Visualize Terabyte Datasets
Light sheet microscopy allows researchers to capture large volume of data with minimal damage to the specimens. These datasets can often be hundreds of gigabytes and even up to multiple-terabytes for a 3D time-lapse. The size of these datasets can exceed even the most powerful GPUs and computers - making visualization of the whole volume very difficult.
In Aivia 6, we are introducing a new rendering pipeline that allows you to interactively review multi-GB/TB volumes on standard computing hardware. We have developed an innovative multi-resolution, multi-block architecture with efficient memory management which work in synchrony to enable you to explore your data in real-time. Visualize and play with your data without interruptions as Aivia swaps in progressively higher resolution volume into your field of view.
Deep Learning Segmentation
Comprehensive 3DEM Segmentation
Segmentation of 3D electron microscopy (3DEM) datasets can be a time-consuming endeavor for many researchers. To create a reconstruction of a single cell, the scientist must go between tens to hundreds of image slices and manually draw the boundary of a given cell on each slice.
In deep learning, the computer takes in pre-labeled training data to generate a simple set of rules and passes the information to the next level for rule-making. The rules from various levels of the hierarchy are then combined to form the final classification outcome. This method enables the creation of complex classifications with few samples.
In Aivia 6, we implemented a deep learning (U-net with densely connected blocks) and trained it for the segmentation of brain samples imaged by electron microscopy (3D). Our pre-trained model lets you fully segment a 1kx1kx100 pixel block in 25 mins - no fancy hardware and coding knowledge needed. Manual segmentation of such a benchmark data set would take days.
Additional Cell Phenotyping
Expanding upon our 3D neuron and object classifiers, Aivia 6 introduces 2D object classification and novelty detection.
The new novelty detection option in Aivia 6 augments the machine learning classifier by looking for outliers in the classified objects. Novelty score, a new measurement provided by the Classifier tool, tells you the likelihood of an object is unique relative to the training data. This enables discovery of hidden phenotypes and new insights to your analysis results.
3D Object Analysis
Better Than Ever
We have rebuilt the 3D Cell Count (now Object Analysis) recipe from the ground up to give you more processing options and better analysis results. In Aivia 6, a new watershed function is incorporated into the processing pipeline to improve partition accuracy; while a new surface smoothing function provides you better-looking surface output.
Furthermore, we have made the recipe more user-friendly by adding a preview option for detection. The recipe generates analysis results in 3D (as surfaces) and 2D (as cross sections) to give you more visualization options.
In addition to the major features above, Aivia 6 brings tens of new improvements to help make the overall Aivia experience better. You will find more feature-rich functions, better data visualization and significantly faster performance in Aivia 6. Here are just some of the improvements you will find:
Object outlines (replacing bounding boxes) for better aesthetics for visualization
Object groups for adjusting display of multiple output
Improved 3D Neuron Analysis recipe
Faster relational measurements between associated object sets
Measurements calculated in the background