https://github.com/menchelab/radipop_scripts
Scripts to train and validate random forest model for HVPG prediction
https://github.com/menchelab/radipop_scripts
Last synced: 11 months ago
JSON representation
Scripts to train and validate random forest model for HVPG prediction
- Host: GitHub
- URL: https://github.com/menchelab/radipop_scripts
- Owner: menchelab
- Created: 2023-10-25T05:17:35.000Z (over 2 years ago)
- Default Branch: master
- Last Pushed: 2023-11-20T15:16:14.000Z (over 2 years ago)
- Last Synced: 2025-04-12T12:38:57.296Z (about 1 year ago)
- Language: Jupyter Notebook
- Size: 20.3 MB
- Stars: 1
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- Citation: CITATION.cff
Awesome Lists containing this project
README
# Radiomics-based prediction of portal hypertension severity and of liver-related events using routine CT scans of patients with cirrhosis
**Hepatic venous pressure gradient (HVPG)** is the reference standard to diagnose portal hypertension. Elevated HVPG is predictive of hepatic decompensation and mortality [Ripoll, 2007], and its measurement is indicated for diagnosis, therapy monitoring and risk stratification. However, HVPG measurement is invasive, relatively expensive and requires specialized medical infrastructure and expertise. Therefore, a non-invasive alternative is highly desirable.
In this project, we developed a radiomics-based model for the non-invasive determination of HVPG > 10mmHg (clinically significant portal hypertension, CSPH) from abdominal CT scans.
This work is published in <<<>>>
## Code base for analysis
This codebase is organized in 3 main folders:
### scripts_0preprocessing
- clean the metadata
- preprocess raw images
- extract radiomics features
### scripts_1ml
- explore the feature space
- feature selection and batch correction
- train and optimize a random forest classifier to predict for HVPG ≥10 mmHg
- evaluate performance of the model
### scripts_cox (Lorenz Balcar/Bernhard Scheiner)
- preform cox regression analysis for prognosis endpoints
## References
Ripoll, C. et al. Hepatic venous pressure gradient predicts clinical decompensation in patients with compensated cirrhosis. Gastroenterology 133, 481–488 (2007)