Auto-detect coverage bounding boxes for brain MRI images project

[devdarsshan]

Hi all,
I'm Devdarsshan, currently in my final year pursuing a Master's in Data Science at Coimbatore Institute of Technology. I am interested in the project: Auto-detect coverage bounding boxes for brain MRI images. I have prior experience working with various ML models and image processing. Additionally, I've worked on a project before which used MRI images to detect brain tumors.

As stated in the project description, I used the Niffler png-extraction module to convert the DICOM data to png after downloading one of the publically accessible MRI images from the Cancer Imaging Archive (TCIA).
Before proceeding further, I have a few questions:

1. There are different angles of MR images of the brain, I would like to know whether the model should be robust to all different angles and identify the bounding boxes or if there is any specific angle(sagittal, axial, etc..) that is enough?
2. How many pre-defined protocol constraints are there?