nnUNet

nnUNet_Segmentation

nnUNet_Colab.ipynb is the nnUNet model running on Colab. From data preparation to training to prediction, the resulting model parameters are mainly saved in the checkpoint_final.pth file

checkpoint_final.pth is the result of the version with the most epoches trained so far, which is about 80 epoches by 40%.

fold_2 progress.png fold2 training process

fold_4 progress.pngfold 4 training process

Segmentation result

Feature extraction and Classifier

Radiomics-features.xlsx All features extracted under the first setting (refer to P ituitary_tumor.ipynb code)

Radiomics-features2.xlsx See P ituitary_tumor.ipynb code for all features extracted in the second setting.

Reduced dimention features:

Classifier result

Tools

TianTan Tools.exe executable file, integrates several methods of data transformation, easy to directly obtain suitable for nnUNet training database

tiantan.ico image used as logo

TianTan.py is the source file from which the exe document is generated

Pituitary_tumor.ipynb, which is the most important code of the entire project, including data processing, feature extraction, classifier design, etc. TianTan Tools is a part of the visualization code.

Pituitary_tumor.pdf, and Pituitary_tumor.ipynb is transferred to pdf

reference paper.pdf is a reference for feature extraction and classifier design in the first way

Funding

This project is supported by Neurosurgery team, Beijing Tiantan Hospital.

https://www.mingyihui.net/hospital_84.html