This model segments the bilateral femur from CT images. The segmentation model was built from convolutional neural network using a Bayesian U-net architecture (https://github.com/yuta-hi/bayesian_unet) and was trained on manual segmentations of 120 cases.
Journal paper can be found at (https://doi.org/10.1007/s11657-022-01063-3)
When using these codes, please cite
- Uemura, K., Otake, Y., Takao, M. et al. Development of an open-source measurement system to assess the areal bone mineral density of the proximal femur from clinical CT images. Arch Osteoporos.(2022). https://doi.org/10.1007/s11657-022-01063-3
- Python 3.6
- keras 2.2.3
- CPU or NVIDIA GPU + CUDA CuDNN
Tested on the following environments: Win10, GPU: GTX1080 CUDA: 10.1, CuDNN: 7.6.5
For details, please see requirements.txt
Codes can be used for research purpose or for educational purpose.
For details, please see LICENCE.txt