S121 - Bias Field Correction in MRI with Hampel Noise Denoising Diffusion Probabilistic Model
Junhyeok Lee, Junghwa Kang, Yoonho Nam, TaeYoung Lee
Non-uniform bias field due to external factors hampers quantitative MR image analysis. For reliable quantitative MR image analysis, appropriate correction for the bias field is necessary. In this study, we propose Hampel denoising diffusion model to effectively correct the bias field from MR images. Compared with N4 and Gaussian denoising diffusion models, the proposed model provided higher PSNRs, SSIMs and lower MSEs. Higher efficiency could be achieved compared to N4 when our model takes 9 times faster in inference time.
Schedule: Wednesday, July 12: Virtual poster session - 8:00–9:00
Poster location: Virtual only