Usama Mirza

Usama Mirza

Ph.D. Student in Electrical and Electronics Engineering

Bilkent University

About

I am a Ph.D. student in Electrical and Electronics Engineering at Bilkent University and a graduate research assistant in the ICON Lab at the National Magnetic Resonance Research Center (UMRAM), working under Prof. Tolga Çukur. I received my M.Sc. in Electrical and Electronics Engineering from Bilkent University, with a thesis on diffusion bridges for MRI reconstruction, and my B.Sc. in Electrical Engineering from the National University of Sciences and Technology (NUST).

My research focuses on generative models, particularly diffusion models and GANs, for computational imaging and inverse problems. I develop diffusion-based methods for accelerated MRI reconstruction, and have also worked on multi-contrast MRI synthesis and federated learning.

The latest version of my curriculum vitae is available here.

Interests
  • Generative Models
  • Diffusion Models
  • GANs
  • Computational Imaging
  • Inverse Problems
  • Accelerated MRI
  • MRI Synthesis
  • Federated Learning
Education

  • Ph.D. in Electrical and Electronics Engineering, 2024–Present

    Bilkent University

  • M.Sc. in Electrical and Electronics Engineering, 2021–2024

    Bilkent University

  • B.Sc. in Electrical Engineering, 2017–2021

    National University of Sciences and Technology (NUST)

Publications

Journal Articles

Usama Mirza, Onat Dalmaz, Hasan A Bedel, Gökberk Elmas, Yilmaz Korkmaz, Alper Güngör, Salman UH Dar, Kader K Oğuz, Tolga Çukur
June, 2026 IEEE TMI
Learning Fourier-Constrained Diffusion Bridges for MRI Reconstruction
Learning Fourier-Constrained Diffusion Bridges for MRI Reconstruction

We introduce the first Fourier-constrained diffusion bridge (FDB) for MRI reconstruction, which learns a dealiasing transformation from undersampled to fully-sampled data instead of relying on a noise-governed generative trajectory.

PDF Cite Code DOI
Onat Dalmaz, Usama Mirza, Gökberk Elmas, Muzaffer Özbey, Salman UH Dar, Emir Ceyani, Kader K Oğuz, Salman Avestimehr, Tolga Çukur
May, 2024 MedIA
One Model to Unite Them All: Personalized Federated Learning of Multi-Contrast MRI Synthesis
One Model to Unite Them All: Personalized Federated Learning of Multi-Contrast MRI Synthesis

We introduce pFLSynth, the first personalized federated learning method for multi-contrast MRI synthesis, which improves reliability against data heterogeneity via model specialization to individual sites and synthesis tasks.

PDF Cite Code DOI

Conference Papers

Usama Mirza, Efe Ozdilek, Esin Öztürk-Işık, Tolga Çukur (2026). Low-Frequency Anchored Diffusion Transition Process for MRI Reconstruction. SIU 2026.

PDF Cite DOI

Usama Mirza, Mustafa Arda Aydın, Tolga Çukur (2025). Frequency-Based Soft Diffusion Model for Accelerated MRI Reconstruction. SIU 2025.

PDF Cite DOI

Usama Mirza, Fuat Arslan, Tolga Çukur (2024). Super Resolution MRI via Upscaling Diffusion Bridges. SIU 2024.

PDF Cite DOI

Onat Dalmaz, Baturay Saglam, Gökberk Elmas, Usama Mirza, Tolga Çukur (2023). Denoising Diffusion Adversarial Models for Unconditional Medical Image Generation. SIU 2023.

PDF Cite DOI

Usama Mirza, Tolga Çukur (2023). Super-Resolution Diffusion Model for Accelerated MRI Reconstruction. SIU 2023.

PDF Cite DOI

Onat Dalmaz, Usama Mirza, Gökberk Elmas, Muzaffer Özbey, Salman UH Dar, Emir Ceyani, Salman Avestimehr, Tolga Çukur (2023). Personalized, Federated, and Unified MRI Contrast Synthesis. ISBI 2023.

Onat Dalmaz, Usama Mirza, Gökberk Elmas, Muzaffer Özbey, Salman UH Dar, Tolga Çukur (2022). A Specificity-Preserving Generative Model for Federated MRI Translation. MICCAI-DeCaF 2022.

PDF Cite DOI

Usama Mirza, Onat Dalmaz, Tolga Çukur (2022). Skip Connections for Medical Image Synthesis with Generative Adversarial Networks. SIU 2022.

PDF Cite DOI

Conference Abstracts & Workshop Papers

Usama Mirza, Onat Dalmaz, Hasan A Bedel, Gökberk Elmas, Alper Güngör, Tolga Çukur (2024). Accelerated MRI Reconstruction with Fourier-Constrained Diffusion Bridges. ISMRM 2024.

Code

Usama Mirza, Onat Dalmaz, Hasan A Bedel, Gökberk Elmas, Alper Güngör, Tolga Çukur (2023). MRI Reconstruction with Fourier-Constrained Diffusion Bridges. MedNeurIPS 2023.

Code

Onat Dalmaz, Usama Mirza, Gökberk Elmas, Muzaffer Özbey, Salman UH Dar, Emir Ceyani, Salman Avestimehr, Tolga Çukur (2023). A Personalized Federated Learning Approach for Multi-Contrast MRI Translation. ISMRM 2023.

Onat Dalmaz, Usama Mirza, Gökberk Elmas, Muzaffer Özbey, Salman UH Dar, Emir Ceyani, Salman Avestimehr, Tolga Çukur (2022). pFLSynth: Personalized Federated Learning of Image Synthesis in Multi-Contrast MRI. MedNeurIPS 2022.

Experience

 
 
 
 
 
Graduate Research Assistant
ICON Lab, National Magnetic Resonance Research Center (UMRAM)
September 2021 – Present Ankara
  • Developed Fourier-constrained diffusion bridges for accelerated MRI reconstruction (first author, IEEE TMI 2026; code: github.com/icon-lab/FDB).
  • Designed frequency-domain diffusion models for accelerated MRI and upscaling diffusion bridges for MRI super-resolution (first author, SIU 2023–2026).
  • Contributed to pFLSynth, a personalized federated learning model for multi-contrast MRI synthesis across institutions (Medical Image Analysis 2024).
 
 
 
 
 
Graduate Teaching Assistant
Bilkent University
September 2021 – Present Ankara
Courses: Engineering Mathematics I & II, Linear System Theory, Neural Networks
 
 
 
 
 
Reviewer
IEEE Transactions on Medical Imaging
January 2024 – Present
 
 
 
 
 
Research Intern
TUKL Research and Development Lab, NUST
June 2019 – September 2019 Islamabad
  • Surveyed FPGA accelerator architectures for deep neural networks.
  • Worked with Vivado HLS (C/C++) and Xilinx Vivado on Zynq FPGA platforms.
  • Prepared neural network models in Python for FPGA deployment.

Honors & Awards

  • 2024
    ISMRM Summa Cum Laude Merit Award
    International Society for Magnetic Resonance in Medicine (ISMRM), Singapore. Oral presentation on Fourier-constrained diffusion bridges for accelerated MRI.
  • 2015
    Outstanding Cambridge Learner Award
    Cambridge Assessment International Education. Highest mark in the world in O-Level Mathematics.