New Columbia Center Will Leverage AI to Enhance Patient Care

Artificial intelligence (AI) offers a pathway to enhance several critical aspects of care for patients with cancer or other diseases. Despina Kontos, PhD, a computer scientist with expertise in artificial intelligence, machine learning, and big data analytics for multi-modality imaging data, believes Columbia University offers both the perfect location and team to build that path.

“I think there is immense untapped potential in this area of research,” says Kontos, professor of radiology and director of the new Center for Innovation in Imaging Biomarkers and Integrated Diagnostics (CIMBID) at Columbia University. “Between NewYork-Presbyterian Hospital, Columbia, and Cornell, the wealth of data and diversity of the population is so unique that I truly believe we can be one of the leading institutions, if not the leader, in this area."

Despina Kontos is a computer scientist with expertise in artificial intelligence, machine learning, and big data analytics for multi-modality imaging data. Her work has contributed to a fundamental transition in the interpretation of breast cancer images, by showing that imaging data can be mined to extract sophisticated phenotypic signatures with independent diagnostic, prognostic, and predictive value. Her lab is also pursuing related research in lung cancer, integrating CT radiomic features with liquid biopsy data to predict response to targeted therapy and immunotherapy.

Under Kontos’ leadership, CIMBID will be dedicated to developing and integrating quantitative imaging and non-imaging biomarkers for disease prediction, particularly in cancer. Kontos, who made seminal contributions in leveraging artificial intelligence tools for risk prediction in breast and lung cancer while serving as professor of research radiology at the University of Pennsylvania, believes that AI can enable better patient care in three major areas:

• Cancer Screening and Protection—Identifying patients who are at a higher or lower risk for getting the disease and personalizing screening to factor in both risk profile and cost-effectiveness.

• Diagnosis, Prognosis, and Planning Therapy—Quantifying information about a particular disease to help physicians select the most appropriate treatment.

• Evaluation and Response to Therapy—Utilizing data from standard-of-care radiology images to assess whether a patient is responding to therapy early in the treatment process and modify it when it is not working.

“Artificial intelligence is trying to program computers to think like humans,” Kontos says. “A doctor looks at a report from a radiologist, looks at the notes, considers the family history, comorbidities, genetic tests, etc., and figures out a course of action. With these models, we can try to do that a bit more comprehensively and in a more quantitative and standardized fashion throughout the different stages of patient care.”

Achieving this goal requires collaboration across multiple disciplines, or “team science,” as noted by Dean Katrina Armstrong, MD, when she announced creation of the center and Kontos' recruitment.

“It’s essential,” Kontos says. “We can’t think about any of this information in isolation. We are leveraging data science and AI to combine the information from various sources, different modalities, and different approaches, and coming up with more personalized precision algorithms so that we can risk-stratify patients and better take care of them.”

CIMBID will develop initiatives in collaboration with multiple areas at Columbia, including biomedical informatics, biomedical engineering, systems biology, the Mailman School of Public Health, and the Herbert Irving Comprehensive Cancer Center. The research and clinical initiatives will share a common goal of using new approaches to understand the behavior of disease and then translate the findings to patient care.

“The ecosystem at Columbia for this type of research is exceptional,” Kontos says. “We have biomedical informatics, one of the leading departments in the country, with such strong expertise in EHR and informatics. We have genetics, systems biology, and bioinformatics, as well as a strong cancer center. CIMBID will create an environment where all these groups can come together and coalesce in a unique way that can accelerate the development of these algorithms.”

Kontos will serve as vice chair of artificial intelligence and data science research in the Department of Radiology and have joint appointments in the Department of Biomedical Informatics at VP&S and the Department of Biomedical Engineering in Columbia’s engineering school. She is also a member of the Herbert Irving Comprehensive Cancer Center and director of biomarker imaging at NYP, which is supporting the development and application of new AI tools.

While research is a major component within CIMBID, Kontos enjoys the training aspect as well.

“Training is an essential part of what we do as academics,” she says. “For me, the two most exciting parts of my work are being around students, young people with fresh ideas who make my day, and interacting with people outside of my discipline. I find that to be invigorating, and it pulls me into new directions.

"All the elements are here, and I’m very excited to be entrusted to pull these all together,” Kontos adds.