Katherine P. Andriole, PhD, a professor of radiological sciences, has spent her career focused on radiological imaging informatics and biomedical data science. She will draw on this expertise as the first associate dean for Health Artificial Intelligence Strategy and Innovation at the David Geffen School of Medicine at UCLA.
Her new appointment, along with being named director of the UCLA Center for AI and SMART Health, are a reflection of the leading role the field of radiology has played in adopting the use of artificial intelligence.
“In health care AI, advances and changes in workflow and the impact on patient care have been greatest in the field of imaging,” said Jonathan Goldin, MD, PhD, professor of radiology, medicine, and physics and biology in medicine, and chair of the UCLA Health Department of Radiological Sciences. “Dr. Andriole is uniquely trained in imaging science to spearhead AI.”
New AI developments provide new opportunities
Radiology has been at the forefront in adopting the use of artificial intelligence in health care, Dr. Andriole noted. Initially, “most of the tools were focused on detection of specific findings in images that could assist the radiologist, for example, detecting stroke in a patient who needs attention urgently, or detecting and measuring lung nodules on a CT chest, or analyzing breast imaging and so on,” she said.
However, with foundational models and multimodal data being used as part of AI models, she noted that its use and opportunities for future applications have quickly expanded.
Given the number of exams radiologists are asked to interpret and the number of images within each examination, the workload for radiologists has been growing, Dr. Andriole said. Using AI can help streamline this, and standardize certain aspects, such as how findings are quantified.
When a radiologist examines a chest X-ray, for example, AI can help them evaluate the results for a number of pathologies, she noted. The new technology can also help by automating quantitative metrics, she said, such as determining the size of a lung nodule and tracking whether it’s growing over time.
There are also numerous workflow efficiencies such as in billing and patient scheduling, including assigning patients to available scanners. AI can improve the process of acquiring images and can take the findings a radiologist dictates and create an impression for them to review and approve.
Patient-focused benefits
The hope for AI is to improve patient care, Dr. Andriole noted.
“Patients have access to their electronic medical records and, sometimes, there are terms in the radiology report that may sound scary but in reality, may not be,” she said. AI can translate the report into patient-friendly language and make the findings more easily understandable.
It can play a similar role in creating clear, comprehensible pre-procedure instructions for patients, she said, and is well-suited for translating information into other languages for patients whose primary language isn’t English.
Ongoing advances in the use of AI
As AI tools continue to be developed, Dr. Andriole expects they’ll be used to assimilate data from numerous sources, aiding radiologists in interpreting imaging data more comprehensively.
Information about a patient is often contained in multiple systems, including their electronic medical record, she noted. AI can eliminate the need for the radiologist to search these systems to gather key information such as recent lab results or procedures, which may provide important context for interpreting imaging data.
In the future, AI may also be used when evaluating imaging to identify aspects that aren’t detectable by the human eye, Dr. Andriole explained. Because AI relies on mathematical pattern analysis, she said, it may be able to home in on specific characteristics that can be helpful in making diagnoses, predicting risk for disease and helping determine the best course of treatment.
Dr. Andriole foresees that radiologists will evolve into knowledge consultants, and that the efficiencies AI will continue to provide will translate to patient care.
“A major goal is taking these tools developed by our researchers and implementing them in the clinical arena,” she said. Evaluating how well the tools work and assessing the benefits they provide will be essential.
Radiologists will play a key role in helping shape the direction of how to best use AI within the field, Dr. Andriole said. This will require educating radiologists on how AI tools work so they can evaluate them critically, know which questions to ask, let the data scientists know about their specific needs and participate in creating the tools, she said.
“The creation of an associate dean in AI speaks to the fact that AI has become central to research, education and clinical operations in the health care system,” Dr. Goldin said. “This is an important moment for UCLA to lead the way in the implementation of AI across all health care disciplines, but starting with a focus on radiology.”