retinalysis-fundusprep 0.1.0

Easy, robust CFI bounds detection and contrast enhancement

Retinalysis fundus preprocessing

Fundus bounds extraction, cropping and contrast enhancement

Basic usage: running from the command line

We include command line utilities for running fundus preprocessing. Two commands:

• preprocess-folder use for running on a folder with input RGB images. Will not recurse into children of the input folder:

  fundusprep preprocess-folder <data_path> [OPTIONS]
  

• preprocess-csv use for more advanced usage to provide arbitrary filepaths and specific filenames (or IDs) for the outputs. Provide an input CSV file with columns path (for filepath to an RGB file) and id (optional) to name/identify the outputs

  fundusprep preprocess-csv ./image_list.csv \
  --rgb_path ./processed_rgb \
  --ce_path ./contrast_enhanced \
  --bounds_path ./metadata/bounds.csv
  

Options

Both commands share the same options:

• --rgb_path PATH: Directory where processed RGB images will be saved
• --ce_path PATH: Directory where contrast-enhanced images will be saved
• --bounds_path PATH: Path to save a CSV file containing image bounds information
• --n_jobs INTEGER: Number of parallel processing workers (default: 4)

Notes

• All output paths are optional - files will only be written when the corresponding path is provided
• Missing image files will be reported but won't stop the processing of other images
• The bounds CSV contains information about how images were cropped for standardization
• All output images are saved in PNG format with the same filename as the input image.

Examples

Processing Folder with RGB Images

To process a folder of fundus images and save only the RGB versions along with the bounds information:

fundusprep preprocess-folder ./original_images \
  --rgb_path ./processed_rgb \
  --bounds_path ./metadata/bounds.csv

Processing with Contrast Enhancement

To process images with both RGB and contrast enhancement:

fundusprep preprocess-folder ./original_images \
  --rgb_path ./processed_rgb \
  --ce_path ./contrast_enhanced \
  --bounds_path ./metadata/bounds.csv

Processing Images Listed in a CSV (No Custom IDs)

Example CSV:

path
/data/images/patient1.jpg
/data/images/patient2.jpg
/data/images/patient3.png

To process images listed in a CSV file:

```bash
fundusprep preprocess-csv ./image_list.csv \
  --rgb_path ./processed_rgb \
  --ce_path ./contrast_enhanced \
  --bounds_path ./metadata/bounds.csv

The outputs will use the same filenames as the input images. For example, the RGB output for /data/images/patient2.jpg will be ./preprocessed_rgb/patient2.png. Note that all outputs will be stored in a single folder, and therefore filenames should be unique. If filenames are not unique, use custom image IDs.

Using Custom Image IDs

The CSV file must include:

• A path column with absolute or relative paths to the image files
• an id column to specify custom identifiers for each image

Example CSV:

path,id
/data/images/patient1.jpg,P1_left
/data/images/patient2.jpg,P2_right

Processing is done in the same way:

fundusprep preprocess-csv ./patient_images.csv \
  --rgb_path ./processed_rgb \
  --ce_path ./contrast_enhanced \
  --bounds_path ./metadata/bounds.csv

The RGB output for /data/images/patient2.jpg will be ./preprocessed_rgb/P2_right.png.