MiRmine provides normalized reads count as reads per million (RPM) for each miR. In this case I would divide all miRs into bins, e.g. lower 25 percentile - not expressed, 25-50 percentile - lowly expressed, 50-75 moderately expressed and >75 highly expressed in the tissue of interest. I took those numbers somewhat arbitrary. It would be better to plot the histograms per each tissue, get the feeling for their distribution and make a more educated guess.

If yo would like to perform analysis yourself on specific dataset, I recommend downloading expression of miRNAs in normal tissues from TCGA. There are processed matrices with normalized reads available (start here: https://portal.gdc.cancer.gov/projects).

Good luck!