Holes in the 3d segmentaion output of BRain tumor data

[Navee402]

Hi, when using openmap t1 for segmentation with Brain tumor data there are so many empty regions in the output. How to overcome this? Are there any special preprocessing tasks to deal with tumor datasets? It looks like the below figure

Thanks once again for this amazing work.

[kei-lab1106]

Thank you for bringing this to our attention and for using OpenMAP-T1 with your brain tumor dataset.

OpenMAP-T1 was trained using brain MRI from the ADNI2 dataset, which does not include brain tumors. As a result, the current model might not perform optimally on datasets containing brain tumors, leading to the partial disruption of the parcellation maps you've observed.

Addressing the challenge of segmenting brain images with tumors or other anomalies is a significant task for us. We are actively exploring the development of a brain tumor extraction model tailored for T1-weighted images. However, for now, there is not any preprocessing for brain tumor datasets.

Thank you again for your support and understanding.

[jhu-cnilab]

Dear Navee402,

Is it possible to share the de-identified Nifti file with us? We would like to use the image as a test bed to improve our algorithm.

Ken

[Navee402]

Hey jhu, sorry for the late response. I have tried to use OpenMapT1 on publicly available BraTS23 dataset.