Image Processing and Analysis:
All images were reconstructed from 2D-slice datasets into 3D volumes,
processed and analyzed using a workflow coded in an in-house Mathematica
notebook. Figure 2 depicts the workflow developed. This workflow
includes:
- MRI-slice selection: In this step, slices outside the region of
interest are visually selected out. This allows maximizing useful
information to render 3D volumes.
- 3D rendering: The remaining slides from step 1 are used to render a 3D
image. An automated algorithm in Mathematica is used to this end.
- Cropping: Cartesian and cylindrical cropping methods are employed to
remove unwanted regions containing elevated levels of noise,
particularly from volumes with essentially no signal.
- Filtering: Sharpening, smoothing and Gaussian filters are applied to
prepare the 3D images for binarization and thereby clustering
analysis.
- Binarization and segmentation: A binary image is produced through the
selection of an adequate threshold value. This step can be conducted
automatically through available library functions in Mathematica.
However, in this work, we explored the threshold selection process by
comparison with alternate sources of data, e.g. gravimetric porosity
determination. More on this is discussed in the Results and Discussion
section.
- Cluster analysis. In this step, morphological components (different
pore systems that are isolated from each other) are clustered and the
overall porosity is calculated. The binary image is dependent on
choosing a threshold value that effectively represents the pore
structure without diluting the signal with low pixel values.