Projects

Overview In recent years, deep learning methods have made significant advancements in fast MRI reconstruction, yielding many promising results. However, challenges are constantly rising. Supervised methods for MR reconstruction are costly because of the collection of k-space label data and a model for one specific application is hard to apply to other applications. Apart from that, worries about the uncertainty caused by undersampling strategies and deep learning algorithms have limited their usage in clinical practice until now, which could lead to hallucinations. The application of generative image priors is one way out.

I have developed a library, called spreco based on TensorFlow, for training generative models. With that, I have trained many medical image priors that are hosted on this huggingface page and used them to address above challenges. Many of my projects are listed below.

Feel free to email me if you are interested in my research. Remote collaboration is also welcome!

Generative Image Priors for MRI Reconstruction Trained from Magnitude-Only Images

generative pretraining, image priors, diffusion models, inverse problem, MR image reconstruction, proximal operator, optimization

Bayesian MRI reconstruction with joint uncertainty estimation using diffusion models

sampling posterior, Bayesian inference, diffusion models, uncertainty estimation, inverse problem, MR image reconstruction, Monte Carlo, Markov chain

Deploy generative image priors for image reconstruction using BART

MR image reconstruction, model deployment, TensorFlow C API, TensorFlow computation graph

MRI reconstruction using deep Bayesian estimation

generative model, PixelCNN, Bayesian estimation, MR image reconstruction, optimization

Generalized deep learning-based proximal gradient descent for MR reconstruction

MR image reconstruction, proximal gradient descent, denoising autoencoder