Modeling of local structure with a dictionary of image patches
- applied to the study of Lung ECM proteins
To reproduce the results from our paper (link to come), you will need to clone this repository or download it as a .zip file. Then, download the input_data and results_justDictionaries folders from DTU Data at https://doi.org/10.11583/DTU.16652065. Place both within the Lung_ECM folder, besides the code folder. Rename results_justDictionaries to results if you want to reproduce the exact results from our paper. Under the same DTU Data link you will also find the complete set of results under results_complete.
The following figure is a schematic view of the content in the results and input_data folders (at DTU Data) and the code in this repository. The table below describes further the content of the code folder found in this repository.
TO DO: ADD schematic
Code contains:
| File name | Description | Input | Output |
|---|---|---|---|
| lung_lif_to_png.py | Transforms Leica microscope format into a png | .lif file pr. sample | input_data/raw |
| microscopy_analysis.py | Helper functions | ||
| image_preprocessing.py | Equalises contrast across images and channels, and computes max. projections | input_data/raw | input_data/maximum_projections and postprocessed_maxProjs |
| lung_feature_patch_allPatients_singleProtein.py | Single-protein analysis | postprocessed max. projs. | results/..._single |
| lung_feature_patch_allPatients_twoProtein.py | Two-protein analysis | postprocessed max. projs. | results/..._dropout |
| lung_feature_patch_allPatients_threeProtein.py | Three-protein analysis | postprocessed max. projs. | results/..._triple |
Code developed by Monica J. Emerson monj@dtu.dk with contributions from Anders B. Dahl abda@dtu.dk.