Git repository for MICCAI 2024
- FID
- MMD
- MS-SSIM
- PCA-tSNE
- Segmentation Quality Control
- AE-style GANs: https://github.com/cyclomon/3dbraingen
- HA-GAN: https://github.com/batmanlab/HA-GAN
- Conditional DPM: https://arxiv.org/abs/2212.08034
- MedSyn: https://ieeexplore.ieee.org/document/10566053
- MONAI: https://github.com/Project-MONAI/GenerativeModels
- Generate your own samples
- Run evaluation.py or simply use relevant evaluation functions
To run the 2-stage Anatomical-based evaluation, you need to follow these steps:
- Generate certain number of Brain MRI images, and real MRI images. Use Synthseg+ to do a whole brain segmentation for both data, SynthSeg+: https://github.com/BBillot/Synthseg
- Run quality control evaluation: Calculate the proportion of each ROI where the quality control exceeds 0.65, and set a threshold to detect the generated results, which we set at 0.95.
- Run subcortical evaluation and cortical evalution: Replace the CSV path with the segmentation output of your generated data and the generated data of the real image, and perform regression and Cohen's d calculation. The scripts are run_eval_aseg.py and run_eval_aparc.py, respectively, for subcortical and cortical segmentation results.