The National Institute of Biomedical Imaging and Bioengineering (NIBIB) has awarded $6.7 million to NYU Grossman School of Medicine to continue operating the Center for Advanced Imaging Innovation and Research. The award marks the tenth anniversary of the center’s founding and extends its mandate as a National Center for Biomedical Imaging and Bioengineering (NCBIB) through 2029. The NCBIB designation is bestowed in five-year terms on select research enterprises that create and share unique, leading-edge technologies to advance the field in basic discovery, translational science, and clinical studies.
“Our mission remains the same, which is to bring people together to create new ways of seeing,” said Daniel Sodickson, MD, PhD, professor of radiology, neuroscience and physiology, chief of innovation in the radiology department, co-director of the Tech4Health Institute, and the founding overall principal investigator of the Center for Advanced Imaging Innovation and Research (CAI2R, pronounced care) at NYU Langone Health. “We are developing new ways to use artificially enhanced vision—in this case medical imaging, mostly MRI—and trying to figure out how to make it faster, better, and ultimately different than it has been.”
Toward Smarter, More Flexible, More Informative Imaging
As the center continues to pursue its longtime aims of improving the speed, versatility, and information content of MRI, it is also beginning to explore new questions made possible by the advent of artificial intelligence. One is whether “image memory”—useful data from prior imaging sessions—can inform new exams. “For some reason, we image you the same way whether we’ve seen you no times before, five times before, or a hundred times before,” said Dr. Sodickson about the current standard of care. “That seems crazy, particularly in the era of AI.” Exploiting connections across time points is already central to accelerated dynamic imaging, which produces movie-like sequences. But linking MRI data across months or years “may be a little more complex than if you’re gathering all the data in one sitting,” Dr. Sodickson said.
In another major research direction, the center’s scientists are working to increase the flexibility and accessibility of MRI hardware by capitalizing on novel materials, auxiliary sensors, and machine learning algorithms. “The next five years is really going to be devoted to figuring out how we can change imaging machines and protocols,” said Dr. Sodickson. “How can we lean more and more on our advanced software to pull out more information from less data?” The answers are likely to reshape not only the devices but also the economics of using them. “There’s a lot of emphasis in the field on making imaging accessible: instead of bringing people to imaging, figuring out how to bring imaging to people,” Dr. Sodickson said. The center’s 2023 MRI4ALL hackathon, aimed at building a low-field scanner in just a few days, exemplifies the growing interest in the potential of simpler, purpose-built, do-it-yourself hardware.
The third significant investigative area of focus for the research center lies in advancing microstructural imaging with a set of techniques rooted in diffusion MRI, which probes the naturally random motion of water inside biological tissue and is therefore sensitive to many physiological processes at cellular level. Although these phenomena occur on the scale of microns—far too small to show up on MR images—they do inform the MR signal collected during scans. Extracting that information has the potential to shed light on the earliest stages of pathology and is akin to “basically using MRI as a kind of in vivo microscope,” said Dr. Sodickson.
The center’s aims are guided by three Technology R&D Projects (TR&Ds), each headed by two co-principal investigators. TR&D 1, focused on image acquisition and reconstruction, is led by Associate Professor of Radiology Li Feng, PhD, and Dr. Sodickson; TR&D 2, dedicated to MRI hardware, is led by Associate Professor of Radiology Ryan Brown, PhD, and Professor of Radiology Christopher Collins, PhD; and TR&D 3, centered on biophysical modeling of tissue microstructure, is run by Associate Professor of Radiology Els Fieremans, PhD, and Associate Professor of Radiology Dmitry Novikov, PhD. In each area, the team draws on a long record of advances.
A Decade of Imaging Innovation
From its inception, CAI2R has pursued a stated goal of pushing MRI beyond traditional stop-and-go scanning protocols and toward what the research team called rapid, continuous, comprehensive imaging. This triple commitment—to speed, pauseless acquisition, and rich information content—remains in the center’s DNA and has already led to important contributions to both imaging research and radiology practice.
The center’s flagship development in this area is an MRI technique called GRASP, which enables high-quality imaging in the presence of natural breathing motion and delivers data characterized by a high degree of flexibility. The technique makes MRI exams more comfortable for patients, simplifies scans for technologists, and produces clearer images for clinicians. It has been incorporated into Siemens MRI machines under the name Compressed Sensing GRASP-VIBE and is available in imaging clinics around the world. The philosophy behind GRASP, which Dr. Sodickson often likens to a shift from still photography to video streaming, “is now out in the field in a number of clinical products—including ones we helped create,” he said.
CAI2R scientists were also among the very first to apply artificial intelligence to the process of transforming raw MRI data into images and to show that such images are as accurate as traditional ones. The breakthrough has “set the stage for a lot of techniques for using AI to accelerate imaging, and there are a number of products now in the marketplace that use that technique,” said Dr. Sodickson. One of those products, a feature on Siemens scanners called Deep Resolve, has its origins in the center’s research. “In fact, AI accelerated imaging is the gold standard now on most clinical MRI machines,” said Dr. Sodickson, adding that the team’s leadership in this area is “something I’m pretty proud of.”
The team’s investigations in engineering of radiofrequency (RF) transmitter and detector coils for MRI have likewise influenced the field by changing assumptions about RF hardware. “We’ve gone through this path from rigid to flexible,” said Dr. Sodickson. “Now, instead of designing big, complex, fixed arrays to try to capture every last ounce of signal-to-noise ratio, we’ve been working more and more on flexible arrays like the glove coil,” he said, referring to a high-impedance array invented by the team in 2018, which opened the door to wearable RF coils that conform to a patient’s anatomy and motion. “This fits beautifully with an overall less-is-more trend across the field.”
Meanwhile, the center’s advances in using diffusion MRI to noninvasively obtain cellular-level information in vivo have “really led the way in the new understanding of microstructural imaging,” said Dr. Sodickson. The effort has yielded theoretical developments, innovative validation techniques, and clinical studies that have linked micron-level phenomena to the microgeometry and physiology of the brain. The team has formulated a unifying theoretical framework known as the standard model of diffusion in white matter and has “helped lead a revolution in microstructural imaging,” said Dr. Sodickson.
A Decade of Partnerships, Collaborations, and Resource Sharing
Swift clinical translation of the center’s most promising technologies is possible thanks to a unique arrangement in which industry scientists embed with the center’s investigators and clinician-researchers. One example is a partnership with Siemens Healthineers, a major manufacturer of MRI machines. Another is a collaboration with Meta AI Research, which has resulted in the largest openly shared set of raw, curated, deidentified MRI data for machine learning research, published in 2019 and augmented several times since.
The center’s researchers also maintain Collaborative Projects with scientists whose studies stand to benefit from its emerging technologies and who in turn test and help advance those very technologies. Since 2014, CAI2R has engaged in more than 80 such collaborations with scientific teams in the U.S. and abroad. In addition, sophisticated research groups can access the center’s unique but less experimental capabilities through arrangements called Service Projects.
Less direct but just as impactful has been the center’s practice of openly sharing useful resources, including software tools for MRI analysis, denoising, and simulations; and image reconstruction code. For research teams that may not have the means inhouse to create efficient, compliant translational study workflows, CAI2R has created platforms like Yarra and Mercure, and is working on additional modules with the goal of supporting the full arc of pulse sequence development, MR data acquisition, image reconstruction, post-processing and data visualization. In addition, the center’s engineers share hardware for research on tracking respiratory motion and for the development of advanced RF coils.
What’s Next Is Different
In a field as inherently interdisciplinary as biomedical imaging it’s difficult to foretell the next paradigm shift, and the center’s prevailing investigative ethos may be described as a combination of scientific rigor and creative open-mindedness.
“I feel like we are at a truly remarkable juncture in the history of imaging,” said Dr. Sodickson. “Things are changing very fast, and assumptions are being questioned left and right.”
Leaning into this sentiment, the center convened a scientific conference in 2023 to contemplate the very nature of imaging, with speakers bringing perspectives from areas as diverse as materials science, machine learning, and radio astronomy.
Building on a decade of working at the vanguard of MRI research—including the move toward rapid continuous acquisitions, the emergence of AI image reconstruction, the shift toward flexible RF coils and scanner “sensorization,” the renewed interest in lower magnetic fields, and the advances in decoding cellular information from MR signal—the center’s scientists now have a renewed mandate to envision and invent the field’s next frontier, where the role of imaging may be bigger and broader than it is today.
“Medicine is becoming more proactive and more preventive,” said Dr. Sodickson. “One of the things we’re trying to do as we move into our tenth year is to deliver the best performance for traditional applications of medical imaging while starting to figure out how to move the field toward emerging new goals like prevention.”
Related Story
Daniel Sodickson, vice chair for research at NYU Langone's radiology department, talks about cultivating talent, finding "a room of one's own," and distrusting startup packages.