Welcome to the second week of Cinemax,

the course on Introduction to Advanced Tomography.

This week, we are going to discuss experimental cases that we are working with.

We will start on reconstruction of 3D volumes,

the methodology behind it,

the mathematics behind it.

Also, we will discuss the computational resources that are required in

order to analyze and process tomography data.

They can be quite computationally heavy,

and we will discuss various strategies for handling these large volumes of data.

That also means that in this week,

we are going to work with the actual data representing the experimental cases.

Cases we have two off,

one of them is composite materials that are used for wind turbine blades.

Composite materials, they consist of various types of

fibers that are glued together with the resin.

The mechanical properties of the materials are

closely linked to the structure of these composite materials.

So, what we need is to establish a 3D model of the structure in

order to be able to model and predict the properties of the material.

The other case is chalk,

natural North Sea Chalk,

which is not a technological material,

but a natural material where we are interested in

how fluids that can be either liquids or gases,

how they interact with the porosity in the material.

In both cases, both the composite materials and the chalk material case,

we need a three-dimensional model so that we

can simulate how the material will react for example to mechanical stress,

and how it will react to pressure of liquid,

for example, passing through the material at a certain flow rate.

To get to such a 3D model,

we need to follow a number of steps

from the data acquisition to the final three-dimensional model.

This workflow we call it or pipeline you may also refer to it as,

is here presented by Vedrana Andersen Dahl,

who is an associate professor at DTU Compute,

the Technical University of Denmark.

A typical analysis pipeline involving images and volumes,

involves a number of steps and spans over a longer period of time,

and it also involves a number of experts.

Usually, we start by a sample, which is scanned.

The data is then reconstructed so that we have a volumetric data,

and then this volumetric data often needs to be segmented.

Also, very often, we are interested in obtaining

a mesh from this segmentation in order to perform some modelling.

All of this is because we want to have

some results about the samples that we have scanned,

so that we can get some insight in how

this object is working or how this object is what is it made of.

So, when we have a sample,

for example from the North Sea Chalk where we

may have core drillings taken up from deep underneath the water,

what we will do is that we will acquire data from that sample.

That is, as we have talked about before in

the first week where we make these projections,

the shadow images so to speak of the sample.

What we need to do then is go through this pipeline that Vedrana just talked about,

where from the recorded projections

generate a three-dimensional volume by 3D reconstruction.

This is the tomography operation you could see.

This is what we will learn more about in the coming lectures.

As we go from the 3D reconstructed volume to the next step,

we will do segmentation as Vedrana also mentioned.

This is this process of taking our now 3D representation of the values say,

contrasts, measures in terms of gray levels,

distributed all this grid in the three dimensions.

We now take a process where we group these volumes,

soft volumes of our 3D volume,

so to speak in labeled domains.

This is the process we call segmentation,

and then we are still in three-dimensions,

but we have now reduced the level of

complexity by grouping into certain colors you could say,

or labels, or identities.