Adding measurements functionalities to inference results

Is your feature request related to a problem? Please describe.
Radiologists need to see quantifiable results of the AI segmentation inference. Getting the mask is great but I am looking for higher level measurements as volumes, Tumor count. If you take a look at this site showing multiple AI models, you will see reports showing key measurements that is useful for radiologist.

Describe the solution you'd like
Lets assume my AI infer can detect kidneys, kidney tumor, liver and liver tumor, I would like a set of post processing functions that calculates measurements below:

1. count of liver tumors, kidney tumors
2. For each tumor what is the volume in mm^3, this is simply pixel count x voxel size
3. Key png image for each tumor. find slice with largest tumor count, then get slices above and below to show tumor location. We should also have the tumor as a burnt contour for each slice.
4. Axial key image is usually sufficient but for Aorta we do need coronal and sagittal key images
5. Average tumor intensity and standard deviation.

List above should be a good start. we could later investigate using packages as pyradiomics for more 1st and 2nd order measurements. For cardiac and vessels we could add vessel length, thickness, diameter, etc.

Once we have this in place we can then create a report for each inference that compiles all this data to a useful report for clinicians. Please take a look at this sample reports to get an idea:

• Brain with tumor https://samples-viewer.one.deepc.pro/?studyUid=CINA_ICH_EU_1&seriesUid=

• Fancy aurta report https://samples-viewer.one.deepc.pro/?studyUid=0a346172-49b2-4686-a19a-18debaeaf7fc&seriesUid=e0bb7132-94df-40eb-beb5-a4d4f2348e30

Describe alternatives you've considered
I have written some post processing for monai label that does this on save to the datastore.

I would be more that happy to share the code

Start by reviewing MONAI's existing inference and post-processing paths, then compare them with the requested measurements: lesion counts, volumes, key images, contours, and intensity statistics. Done would require an agreed scope and a tested implementation that produces these measurements and supports the requested report inputs.