These are general questions that are not specific to MRIcron. I would suggest you attend a FSL, SPM, FreeSurfer, etc workshop to learn more. You may also want to check out the powerpoints, demos, web pages and tutorials for my Image to Inference class.
  

1.) I have never of '.imwirte' or 'imread' - do you refer to the Matlab functions imread and imwrite or the Python SicPy functions imread and imwrite? Regardless, as I stated earlier I would only use PNG output for publication, as it downsamples your data to 8 bits (while MRI and CT typically have 16-bit precision). For Matlab you would use SPM's spm_vol and spm_read_vols to read the NIfTI data directly
 hdr = spm_vol('myimg.nii');
 img = spm_read_vols(hdr);
for Python you could use NiftiLib.

2.) Typically images outside the brain are removed when you normalize your data. This is an automated process, which will reslice your data to standard space. 

3.) The initial coordinates shown when you load an image in my software is the origin of the image as determined by the header S/QForm. For raw data this is the isocenter (MRI) or table center (CT), for normalized brains this should be near the anterior commissure. You can change the S/QForm to set this as you wish.