of different model organisms including fruit fly, Caenorhabditis elegans, and mouse. In our experiments, the GD method outperformed several comparison methods in reconstruction accuracy, consistency, robustness and speed. We further used GD in two real applications, namely cataloging neurite morphology of fruit fly to build a 3D ‘standard’ digital neurite atlas, and estimating the synaptic bouton density along the axons for a mouse brain.Availability: The software is provided as part of the V3D-Neuron 1.0 package freely available at http://penglab.janelia.org/proj/v3dContact: pengh@janelia.hhmi.org

A Broadly Applicable 3-D Neuron Tracing Method Based on Open-Curve Snake

“…The radius with a maximum of this medialness measure is used as an estimate of the local radius. Peng et al (2010a) estimated the radius by fitting a radius-adjustable sphere centered at the point of skeleton trace into local image evidence. In this paper, we employ a boundariness measure (Pock et al 2005) and estimate the radius to be the one maximizing this measure.…”

“…Broadly speaking, pre-DIADEM neuronal reconstruction algorithms can be categorized into following types: (i) Sequential tracing (Al-Kofahi et al 2002, Aylward andBullitt 2002); (ii) Skeletonization (Cohen et al 1994, He et al 2003; (iii) Minimal Cost Path (Meijering et al 2003, Peng et al 2010a); (iv) Minimum Spanning Tree (MST) (Yuan et al 2009, González et al 2010, Xie et al 2010); (v) Active Contour based tracing methods (Schmitt et al 2004, Vasilkoski and Stepanyants 2009); (vi) special-purpose algorithms for tracing Neuromuscular Projection Fibers (Cai et al 2006, Srinivasan et al 2007, Wang et al 2007, Cai et al 2008, Lu et al (2009, Srinivasan et al 2010).…”

“…Minimal path based tracing approaches (Meijering et al 2003;Peng et al 2010a) are scalable because they limit computation to user-defined image subregions, rather than to the entire image. They usually define a cost or "energy" measure based on local image features and curve regularity.…”

“…The optimal path is found by the Dijkstra's shortestpath algorithms (Dijkstra 1959). Peng et al (2010a) designed a geodesic metric function measuring medialness for each voxel. After a graph is constructed with the image voxels as the vertexes and the edges among vertexes having weight defined with the geodesic metric function, the centerline is estimated by solving the shortest path problem on the graph using the Dijkstra algorithm.…”

SmartScope2: Simultaneous Imaging and Reconstruction of Neuronal Morphology

“…First, for three isolated EGFPexpressing neurons in mouse (Supplemental Table S1, Scans 1-3), we compared S2 to manual reconstructions generated by three annotators who traced the neuron structure manually using Vaa3D-Terafly 15 . We compared the similarity between the S2 reconstruction and the manual reconstructions as well as the similarity among manual reconstructions, using the best average spatial distance score d 16 (Supplemental Text and Table S2). Overall, the average bi-directional distance scores between S2 and three manual reconstructions are very close to the respective scores between three pairs of manual In scan 2, the high bi-directional score was actually due to a high average directional distance score from S2 to manual (13.42 voxels vs 3.83 voxels) on average, indicating that in that particular case, S2…”