ambientmemory / bloodcelldb

Visual Databases project that uses machine learning to identify whether images of blood slides contain acute erythroid leukemia or not.

bloodcelldb

Visual Databases project that uses machine learning to identify whether images of blood slides contain acute erythroid leukemia or not.

Training Set Creation:

Data sourced from: http://imagebank.hematology.org/ as well as http://library.med.utah.edu/WebPath/HEMEHTML/HEMEIDX.html#6

• How to create/use this dataset:
  • First, we run a resizing script that forces each image into a 300x300 thumbnail (consistency in processing)
  • The infected images and the healthy images are spearately resized and labeled with either 'L' or 'N' headings (refer to img_proc_leuk.py for details)
  • Some healthy and some infected cells are selected from the pool and then a renaming/label-creating script is run to randomize their names as well as assign a 0-1 label for further supervised learning.

TODO:

• UI:
  • Test with real response from server
• Server:
  • Call Python on the uploaded files by passing them via the command line
  • Test writing to SQL
• ML Proc:
  • Test