GitHub - skill-diver/segmentation_refiner: Addressing the limitations of deep learning methods, the algorithms and anatomical principles enable automated detection and repair of missegmentation in 3d medical image data, the repair results can reach 90% dice score (evaluated by human experts)

Branches Tags

Name		Name	Last commit message	Last commit date
Latest commit History 1 Commit
sample		sample
ssm		ssm
statistics		statistics
third_party		third_party
train		train
__init__.py		__init__.py
classifier.py		classifier.py
consistency_loop.py		consistency_loop.py
environment.yml		environment.yml
eval.py		eval.py
eval.sh		eval.sh
graph.py		graph.py
identify.py		identify.py
locate.py		locate.py
metrics.py		metrics.py
readme.txt		readme.txt
regressor.py		regressor.py
requirements.txt		requirements.txt
run_test.out		run_test.out
run_test.py		run_test.py
segment_spine.py		segment_spine.py
segment_vertebra.py		segment_vertebra.py
segmentor.py		segmentor.py
test.py		test.py
utils.py		utils.py

Repository files navigation

1. Set up environment
pip install -r requirements.txt

2. Do refinement.
python test.py -D <input_dir> -B <spine_binarymask> -S <output_dir> -P<vertebrae path which you want to refine>

About

Addressing the limitations of deep learning methods, the algorithms and anatomical principles enable automated detection and repair of missegmentation in 3d medical image data, the repair results can reach 90% dice score (evaluated by human experts)

Readme