Hey,
I understand that we get sub-volumes of the MR data, which are in DICOM format, and later, how do we actually get the candidate patch, and how do we get an image representing the enhanced tumor part of the label? Is it already given in the label? I understand the later part of the sections, like all the Soft Dice Loss and all.
I have one more lingering doubt(or a small confirmation) that is in the Diagnosis (Identification of illness), here, how do we identify an illness like an enhancing tumor, stroke, etc. (Is it similar to the Supervised learning where all the images have labels and then it gets trained, and when a new input is given, it detects). If possible, can you please explain the workflow like a kind of flowchart?
ex: {1. Data taken(Image,Label)
2. Patches Made (Sub Volumes)
3. Individual Patches Trained
and so on.}
Also, can anyone give me a real-life, relatable explanation for using standardization patches and how the Expected Output is actually generated in the C1W3_Assignment, where it has 2 sections, one is our input image, beside that, we also have a detected patch.
Sorry if my ambiguity sounds very silly, but I wanted to know the process in short, crisp steps.