DeepSTEMI

A Novel Deep Learning System for STEMI Prognostic Prediction from Multi-Sequence Cardiac Magnetic Resonance

User Guide

Prediction Objective:

    2-Year MACE Event Probability Prediction from Multi-Sequence Cardiac Magnetic Resonance

File Format Requirements:

    Image: JPG(.jpg)/PNG(.png). DICOM/NII images need to be converted into JPG/PNG format before being uploaded.
    Table: Excel file, xls/xlsx

File Preparation:

  • Make a parent-folder with 4 required child-folders (all in lower case):
    • cine : images from the CINE sequence (For the cine images, group 30 frames as one slice.)
    • lge : images from the LGE sequence
    • t2 : images from the T2 sequence
    • tab : table of case information (Excel .xlsx file), including 5 required columns:
      1. name
      2. age
      3. sex : 0 for female, 1 for male
      4. grace
      5. killip
  • The filenames of the images should be numbered sequentially, such as
    • 093-img-00012-00241.jpg
    • 093-img-00012-00242.jpg
    • 093-img-00012-00243.jpg
    • 093-img-00012-00244.jpg
    • 093-img-00012-00245.jpg
  • Compress the parent-folder into a zip file before uploading.
  • Here is an example for demonstration:
    1. Download the demo1.zip.
    2. If the default name of downloaded file is _static_gbdt_file_demo1.zip, unzip the download file.
    3. You will get a demo1 directory with 4 sub-directories metioned above.
    4. Zip the demo1 directory into a zip file like demo1.zip before upload it.
    5. Please upload the zip file and perform subsequent operations as instructed below.

Click here to upload the zip file