ukk_mets

Introduction

This repository predicts the genotype and histology of brain metastases

Installation

Clone this repository: git clone https://github.com/rgutsche/ukk_mets.git

How to use it

Prerequisites

For Preprocessing

Data should be in the following structure:

└── PID_1
    ├── T1C
    │   ├── IM-0305-0001.dcm
    │   ├── IM-0305-0002.dcm
    │   └── IM-0305-0003.dcm
    ├── T2
    │   ├── IM-0309-0001.dcm
    │   ├── IM-0309-0002.dcm
    │   └── IM-0309-0003.dcm
    ├── FLAIR
    │   ├── IM-0307-0001.dcm
    │   ├── IM-0307-0002.dcm
    │   └── IM-0307-0003.dcm
└── PID_2
      ├── T1C
      ├── T2
      └── FLAIR
└── ...

Tumor Segmentation

Please create your segmentation according to the following format:

• Label 1: Contrast enhancing tumor
• Label 2: Non-enhancingT2/FLAIR abnormalities (Edema)
• Label 3: Necrotic like parts

For prediction

Data should be in the following structure (IMPORTANT! Add tumor segmentation in nifti format!):

└── PID_1
    ├── IMG_DATA
    │   ├── PID_1_0001.nii.gz
    │   ├── PID_1_0002.nii.gz
    │   ├── PID_1_0003.nii.gz
    │   └── PID_1_tum_seg.nii.gz
└── ...

Terminal commands

Terminal arguments:

• '-preprocess' must be 'Y' or 'N' if dicom files should be converted to nifti, registered and n4-bias-field corrected
• '-feat_extract' must be 'Y' or 'N' if features should be extracted
• '-age' patient age for the prediction
• '-input_path' must be the path to the patient folder

Example:

Only preprocess

main.py -preprocess Y -feat_extract N -age 0 -input_path /Users/robin/data/PID

Prediction and Feature extraction

main.py -preprocess N -feat_extract Y -age patient_age -input_path /Users/robin/data/PID

Only prediction

main.py -preprocess N -feat_extract N -age patient_age -input_path /Users/robin/data/PID

Output

Final output

The final output will create feature folder containing radiomics features as Excel file and the prediction with a preview of the tumor and an Excel file with the predictions.

└── PID_1
    ├── IMG_DATA
    ├── FEATURES
            ├── PID_1_features.xlsx
            ├── PID_1_preview_tumor.png
    ├── PREDICTION
            ├── PID_1_prediction.xlsx
└── ...