VICOROB DBT Challenge participation

Participation on the DBT Challenge http://spie-aapm-nci-dair.westus2.cloudapp.azure.com/competitions/4 from the VICOROB research group (University of Girona, Spain).

Contact: Robert Marti, robert.marti@udg.edu

Using functions from duke-dbt-data

https://github.com/MaciejMazurowski/duke-dbt-data

Files

• pre_process_rm.ipynb

Pre-processing of DBT images to generate 2D slices and bounding boxes to be trained in detectron

• DBT_detectron.ipynb

Implementation of detectron training with the OMIDB and DBT dataset

• inference_DBT.ipynb

Inference of new images (validation provided in the web page)

From dubke-dbt-data

read dicom image

dcmread_image.ipynb notebook shows how to read image data from a DICOM file in the coordinate system that maches the ground truth bounding boxes.

draw bounding box

draw_box.ipynb notebook shows how to draw a bounding box on the image.

helper functions

To use helper functions from the notebooks, simply copy the duke_dbt_data.py file to your project.