Impact: Quantification of the information contribution of scans in an MRI exam can assist the operator/radiologist in optimizing exam protocols for high efficiency with minimum information loss. Such optimized exams could reduce strain on the patient, while increasing the hospital’s patient-throughput.

Impact: We used a multi-center paired spinal MR/CT dataset to create an MR-to-CT network, generalizing to three out-of-distribution centers. Our findings facilitate radiation-free 3D bone visualization without changes to existing diagnostic scan protocols, unlocking its potential for surgical applications.

Impact: We show that quantitative susceptibility maps can be reconstructed from standard multi-parametric MRI data without phase information. It demonstrates a new way to recover otherwise lost tissue information, potentially enabling retrospective analysis of existing imaging datasets.

Impact: The proposed network enables MR image translation (T1w⟷T2w) based on metadata (image contrast and scan parameters). By synthesizing parameter-specified contrasts from limited acquisitions, this approach can reduce scan burden while offering flexibility in image contrast determined by target scan parameters.

Impact: We developed a graph-attention fusion framework for high-quality synthesis of brain MRI images. This approach delivers superior synthesis quality in motion artifact removal and cross-modal translation, highlighting its potential to enhance neuroimaging analysis when high-quality or comprehensive data are lacking.

Impact: This study enables contrast-free, multi-phase breast DCE-MRI generation using non-contrast inputs, reducing GBCA dependence and patient risk while preserving diagnostic fidelity, paving the way for safer, faster, and more accessible breast cancer imaging and future AI-driven contrast synthesis research.

Impact: This study demonstrates the feasibility of generating synthetic T1CE images from non-contrast MRI, thus reducing reliance on contrast agents and improving clinical workflows. This approach provides a non-contrast alternative to endometrial cancer detection and diagnosis.

Impact: This research transforms synovitis management by establishing pathophysiologically distinct subtypes through integrated MRI. This refined stratification enables precise treatment matching—directing anti-angiogenic therapy to hypervascular subtypes while avoiding overtreatment in fibrotic cases—optimizing therapeutic efficacy and accelerating clinical trials for targeted therapies.