This repo contains the code needed to perform the emphysema experiment. The title of the paper is 'Effect of emphysema on AI software performance for lung nodule detection compared to a human reader in low-dose chest CT'.
For this experiment, the file that needs to be executed is the reviewing_info_extract_emphysema
This will also execute patient_selection_emphysema_experiment. In order for this to work we need the following folders:
- 2 folders named emphysema_reviewed and no_emphysema_reviewed. The emphysema folder should contain 3 subfolders for each emphysema degree (advanced, confluent, moderate). Inside each of those subfolders, as well as in the main no_emphysema_reviewed folder, there should be subfolders named with the participant IDs that were reviewed by a consensus panel of radiologists. In each of those subfolders txt files with the results of that review should exist. An example of such a txt file can be found here
- A folder with the Excel files of the manual annotations/comparison between REDCap and AI (4 in total, one for each degree of emphysema).
- A folder with the CSV REDCap extractions.
These folders are not provided due to privacy regulations.
Pull requests are welcome. For major changes, please open an issue first to discuss what you would like to change.