Neuron analysis with Aivia

Imaging techniques like electron microscopy (EM) and lightsheet microscopy allow us to see whole neural network in very fine detail. However, the task of extracting information from these datasets is often time-consuming and tedious to perform manually. With Aivia, we have developed a number of tools to automate the analysis process and extract key information.

__________________________________________________________________________________

“I’ve tested several open-source and commercial solutions for my neuron analysis, but after accounting for all pros and cons Aivia was the best solution. It is stable, can handle big datasets and is flexible. I can auto-trace then switch to manual-edit mode if necessary. Another bonus is that Aivia values my ideas for future improvements.”

Laura Korobkova, USC Laboratory of Neuro Imaging

__________________________________________________________________________________

Since the 3D Neuron Analysis recipe was launched in 2017, we have continually improved the way neuron tracing is accomplished in Aivia. You can trace thousands of neurons and detect spines using Aivia's automated 3D Neuron Analysis recipe. The introduction of ROI processing in 2019 allows you to isolate specific volumes to apply a recipe, crucial for datasets with multiple neuron types or large volumes.

For users who have smaller datasets and desire higher accuracy, Aivia provide intuitive tracing tools in the Neuron Composer. You can trace dendrites completely manually or semi-manually using the Predict Dendrite tool. With this tool you select a starting point along the dendrite, Aivia automatically predicts (and connects) the possible tracing paths. As you validate the predicted segment, Aivia will predict the next section - allowing you to trace the dendrite more quickly. You can also click to identify the next point along the dendrite and let Aivia figure out the path.

See the manual tools in action

Read about the Neuron Composer on our poster presented at ASCB 2019

Traditionally, EM datasets are annotated manually due to the lack of automated tools and approach to detect and segment objects. In 2017, Aivia became the first commercial software to deploy deep learning for automatic segmentation in 3D EM images. Since then, we have further expanded the deep learning capability in Aivia so you can augment our pre-trained model or create your own model to match your image/experimental needs.

The Mesh Contour Editor in Aivia allows you to manually annotate the 3D EM volume in case adjustments are needed for the automatic results. This can also be used to create ground truths for the creation of your own deep learning model.

Read more about how the deep learning training process works on the cloud (now also available using local resources).

As new, powerful imaging platform emerges, the quantity and size of data grow correspondingly. Large, multi-terabyte image volumes composed of whole organs imaged at a single cell resolution is no longer the province of a handful of advanced imaging cores. We have developed an optimized pipeline for viewing these large datasets. By breaking the image down to multiple image blocks with varying levels of resolution, Aivia is capable of rendering 2TB, single time-point, 3D volumes using a single consumer-grade GPU (Nvidia RTX 2080Ti with 11GB of VRAM).

The neuroscience field continues to grow as new, high resolution imaging platforms enable researchers to visualize the connectome in ever-finer details. As the resolution gets better, image visualization and analysis tools must also evolve to tackle the challenges of these datasets.

Through collaborative R&D projects with leading researchers, we are working to address these challenges. Our current projects include:

Whether you are looking for neuron analysis for your work today or looking towards further work, Aivia has the tools to help you advance your research.

Visit www.aivia-software.com/demo to request a demo license.