Taipei-1 Usage Convert Scripts

1. Requirements

• pydicom
• nibabel
• pandas
• numpy

2. Using to convert *.dcm file to *.nii.gz file

• dcm2niix_HsinChu.py

  • This script is using to convert *.dcm file that come from HsinChu Hospital

  • You might notice that there is no extension name in the raw file, and we also can't read each file directly; fortunately, dcm2niix.exe is still working!

  • The dictionary from HsinChu be like:

    /mnt/HsinChu
    ├── batch1 <-- the data_root must indicate this location
    │   ├──S000001  <-- The dictionary at this layer is meaning a patient
    │   ├──S000010
    │   └──......
    │       
    └── batch2
        ├── 2023.01-2023.12.... <-- the data_root must indicate this location
        │   ├── S000001  <-- therefore, this meaining patient
        │   ├── S000010
        │   └── ......
        ├── 2024.01-2024.03.... <-- this one also indicate as data_root
        │   └── ......
        └── .....
    

    • The batch1 is already uploaded on doctor's dropbox.

  • The storage structure will same as original

    • ex: <the path for patient>/<series number>/<instance uid>/<cardiac phase>/<name generated by dcm2niix.exe>.nii.gz

  • Example:

    Follow the example dictionary structure, suppose we want to convert /mnt/HsinChu/batch2/2023.01-2023.12, and the target dictionary called /mnt/hsinchu_dst, the dcm2niix.exe place at /mnt/lib/dcm2niix.exe, then we simply named out, buf, err for output, buffer, error storage dictionary, and using 10 process, the command will look like:

    python dcm2niix_HsinChu.py --data_root=/mnt/HsinChu/batch2/2023.01-2023.12 --out_dir=/mnt/hsinchu_dst/out --buf_dir=/mnt/hsinchu_dst/buf --err_dir=/mnt/hsinchu_dst/err --num_workers=10 --dcm2niix=/mnt/lib/dcm2niix.exe
    

• dcm2niix_Taipei.py
  • This script is using to convert *.dcm file that come from Taipei Hospital
  • The dictionary from Taipei be like:
    • The 502-dicom:

      /mnt/usbB/'502-CT(dicom-format)  <-- That is for argument --data_root
      ├── 0001  <-- here are the patient
      ├── 0003
      └── ......
      /mnt/usbB/'CCTA Result' <-- That one should pass into --isp_root
      ├── 0001  # Almost like 502-CT
      ├── 0003
      └── ......
      

      • PS: Because in this part got some chaos
        1. I already convert around 4 / 5 of raw data(the *.dcm)
        2. But I just put around 300 patient on doctor's dropbox
        3. For ignore already processed part, the --meta_dir should point to the /mnt/usbA/502CT/meta
    • But, for 2500+ dicom, that got different:

        /mnt/usbC 
        ├── CGMH  <-- That dictionary is for argument --data_root
        │   ├── 1557.zip  <-- I suggest unzip after uploaded, send single file should faster than single dictionary.
        │   ├── 0003
        │   └── ......
        ├── CMUH  <-- That is for argument --data_root
        │   ├── 0001  <-- here are the patient
        │   ├── 0003
        │   └── ......
        ├── MMH
        │   └── ......
        ├── MMH
        │   └── ......
        ├── ......
        └── CT標註result  <-- That one is --isp_root
      

  • The progress
    • if the status show Skip ISP it represent the isp folder didn't contain any tissue labels.

3. Extract each coronary artery centerline coordinate and name, also plaque.

python store_each_plq.py ...

4. Merge the duplicate plaque label

python combine_multi_plque.py ... 

PS

• The dcm2niix program just download from dcm2niix