• Development of quantitative imaging biomarkers to guide the adaptive use of immunotherapy, radiation and chemotherapy against solid tumors.
    • Development of differential diagnostics for cancer by machine learning on multiparameter Magnetic Resonance Imaging (mpMRI) and digitized multiplexed immunohistochemistry (MIHC) images.
    • Joint reconstruction of mpMRI images from co-registered sparse k-space data to optimize performance of mpMRI machine learning diagnostic vs. total scan time.
    • Synthetic MRI and image prediction to guide radiotherapy of solid tumors.



To close the loop between imaging, image analysis, and biomarker/predictive model, so as to inform the optimal task-specific acquisition and reconstruction of raw image data.




Imaging in Prospective Clinical Trials


DW-MRI and DCE-MRI in a Phase 1 clinical trial of Crolibulin 









Ferumoxytol-enhanced MRI as a biomarker of tumor-associated macrophages



Multispectral Analysis of Multiparameter MRI


Multiparameter MRI phenotypes of normal tissues and intratumoral habitats in pancreatic ductal adenocarcinoma





Tissue type maps of normal brain and recurrent glioblastoma at various times post-initiation of treatment (C1D1)




Joint reconstruction of undersampled co-registered mpMRI raw data to accelerate the acquisition of intratumoral habitat maps.












Machine Learning on Multiparameter MRI

Self-Organizing Maps for Differential Diagnosis. Activation maps of two classes of tumors are shown, with the top five neurons most characteristic of each tumor class indicated by green X's and red diamonds. Also shown are the normalized intensities on T2W, T1W unenhanced, T1W-CE arterial phase, and T1W-CE venous phase MRI that correspond to each neuron.






Molecular Imaging

pH and redox imaging using responsive gadolinium complexes