I am wondering what the difference between pixel based ML and structure based ML is, in the context of Medical image segmentation (of CTA images). I have read that structure based ML will make classifications of objects based on their input features (contrast, area...). And that pixel based ML will rely on directly using pixels/ voxels as input information instead of structures.
But I am wondering, where the difference is then? as structure based ML in my opinion would also need to gain understanding of the structure by analyzing the input pixels, so I don't really get where they differentiate? So my question is, where the difference is between pixel based, where pixels as input are used and structure based, where a structure made out of pixels is used as input?
