Using Advanced Computing to Create AI-ML Diagnosis Toolbox

From left, Miller School of Medicine radiologists Dr. Fernando Collado-Mesa and Dr. Alex McKinney examine a breast MRI at the Taylor Breast Health Center at Jackson Memorial Hospital. Photo: Diego Meza-Valdes/University of Miami

Using Advanced Computing to Create AI-ML Diagnosis Toolbox

Over the years, new technology has helped Radiologists diagnose illness, but it has also changed their jobs. To diagnose illness in the past, physicians spent time speaking with patients. Today, they spend more time scrutinizing images, patient electronic medical records, and other data sources.

“We believe the next iteration of AI
should be contextual in nature”

Fernando Collado MesaCurrent available artificial intelligence (AI) tools are limited to a specific type of medical image, and cannot, for example, analyze both MRI (magnetic resonance imaging) and ultrasound at the same time. In addition, the patient data used is typically not inclusive of a range of demographic groups, which can lead to bias in health care. To address these limitations, University of Miami Miller School of Medicine Radiologists Dr. Fernando Collado-Mesa (pictured at left) and Dr. Alex McKinney (pictured below) are working to create an AI tool based not only on imaging data, but which also considers a patient’s unique background and circumstances including risk factors like race and ethnicity, socioeconomic and educational status, and exposure. “We believe the next iteration of AI should be contextual in nature, one that takes into account all of a patient’s risk factors, lab data, past medical data, and that helps us follow the patient,” said Dr. McKinney, who is also Chair of the Department of Radiology. “Additionally,” added Dr. Collado-Mesa (Associate Professor of Radiology and Breast Imaging, and Chief of Innovation and AI for the Department of Radiology), “it will become a form of augmented interpretation to provide better patient care. This tool will not just say yes or no, disease or no disease. It will point to the data around it to consider a variety of issues for each individual patient, and put its findings into context, including future risk.”

URIDE logoThese goals and efforts led Dr. McKinney and Dr. Collado-Mesa to the Institute for Data Science and Computing (IDSC) Director and Vice Provost for Research Computing and Data, Dr. Nick Tsinoremas. Dr. Tsinoremas (also a Professor of Biochemistry and Molecular Biology, Computer Science, and Health Informatics) and IDSC’s Advanced Computing team came up with the idea to add the deidentified images from the Department of Radiology to an existing UM tool, URIDE, the University Research Informatics Data Environment (a web-based platform that aggregates deidentified patient information for faculty research).

A first version of the toolbox will be unveiled later this summer and new elements will be added as more imaging data is gathered. It will include millions of CT scans, mammograms, and ultrasound and MRI images, along with radiographs, Dr. McKinney pointed out. “We don’t want to rush this because we want it to be a high-quality, robust toolbox,” said Dr. Collado-Mesa.

Both Radiologists and Dr. Tsinoremas hope this new AI tool will help answer vital research questions, like: Which risk factors lead to certain brain tumors? Or, What are the most effective treatments for breast cancer in certain demographic groups? It will also use machine learning (a technique that constantly trains computer programs how to utilize a growing database), so it can “learn” the best ways to diagnose certain conditions. “Creating this tool can help with diagnosis and will allow predictive modeling for certain illnesses, so that if a person has certain image characteristics and clinical information that is similar to other patients from this database, doctors could predict the progression of a disease, the efficacy of their medication, and so on,” Dr. Tsinoremas said.

“AI has the potential to advocate for the patients,
rather than a one-size-fits-all approach”

Alexander McKinney

To ensure the new tool will be unbiased, the team also plans to add more images and data on all population groups in the community, as it is available, as well as monitor the different elements constantly and systematically within the toolbox to make sure it is performing properly. The project will focus, first, on illnesses that have a high mortality or prevalence in the local population—like breast cancer, lung cancer, and prostate cancer—and will add others over time. This will allow doctors to spend more time with patients and offer more personalized, precision-based care based on the patient’s genetics, age, and risk factors. “AI has the potential to advocate for the patients, rather than a one-size-fits-all approach to medicine based on screening guidelines,” Dr. McKinney said. “This could help us get away from that, and will offer more hope for people with rare diseases.”

As data is added in the future, the researchers hope that other physicians across the University will also use it to conduct medical research. “This will be a resource that any UM investigator could potentially access, and it could spark a number of different research inquiries to describe the progression of disease and how patients respond to different treatments in a given time period—and these are just some of the questions we can ask,” Dr. Tsinoremas said.