BigNeuron: a resource to benchmark and predict performance of algorithms for automated tracing of neurons in light microscopy datasets

BigNeuron is an open community bench-testing platform with the goal of setting open standards for accurate and fast automatic neuron tracing. We gathered a diverse set of image volumes across several species that is representative of the data obtained in many neuroscience laboratories interested in neuron tracing. Here, we report generated gold standard manual annotations for a subset of the available imaging datasets and quantified tracing quality for 35 automatic tracing algorithms. The goal of generating such a hand-curated diverse dataset is to advance the development of tracing algorithms and enable generalizable benchmarking. Together with image quality features, we pooled the data in an interactive web application that enables users and developers to perform principal component analysis, t-distributed stochastic neighbor embedding, correlation and clustering, visualization of imaging and tracing data, and benchmarking of automatic tracing algorithms in user-defined data subsets. The image quality metrics explain most of the variance in the data, followed by neuromorphological features related to neuron size. We observed that diverse algorithms can provide complementary information to obtain accurate results and developed a method to iteratively combine methods and generate consensus reconstructions. The consensus trees obtained provide estimates of the neuron structure ground truth that typically outperform single algorithms in noisy datasets. However, specific algorithms may outperform the consensus tree strategy in specific imaging conditions. Finally, to aid users in predicting the most accurate automatic tracing results without manual annotations for comparison, we used support vector machine regression to predict reconstruction quality given an image volume and a set of automatic tracings.

Functional and anatomical specificity in a higher olfactory centre

Most sensory systems are organized into parallel neuronal pathways that process distinct aspects of incoming stimuli. In the insect olfactory system, second order projection neurons target both the mushroom body, required for learning, and the lateral horn (LH), proposed to mediate innate olfactory behavior. Mushroom body neurons form a sparse olfactory population code, which is not stereotyped across animals. In contrast, odor coding in the LH remains poorly understood. We combine genetic driver lines, anatomical and functional criteria to show that the Drosophila LH has ~1400 neurons and >165 cell types. Genetically labeled LHNs have stereotyped odor responses across animals and on average respond to three times more odors than single projection neurons. LHNs are better odor categorizers than projection neurons, likely due to stereotyped pooling of related inputs. Our results reveal some of the principles by which a higher processing area can extract innate behavioral significance from sensory stimuli.

Clonal analysis of olfaction in Drosophila: immunochemistry and imaging of fly brains

Clonal analysis with the MARCM (mosaic analysis with a repressible cell marker) system can be used for studying cell lineage, development, and anatomy in the Drosophila olfactory system and other parts of the fly brain. This protocol describes the dissection, staining, and imaging of brains from Drosophila with mosaic labeling. Staining for the presynaptic marker Bruchpilot (nc82) is performed in the example given here. The well-stained whole brain images that are obtained can be used to examine neuronal morphology. They are of sufficient quality to be used for image registration, which allows one to compare confocal images of labeled neurons in different brains.

Clonal analysis of olfaction in Drosophila: image registration

Clonal analysis with the MARCM (mosaic analysis with a repressible cell marker) system can be used for studying cell lineage, development, and anatomy in the Drosophila olfactory system and other parts of the fly brain. To compare confocal images of labeled neurons in different brains, it may be desirable to register them to a template or standard brain. There are various image registration approaches available. Some depend on manually specifying landmarks on the brains to be registered. Others depend only on the grayscale intensity value of one of the channels in the confocal image. Another important difference between registration approaches is whether they apply linear or nonlinear (warping) transformations. Linear transformations typically include translation, rotation, and scaling along each axis. Nonlinear transformations are much more computationally intensive, but are required to register brains with different shapes. Here we describe the practical steps required for an intensity-based nonlinear registration that has been used to map the higher olfactory centers of the Drosophila brain using the staining for the presynaptic marker Bruchpilot (nc82). This registration is in fact a two-step process. The first step is a linear transformation that roughly aligns the two brains, followed by a second nonlinear step that allows different parts of the brain to move in slightly different directions.

Clonal analysis of olfaction in Drosophila: generation of flies with mosaic labeling

Clonal analysis with the MARCM (mosaic analysis with a repressible cell marker) system can be used for studying cell lineage, development, and anatomy in the Drosophila olfactory system and other parts of the fly brain. This protocol gives a method for generating flies with mosaic labeling. It describes how to establish a mating cage for MARCM in PNs (projection neurons) of the fly antennal lobe and then select appropriate flies for dissection and staining using immunohistochemistry. The protocol can be adapted to determine the birth order of neuroblast lineages or individual cells. Alternatively, it can be used to dissect a complicated Gal4 line into its component neuroblast lineages to help elucidate projection patterns and connectivity. Collecting newly hatched larvae during a short time window allows for precise control of the stage during development at which the heat shock is applied.