epierson9 / pain-disparities Goto Github PK

Hi, I'm trying to figure out this line of code and why it comes after the 'for' loop and not within the 'for' loop. Since it runs after the 'for' loop then it will only grab the last 'dataset_name' value from the 'for' loop (which does not seem correct).

• thank you

Howdy, thank you for sharing this amazing work. However, I do face difficulties in getting the data. I was not able to download data from NDA query tools, due to the NDA download manager issue ( I think that might be NDA's issues), and if I use OAICompleteData_ASCII, will that be sufficient to reproduce the paper experiments? Many thanks!

Hi @epierson9
I am looking through the image_processing.py code and the repository seems to be missing a 'detector' module as referenced in line 107 of image_processing.py.

from detector import KneeLocalizer

The 'KneeLocalizer' function but it does not appear in the any of the repository scripts and 'detector' does not appear to be an external package since it is not mentioned in 'requirements.txt'.

Hello! Truly excellent work.

I noticed that the code here is not licensed. I was wondering if it would be ok to use it in another open source (MIT licensed) project with citation and acknowledgement?

Thank you!

Hi @epierson9,
I'm working with your code once again for a project. I can't figure out the contents.csv file. It looks like the code is reading a pre-populated CSV file, but is this one you generate yourself or was found in an OAI archive? It sounds like an OAI provided file but it does not appear in my archives (which have probably changed considerably since you downloaded the OAI archives). I know you have this function call commented out but was just wondering.

• thank you

On a side note, since I am working with a set of radiograph images with DICOM headers that differ from OAI images, I am planning to create what is probably similar to the contents.csv file and would contain the radiograph metadata of images being processed. I will use this instead of reading the DICOM headers as done below so I can create an easier way to use radiograph metadata in the image processing script, regardless of what your DICOM headers look like. Bulk updating the headers of thousands of DICOM images seemed out of my reach.

Looking to recreate some of your team's findings. Is it possible to share the same patient ID splits that were used in these experiments? Thank you for your consideration.

Hi @epierson9, this is not an issue with your code, but I downloaded the latest OAICompleteData_ASCII from OAI and several of the MIF text files contained non-ASCII characters which caused errors running the code (specifically the code would throw UnicodeDecodeError: 'utf-8' codec can't decode byte 0x80 in position 15: invalid start byte errors within the non_image_data_processing.py script when trying to open the files). I found the non-ASCII characters by running:

cd <your non-images data directory>                                             # change to the 'non-images data directory'
LC_ALL=C find . -type f -exec grep -c -P -n "[^\x00-\x7F]" {} +       # from within the non-images data directory, list the files and number of non-ASCII characters in them (this assumes you only have the .txt files in this directory)
LC_ALL=C grep --color='auto' -P -n "[\x80-\xFF]" somefile.txt         # shows where non-ASCII characters are found in the file (NOTE: copy the results to an empty text file or somewhere to reference)

I thought I would share this since this will likely come up again for others unless OAI removes the non-ASCII characters from those files.