Inquiry about BraTS2021 Implementation Strategy in Medical SAM 2

[ObliviateHe]

Dear author

Could you please provide guidance on the correct implementation approach for BraTS2021, or share the specific code/configuration used in your experiments? This would greatly help the research community reproduce your excellent results. Specific questions are as follows:

• Did you use multi-target training (training ET, TC, and WT regions separately) or single-target training (training only whole tumor)?
• Did you generate one bounding box per slice for the whole tumor, or multiple bounding boxes (one for each region: ET, TC, WT)?
• During training, did you use the original BraTS2021 labels (0,1,2,4) directly, or did you convert them to binary masks for each region?

Best Will