Meningioma

Goals

1. To automatically segment meningiomas pre- and post-surgery.
2. To assess brain's spatial recovery post-surgery.

• Data is available on Brainbox
• Ongoing literature search

Table of contents

1. Ground truth segmentation
2. Segmentation with slicer
3. ANTs brain extraction
4. t-SNE on synthetic brain

Ground truth segmentation

case	segmented	reviewed
001	✔
002	✔
003	✔

Segmentation with Slicer

An example of semi-automatic segmentation with Slicer's Segmentation Editor module.

<img src='/images/project_week/case_052_2_slicer_seg.png?raw=true', height=400>

ANTs brain extraction

We attempted to remove the skull with ANTs antsBrainExtraction.sh. Results were mixed. In some cases, the script performed well. In other cases, parts of the brain were removed, or parts of the skull (usually close to the meningioma) remained.

<img src='/images/project_week/case_052_ants_brain.png?raw=true', height=300>

Above: successful brain extraction.

<img src='/images/project_week/case_001_ants_brain_failure.png?raw=true', height=300>

Above: unsuccessful brain extraction.

t-SNE on synthetic brain

[source]

• blue == background
• green == skull
• red == brain
• cyan == tumor