PET (Positron Emission Tomography) is an advanced cardiac imaging modality for studying cardiac functionality and blood flow dynamics. This project aims to explore the impact of various sub-supervised denoising techniques on Cardiac Dynamic PET images to optimize image clarity and analytical accuracy.
git clone [email protected]:yoko19191/cardiac-dynamic-pet-denoising.git
cd cardiac-dynamic-pet-denoising
virtualenv .env --python=python3.11
cd .env
source /bin/activate
pip install -r requirements.txt
- Block-Matching and 4D filtering
- Noise-As-Clean(NAC)
- Noise2Void(N2V)
- Neighbor2Neighbor(Nb2Nb)
- Zero-shot Noise2Noise(ZS-N2N)
- Neighbor2Neighbor(Nb2Nb) (extend to 2.5D)
- Zero-shot Noise2Noise(ZS-N2N) (extend to 2.5D)
- Peak signal-to-noise ratio(PSNR)
- Structural similarity(SSIM)
- Blind/Referenceless Image Spatial Quality Evaluator (BRISQUE)
- Time-activity curve(TAC)
you may downlaod the example dataset here
PSNR
SSIM
BRISQUE
TAC