Alternate header for print version

Segmentation and Cellular reconstruction

Label data
Label data, also known as groundtruth, forms the baseline annotation and understanding in an image volumes. Structures are identified by their pixel features, such as agglomerations of protein that appear darker and of globular shape. These bio-structures are painted to be used to identify what structures are for the purposes of analysis and potentially to teach A.I.

Read more

CDeep3M makes deep learning algorithms available to the community and improve reproducibility. CDeep3M is a containerized tool, using deep learning for large-scale image segmentation tasks. It is an open source development and the software is free to use. You can run CDeep3M on your local platforms, on cloud providers, on GPU clusters or with free GPU resource.

Article about CDeep3M

Reconstruction data
Final forms of cells and their structures are converted into mesh- or voxel-based renderings to visualize the structures in an image, as well as the high-resolution details of what subcecullar structures are inside of them as well as their tissue-context and what other cells they interact with.

These data-rich scenes display the attributes of the tissue, such as how cells appear in the tissue, how the subcellular contents of the cells organize, and how the pathological structures could be interfering. Studying the distribution of paired-helical filaments (PHF) and β-amyloid plaques within the sophisticated organization of neurons, blood vessels, and brain structures can lead to profound insights regarding the progression of AD.

Scaling and Quantitative Analysis

Inference data
A trained deep learning model is asked to inspect an entire image volume and returns a map of values that indicate which pixels belong to which kind of bio-structure in the volume. This data can result in larger-scale analysis that can involve hundreds of cells and thousands of synapses.

Read more

Models & Quantitative Data
Labeled image data is reduced into quantitative factors, and paired with trained A.I. models, presents the exact analysis workflow that was applied to yield a given set of quantative data that can help understand the underlying AD mechanisms.

Markdown and notebook formatted files contatin detailed explanations of the tools used and how they are used. They can also execute on the locations where data is available and required resources.

Read more