Segmentation

[mohammad maskani]

Hi 

I have problem with segmentation of Tumor from Brain CT scan images ,

Which method is better for Radiomics ? single-slice based (2D) or Segment tumor from all slice (3D) 

[Joost van Griethuysen]

There have been some studies here and there investigating the influence of the type of segmentation on the performance of the models, but I don't know of a specific one for brain.

Basically, the best (reproducible) result was obtained with whole volume, followed by single slice. Worst was the circular ROI on a single slice.

This may be due to the fact that readers will not always select the same slice as "largest" slice, and there is the heterogeneity of texture over the entire region to consider.

That said, there are also several studies that only have single slice segmentations, and achieve some performance. The main upside of single-slice is that it takes less time to do so manually.

PyRadiomics supports both single-slice as whole volume segmentations, though in case of single-slice, you're better off using shape2D (disabled by default) instead of shape (3D, enabled by default).

Regards,

Joost

Op vrijdag 1 november 2019 08:43:26 UTC+1 schreef mohammad maskani: