The QBIS Lab, led by Dr. Rachel K. Surowiec at Purdue's Weldon School of Biomedical Engineering, works to close the loop between what happens inside a tissue and what we can see in a patient. Clinical patterns point us toward mechanisms worth understanding. Mechanistic and preclinical work tells us what our imaging should be measuring. Sharper measurements then feed back into how we read a scan and judge who is at risk. The lab is built to run in both directions rather than sit at either end.
That loop rests on a premise about bone: how much a person has tells you less than what condition it is in. Bone mineral density, the standard measure, misses much of what determines whether a bone actually fails. We read bone through the water bound in its matrix, its composition, and its mechanics, because those signals change early, and because they may carry a record of what a skeleton has lived through that density cannot see. A central question in the lab is whether bone retains a readable trace of prior systemic stress, and whether that history changes how it responds later. It is the kind of mechanistic question our imaging is built to answer and then to detect in patients.
Turning the loop takes measurement most tools cannot provide. We apply ultrashort echo time MRI and advanced k-space trajectories, including PETALUTE, to quantify tissues whose signal disappears too fast for conventional imaging. We pair these with near-infrared imaging, Fourier-transform spectroscopy, and thermogravimetric analysis to resolve bone's building blocks in space and follow how they change. From these acquisitions we extract quantitative biomarkers of tissue condition, not only images of it.
Artificial intelligence runs through the whole loop. We build machine-learning methods that make hard imaging usable, including motion correction and microstructure recovery in HR-pQCT, non-rigid motion correction in lung MRI, and reconstruction of sparse-view CT. On the analysis side, we use radiomics to turn imaging of bone and muscle into a minable dataset for biomarkers of both risk and resilience: who is likely to decline, and who holds up against it. We favor interpretable models, including attention maps that show which regions of a tissue drive a prediction, so the result points back to a location and a mechanism rather than a number alone. That interpretability is what lets a clinical finding become the next mechanistic question.
Much of this work serves people what the current tools serve poorly: those with X-linked hypophosphatemia, chronic kidney disease, hepatic disease, and postmenopausal osteoporosis, and the growing number undergoing rapid weight loss after bariatric surgery or on GLP-1 therapies, where the skeletal consequences are real and largely unwatched.
We work across bone, lung, and beyond, with a team of PhD students at the intersection of imaging, computation, and biomechanics.
