RSNA 2023 Abdominal Trauma Detection Competition

RSNA 2023 Abdominal Trauma Detection Competition

---

Introduction

Hardware:

• CPU : 13th Gen Intel(R) Core(TM) i7-13700K 8 cores 24 threads
• memory : 64GB
• GPU : NVIDIA GeForce RTX 3090 GPU with 32 GB VRAM
• number of GPUs : 1

OS/platform : Windows 11

---

Prerequisites

• Clone the repository

• Setup the environment: pip install -r requirements.txt

• Download the data in the data folder:

  • download Competition data and change folder name to rsna-2023-abdominal-trauma-detection

    data/rsna-2023-abdominal-trauma-detection
    ├── image_level_labels.csv
    ├── sample_submission.csv
    ├── segmentations [206 entries exceeds filelimit, not opening dir]
    ├── test_dicom_tags.parquet
    ├── test_images
    ├── test_series_meta.csv
    ├── train.csv
    ├── train_dicom_tags.parquet
    ├── train_images
    └── train_series_meta.csv
    

Run The pipeline

Train model

1. Train 3D segmentation model

1. cd src/Segmentation : to run segmentation file

2. python train.py : to train segmentation model

3. python process_data.py : generate classification data based on segmentation model

   • segmentation model folder

     ├── results
     │   ├── models
     │   │   ├── segmentations
     │   │   │    └── train
     

   • segmentation logs folder

     ├── results
     │   ├── logs
     │   │   ├── segmentations
     │   │   │    └── train
     

   • data structure of segmented output will be :

     ├── results               
     │   ├── data
     │   │   ├── segmentation
     │   │   │   ├── data        
     │   │   │   │   ├── liver
     │   │   │   │   ├── kidney
     │   │   │   │   ├── spleen
     │   │   │   │   ├── bowel
     

2. Train Bowel Model

1. cd src/Classification_bowel : to run bowel classification file

2. python train.py : to train bowel model

   • bowel model folder

     ├── results
     │   ├── models
     │   │   ├── bowel
     │   │   │    └── train
     

   • train bowel logs folder

     ├── results
     │   ├── logs
     │   │   ├── bowel
     │   │   │    └── train
     

3. Train Kidney Model

1. cd src/Classification_kidney : to run kidney classification file

2. python train.py : to train kidney model

   • kidney model folder

     ├── results
     │   ├── models
     │   │   ├── kidney
     │   │   │    └── train
     

   • train bowel logs folder

     ├── results
     │   ├── logs
     │   │   ├── kidney
     │   │   │    └── train
     

4. Train Liver Model

1. cd src/Classification_liver : to run liver classification file

2. python train.py : to train liver model

   • liver model folder

     ├── results
     │   ├── models
     │   │   ├── liver
     │   │   │    └── train
     

   • liver logs folder

     ├── results
     │   ├── logs
     │   │   ├── liver
     │   │   │    └── train
     

5. Train Spleen Model

1. cd src/Classification_spleen : to run spleen classification file

2. python train.py : to train spleen model

   • spleen model folder

     ├── results
     │   ├── models
     │   │   ├── liver
     │   │   │    └── train
     

   • spleen logs folder

     ├── results
     │   ├── logs
     │   │   ├── liver
     │   │   │    └── train
     

6. Train all Model

1. cd bash : to run bash file
2. sh train_all.sh : to train all model