Radiology Research Partnerships
Technological innovation in biomedical imaging requires a breadth and depth of expertise that can only be mobilized by highly interdisciplinary and intensely collaborative teams.
The translation of new technologies to the clinic requires close cooperation among experts from disparate fields and organizational cultures, ranging from the basic sciences and medical subspecialties to mechanical, electrical, and computer engineering and medical device manufacturing. NYU Langone’s Department of Radiology accelerates this process through our wide-ranging partnerships.
Collaboration begins in our department, where clinician–researchers are embedded in our translational teams, extends throughout the institution with interdepartmental investigations, and reaches across the world in the form of partnerships that include medical research and medical industry enterprises on four continents.
The Department of Radiology sponsors protected research time for a number of clinical radiologists from a variety of specialties, including breast imaging, neuroimaging, musculoskeletal imaging, and abdominal imaging.
Clinician–researchers in our 50–50 program work closely with physicists, engineers, and computer scientists to define pressing clinical problems and to investigate, develop, and evaluate promising imaging solutions.
Examples of recent work include five-dimensional imaging of the heart, artificial intelligence methods for detection of breast cancer, and abdominal MRI that produces clear images without requiring patients to hold their breath in the scanner.
Worldwide Collaborative Projects
We operate the Center for Advanced Imaging Innovation and Research (CAI2R), a National Institutes of Health Center for Biomedical Imaging and Bioengineering. The center’s mission is to create and disseminate leading-edge technologies in order to support scientific research and advance the state of the art in radiology. CAI2R (pronounced care) is founded on a unique model that combines basic research with ongoing exposure to unsolved biomedical problems, accelerates clinical translation of vanguard technologies, and relies on close interdisciplinary collaboration among academia, medicine, and industry.
Through CAI2R, we are engaged in more than 60 collaborative projects with partners at dozens of institutions in North America, South America, Europe, and Asia. In these collaborations, we make sophisticated emergent technologies available to partners whose investigations stand to benefit from early access to novel applications in neuroimaging, musculoskeletal imaging, cancer imaging, and cardiovascular imaging. In turn, our partners’ investigations serve as test beds for the emergent technologies themselves. The resulting iterative development cycle helps advance our partners’ research while significantly expediting maturation of our innovations.
We are also proud to offer imaging capabilities to investigators around the world through a similar number of service projects. Here, we make more mature technologies available to research teams that may benefit from access to sophisticated applications. In these projects we provide advanced technologies as a service rather than engaging in iterative development.
Partnerships with Industry
Although academic research can be a hotbed of innovation, commercial enterprise is often the only vehicle able to convey novel technologies to the healthcare market at scale. Transition of technology from lab bench to factory floor, however, is a complex and often lengthy process.
Our unique model of collaboration allows medical device manufacturers to embed scientists in our research teams, promoting rich and frequent communication, as well as a high level of shared understanding across organizational cultures. This model helps us identify promising ideas earlier, iterate prototypes faster, and work together more effectively in bringing innovations to the clinic.
One example is our partnership with Siemens Healthineers, a major manufacturer of MRI scanners. We developed an MRI method called GRASP that produces high-quality abdominal images without requiring patients to hold their breath during exams. As a result of the partnership, GRASP has become available around the world as a standard feature on Siemens MAGNETOM Sola and Vida MRI systems under the name Compressed Sensing GRASP-VIBE. Clinical availability of this method is a significant breakthrough for the pediatric patients and adults for whom repeated breath holds are challenging.
Siemens is now adapting to its scanner platforms a technology we recently created for correction of motion effects on PET/MR images. Such fast and broad translation of imaging innovations would be impossible without collaborations between academia and industry.
Partnerships are also critical when private industry commands major concentrations of highly specialized resources and expertise that are unavailable in academia and that can facilitate innovation by helping answer well-defined research questions.
A case in point is our collaboration with Facebook Artificial Intelligence Research (FAIR) to develop fastMRI—deep learning methods for the acquisition and reconstruction of medical images—with the goal of making conventional MRI 10 times faster. This partnership brings our imaging researchers together with some of the world’s foremost experts in AI and allows us to leverage FAIR’s vast computational resources in the service of advancing biomedicine. All resulting code is freely shared. Moreover, in order to further encourage innovation in this area, we have made available the largest, rigorously deidentified open-source set of MRI data for AI research and are holding reconstruction competitions that are open to research teams anywhere.
Open Source and Open Data
We share source code and data to stimulate technological innovation, facilitate scientific reproducibility, and advance clinical translation throughout the field of biomedical imaging.
CAI2R offers dozens of research-oriented software downloads, including MRI reconstruction code, RF simulation tools, image analysis software, and image datasets. We also created the Yarra Framework, a suite of open-source MRI reconstruction tools to support research and custom translational implementations of computationally demanding techniques, including GRASP, which is itself available as a free download. CAI2R has also sent scientist envoys to train local personnel in the setup, configuration, and use of these technologies at many clinics and research centers around the world.
Our fastMRI dataset addresses the pressing need for large, well-documented, and open datasets for research into AI-based MRI reconstruction. By sharing high-quality, meticulously deidentified, and voluminous data, we galvanize innovation and encourage peer institutions to contribute to open data initiatives. Such resources strengthen research findings, boost reproducibility of AI studies, and move the field forward. The demand for substantial, clean, and reliable data is high. Our fastMRI resources were downloaded by more than 2,000 researchers within the first 18 months of going live.