Abstract
The aim in high-resolution connectomics is to reconstruct complete neuronal connectivity in a tissue. Currently, the only technology capable of resolving the smallest neuronal processes is electron microscopy (EM). Thus, a common approach to network reconstruction is to perform (error-prone) automatic segmentation of EM images, followed by manual proofreading by experts to fix errors. We have developed an algorithm and software library to not only improve the accuracy of the initial automatic segmentation, but also point out the image coordinates where it is likely to have made errors. Our software, called gala (graph-based active learning of agglomeration), improves the state of the art in agglomerative image segmentation. It is implemented in Python and makes extensive use of the scientific Python stack (numpy, scipy, networkx, scikit-learn, scikit-image, and others). We present here the software architecture of the gala library, and discuss several designs that we consider would be generally useful for other segmentation packages. We also discuss the current limitations of the gala library and how we intend to address them.
Highlights
Connectomics, the elucidation of complete neuronal circuits, requires resolutions as low as 5–10 nm to distinguish the smallest neuronal processes, and fields of view hundreds of micrometers across or more, as neurons can span those distances
Since we have described the details of the GALA algorithm in detail elsewhere (Nunez-Iglesias et al, 2013), in this paper we focus on the design aspects of our implementation
We briefly present the Python application programming interface (API) and the command-line interface (CLI), before delving more deeply into useful design choices, and discussing the current limitations of the library and future directions
Summary
Connectomics, the elucidation of complete neuronal circuits, requires resolutions as low as 5–10 nm to distinguish the smallest neuronal processes, and fields of view hundreds of micrometers across or more, as neurons can span those distances. This size disparity results in large image volumes of at least 10 gigavoxels and often orders of magnitude larger. We developed a new machine learningbased algorithm for image segmentation (Nunez-Iglesias et al, 2013) that provides state of the art automatic segmentation accuracy and directs proofreaders to likely areas of error in the segmentation This has dramatically sped up proofreading and reconstruction speed (5–16-fold, in our anecdotal observations)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have