lung nodule classifier for 9 attribute
python 3.5.2
pytorch 0.2.0
- malignancy (range: 1~5)
- sphericity (range: 1~5)
- margin (range: 1~5)
- spiculation (range: 1~5)
- texture (range: 1~5)
- calcification (range: 1~6)
- internal structure (range: 1~4)
- lobulation (range: 1~5)
- subtlety (range: 1~5)
| MAE under 0.1 | MAE under 0.2 | MAE under 0.3 | MAE under 0.4 | MAE under 0.5 | MAE under 0.6 | MAE under 0.7 | MAE under 0.8 | MAE under 0.9 | MAE under 1 | |
| malignancy | 15.24% | 24.76% | 42.86% | 60.00% | 71.43% | 80.95% | 86.67% | 92.38% | 93.33% | 94.29% |
| sphericity | 17.14% | 29.52% | 50.48% | 60.00% | 67.62% | 72.38% | 80.95% | 84.76% | 89.52% | 92.38% |
| margin | 19.05% | 38.10% | 48.57% | 60.95% | 67.62% | 77.14% | 82.86% | 85.71% | 86.67% | 89.52% |
| spiculation | 21.90% | 37.14% | 63.81% | 72.38% | 76.19% | 81.90% | 83.81% | 87.62% | 88.57% | 91.43% |
| texture | 22.86% | 59.05% | 76.19% | 80.95% | 84.76% | 84.76% | 86.67% | 89.52% | 91.43% | 92.38% |
| calcification | 89.52% | 89.52% | 92.38% | 93.33% | 93.33% | 93.33% | 97.14% | 99.05% | 99.05% | 100.00% |
| internal_structure | 100.00% | 100.00% | 100.00% | 100.00% | 100.00% | 100.00% | 100.00% | 100.00% | 100.00% | 100.00% |
| lobulation | 19.05% | 31.43% | 52.38% | 67.62% | 75.24% | 79.05% | 85.71% | 87.62% | 91.43% | 92.38% |
| subtlety | 15.24% | 32.38% | 50.48% | 57.14% | 67.62% | 71.43% | 80.95% | 86.67% | 86.67% | 88.57% |
- https://luna16.grand-challenge.org/download/
- download data and candidates_V2.csv
- https://wiki.cancerimagingarchive.net/display/Public/LIDC-IDRI
- download Radiologist Annotations/Segmentations (XML)
- set luna and lidc path: set config_training.py
- export path: source export_path.sh
- train, val idx npy make: python preprocess/make_validate_npy.py
- preprocess data make: python preprocess/prepare.py
- train: sh train.sh
https://github.com/lfz/DSB2017
https://github.com/juliandewit/kaggle_ndsb2017