Medical image dataset bounding box in ROI

I have a large number of medical x_ray images as a dataset. I want detect a part of object and surround it by bounding box automatically without using image annotation tools to determinate the (x1 , y1) (x2 , y2). Can you please advise me.

template matching, maybe:

train a neural network.

you can make it predict boxes or segmentations. I’d suggest segmentations. just paint the dense parts of the bone.