AI-powered projects targeting blindness, cervical cancer to get boost through Penn Medicine accelerator program

PHILADELPHIA— Artificial intelligence (AI) is having a moment, partly due to its potential to address health outcomes and health care efficiency, and new applications of the technology factor prominently into the pair of projects chosen for the latest round of Penn Medicine’s Innovation Accelerator Program. Run through the Center for Health Care Transformation and Innovation (CHTI), this year’s iteration of the program will feature a project centered on using AI -enabled image analysis from advanced cameras to prevent blindness among patients with diabetes, and one designed to use an AI language processing technique to increase the speed and accuracy at which reports from cervical cancer screens are interpreted and communicated.

“We encourage teams to ‘fall in love with the problem, not the solution,’ so we’re excited that we have the chance to do things like reduce blindness and the cost of care by catching diabetic retinopathy earlier, and improve cancer outcomes through creating more efficient screening processes,” said Roy Rosin, MBA, chief innovation officer at Penn Medicine. “Our goal in both these cases is to help patients and reduce burdens for clinicians, who want to help more patients. AI is simply a new way to go after old, important problems like these more effectively.”

The AI-powered projects selected this year are “Diabetic Eye Screening Interventions with Teleretinal Exams in Primary Care (eyeSITE),” led by Corinne Rhodes, MD, an associate professor of General Internal Medicine, and Lama Al-Aswad, MD, MPH, a professor of Ophthalmology, as well as a PRECISE Center faculty member in Penn’s School Engineering and Applied Sciences, and “C3P3 Centralized Cervical Cancer Prevention Program at Penn,” led by Danielle Burkland, MD, an associate professor of Clinical Obstetrics and Gynecology.

About eyeSITE

Rhodes and Al-Aswad’s eyeSITE project will use AI-guided image analysis from advanced cameras to screen for and flag patients with diabetic retinopathy, one of the most common form of blindness in working-age adults in developed countries. Even if a patient doesn’t lose their sight to the condition, the costs of addressing diabetic retinopathy escalate quickly the condition is not caught early. However, only 62 percent of people with diabetes are screened for retinopathy, so the eyeSITE team hopes to create easy-to-implement procedures for using the cameras to quickly and efficiently screen and diagnose more patients.

“Our greatest worry is that clinics may receive these cameras and despite their powerful potential, leave them unused without the process we’re hoping to design,” said Al-Aswad. “Through the next months of development with the CHTI, we think we can make something that would be easy to transfer to any clinic with the proper equipment.”

Applying eyeSITE could result in earlier intervention for many people, potentially preventing blindness and the high costs of delayed detection and treatment.

“We hear firsthand about the challenges that our patients face in caring for their health, and see how screening tests, like diabetic retinopathy screening, often get delayed or never completed,” Rhodes said. “We think this is something that could really change the dynamic.”

About C3P3

Cervical cancer is the most common cancer across the world and ranks fourth in cancer-related deaths, so screening for it is vital. However, reviewing results can be time-consuming and difficult to interpret. The C3P3 project aims to simplify the process by employing an AI-guided technique called “natural language processing” to speed up and optimize how cytology and pathology reports are read, increase the accuracy of the interpretation, allow doctors to spend more time on abnormal results, and communicate next steps to patients quickly.

“In the United States, cervical cancer prevention has plateaued for the last 20 years, even with advances in screening technology,” Burkland said. “This is a cancer that disproportionately affects marginalized communities, so our goal for CP3P is to improve management of these results so much that we make a noticeable dent in the number of patients who we lose before follow-up or have significant delays in their treatment.”

Both projects will receive the full support of the CHTI’s Innovation Accelerator team, which includes funding, guidance on implementing change and staff assistance on things like the design of interventions or implementing technologies.

More than 30 other projects have received the same kind of support from the CHTI’s Innovation Accelerator, whose first class was selected in 2013. The idea was to solve problems in the delivery of health care—with an eye toward improving patient outcomes and/or clinician processes—via interventions that could eventually be scaled to the whole health system or, even, beyond.

Past projects have included the LiveAware liver cancer screening program (recently adapted by Independence Blue Cross for their network) and Healing at Home, which was designed to safely send patients home from the hospital to recover from childbirth sooner (and has since expanded to include remote screening for high blood pressure).

Project proposals are collected and considered by the Innovation Accelerator’s team from across the University of Pennsylvania Health System. Out of 30 teams that applied this year, eyeSITE and C3P3 were the chosen winners, but seven others will continue to receive other forms of support from the Center for Health Care Transformation and Innovation.

Over the next several months, the team will meet to discuss and tweak their ideas, implement small pilots testing their work, and work toward a goal of successfully implementing their projects across the whole health system.