Research

EPINR ‑ Distortion Correction in Clinically‑Constrained Echo Planar MRIs with INR‑Based Registration

University of Virginia

Ongoing

• Correction of magnetic susceptibility distortion in EPIs (BOLD/fMRI and DWI/dMRI) without fieldmaps or reverse phase‑encoded acquisitions.
• Uses implicit neural representations (INRs) to learn physically‑plausible deformation fields from T1w MRIs to distorted EPIs.

MedIL ‑ Medical images from Implicit Latent spaces

University of Virginia

2024 ‑ 2025

• A first‑of‑its‑kind autoencoder that encodes medical images at their native resolution, without resampling, using implicit neural representations.
• Used alongside diffusion generative models to generate medical images of any resolution.
• Demonstrated exceptional or competitive performance versus previous methods on both T1w brain MRIs and lung CTs.

Diffusion MRI features for predicting progression of multiple sclerosis

University of Virginia

Ongoing

• Collaboration with Dr. Myla Goldman M.D. at Virginia Commonwealth University Department of Neurology.
• Studied image features in MS patient diffusion MRIs that correspond with progression or non‑progression of MS symptoms.
• Found that extra‑cellular content in MS lesions corresponds to patient performance in a six minute walk diagnostic test.

FENRI ‑ Fiber orientations from Explicit Neural RepresentatIons

University of Virginia

2023 ‑ 2025

• A novel spatially‑continuous super‑resolution neural network to improve white matter tractography in clinical‑grade diffusion MRIs.
• Outperforms current methods in resolution enhancement of in vivo noisy, low‑resolution diffusion MRIs.
• Beats other methods at recovering white matter tracts from low‑resolution simulated diffusion images.