https://github.com/wdbm/dosimetric_correlations
https://github.com/wdbm/dosimetric_correlations
cancer tensorflow
Last synced: 2 months ago
JSON representation
- Host: GitHub
- URL: https://github.com/wdbm/dosimetric_correlations
- Owner: wdbm
- License: gpl-3.0
- Created: 2017-06-19T21:08:11.000Z (about 9 years ago)
- Default Branch: master
- Last Pushed: 2018-04-27T17:37:24.000Z (about 8 years ago)
- Last Synced: 2025-03-28T03:46:24.639Z (over 1 year ago)
- Topics: cancer, tensorflow
- Language: Jupyter Notebook
- Size: 2.25 MB
- Stars: 0
- Watchers: 2
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# dosimetric_correlations
# note
This project is a work in progress.
# setup
Install the following:
- NumPy
- pandas
- Scikit-learn
- TensorBoard
- TensorFlow
# use
The input data could be CSV with the following fields:
- features:
- PTV vol (cc)
- Lungs-GTV vol (cc)
- Lungs-GTV-PTV vol (cc)
- Lungs-GTV in PTV vol (cc)
- KBP Lungs (cc)
- Lungs-GTV - KBP Lungs (cc)
- Heart vol (cc)
- Heart in PTV vol (cc)
- targets:
- V5 (%)
- V20 (%)
- Mean Lungs-GTV (Gy)
- V30 (%)
- Mean (Gy)
It could be CSV with the following fields:
- features:
- Dose/#
- Prescription
- PTV vol (cc)
- Lungs-GTV vol (cc)
- Lungs-GTV-PTV vol (cc)
- Lungs-GTV in PTV vol (cc)
- KBP Lungs (cc)
- Lungs-GTV - KBP Lungs (cc)
- targets:
- V5 (%)
- V20 (%)
- Mean Lungs-GTV (Gy)
The rightmost columns should be the targets. The number of targets can be specified as an argument for a neural network script.
Manually remove missing values from data.
Preprocess the CSV data such that all features are scaled to (-1, 1):
```Bash
./preprocess_CSV_file.py --infile=data.csv --outfile=preprocessed_data.csv
```
Train and evaluate on preprocessed CSV data with TensorBoard:
```Bash
./cures_cancer.py --help
./cures_cancer.py --infile=preprocessed_data.csv --TensorBoard
```