LITTLE ROCK — The University of Arkansas for Medical Sciences (UAMS) will have a key role in a major international effort using artificial intelligence (AI) to teach computers at local research centers how to recognize cancer.
UAMS has received a four-year grant of $894,074 as part of EuCanImage, a nearly $9 million cancer research project based at the University of Barcelona in Spain.
A UAMS team led by Fred Prior, Ph.D., professor and chair of the College of Medicine Department of Biomedical Informatics, will lead the United States portion of the effort along with colleagues at Emory University in Atlanta.
“What we learn in this European project and the tools we develop will directly impact cancer patients in Arkansas by providing new understanding of the biology of cancer and new ways to target therapies,” Prior said. “It also brings European resources to Arkansas to fund research at the UAMS Winthrop P. Rockefeller Cancer Institute.”
Scientists are learning a lot about cancer by applying AI to the analysis of images, Prior said. The specific AI for the project is called machine learning, in which the computer learns how to recognize cancer, determines if the cancer is responding to treatment, and even understands changes in cancer genes during treatment.
“To create precision diagnostic tests and targeted therapies we need to understand the variability in the global human population and the variability in the disease itself,” Prior said.
Prior is the principal investigator of The Cancer Imaging Archive (TCIA), a National Cancer Institute (NCI)-funded program and a critical part of the EuCanImage project. The archive contains freely accessible, de-identified cancer radiology and pathology images and data. It includes fully anonymous information on patient histories, outcomes, treatment details, the results of machine and human expert analyses, and links to genetic and many other types of information for the same people but stored in other repositories..
Prior said new ways must be developed to both access the huge amounts of cancer data while maintaining patient privacy, and to train artificial intelligence-based analysis tools to make them more precise.
“This project explores answers to these questions by studying innovative ways of linking together data warehouses in Europe and the U.S., and sending the machine learning code to the data rather than bringing the data to it,” he said. “You can think of this as having our machines learn from teachers all over the world.”
As part of the project, EuCanImage plans to build a federated, secure and scalable cancer imaging platform, with capabilities that will enhance the potential of artificial intelligence in oncology.
Karim Lekadir, Ph.D., EuCanImage project coordinator from the University of Barcelona, said the project will leverage world-renowned expertise in the fields of radiomics/radiogenomics, artificial intelligence, data protection, data management and clinical oncology.
“EuCanImage is an unprecedented opportunity to create a long-awaited European platform for sharing and leveraging large-scale cancer imaging and non-imaging data,” he said.
Prior said the project has potential to include other parts of the world as well. He has worked with scientists in India, for example, to develop image repositories that might be linked using the tools developed as part of the project.
“The exciting prospect of having distributed data storage and distributed machine learning really opens the door to sampling the variability in the human population and in the expression of cancer by using data from around the world,” he said. “Ultimately we hope to learn new information about the relationships between things we see in images and the way cancer genes work and respond to therapy.”
Other members of Prior’s team include Lawrence Tarbox, Ph.D., assistant professor; Kirk Smith, instructor; William Bennett, instructor; and Tracy Nolan, instructor.
EuCanImage includes 20 research institutions, companies and clinical centers across Europe and the U.S. The study will leverage world-renowned expertise in radiomics/radiogenomics, artificial intelligence, data protection, data management and clinical oncology.
Other EuCanImage consortium partners are:
- Universitat de Barcelona, Spain (Coordinators)
- Universiteit Maastricht, Netherlands
- Erasmus Medical Center Rotterdam, Euro-BioImaging Node, Netherlands
- Barcelona Supercomputing Center, Spain
- Centre for Genomic Regulation, Spain
- Biobanking and Biomolecular Resources Research Infrastructure, Austria
- Universidad del País Vasco, Spain
- Lynkeus Srl, Italy
- Collective Minds Radiology AB, Sweden
- OncoRadiomics SA, Belgium
- Siemens Healthcare GmbH, Germany
- European Institute for Biomedical Imaging Research, Austria
- European Society of Oncologic Imaging, Austria
- European Association for Cancer Research, United Kingdom
- Università di Pisa, Italy
- Fundació Clínic Recerca Biomèdica / Hospital Clínic de Barcelona, Spain
- Umeå Universitet, Sweden
- Gdański Uniwersytet Medyczny, Poland
- Kauno Klinikos, Lilthuania