Campus News

AI and Ophthalmology: USC Roski Eye Institute Faculty Aim to Revolutionize Patient Care

Dr. Melinda Chang and Dr. Benjamin Xu are two physicians exploring the use of artificial intelligence (AI) in ophthalmology to automate clinical tasks and improve patient care.

Eric Weintraub June 15, 2023
Artificial intelligence AI bot binary eye digital futuristic looking.
Melinda Chang, MD, and Benjamin Xu, MD, PhD

The aging population of the United States is expected to experience a significant shortage of healthcare providers in the coming decades, particularly in the field of ophthalmology. However, some physicians and researchers are exploring the potential of using artificial intelligence (AI) to automate aspects of patient care. If successful, their research could enable healthcare providers to allocate medical resources more efficiently, prioritizing patients with more severe diseases.

At the USC Roski Eye Institute, Dr. Melinda Chang, an Assistant Professor of Clinical Ophthalmology specializing in pediatric neuro-ophthalmology, and Dr. Benjamin Xu, an Assistant Professor of Clinical Ophthalmology specializing in glaucoma, are leading research using AI. While their focuses differ, both doctors are interested in leveraging AI to automate clinical tasks, streamline healthcare services, and provide high-quality care for patients in need.

“Our primary use of AI is to assist with data analysis and determine if it outperforms humans in certain challenging tasks,” explained Dr. Chang.

Differentiating Minute Details

Dr. Chang’s research specifically aims to employ AI to distinguish between papilledema and pseudopapilledema in fundus photos. Papilledema is a serious neurological condition that can be linked to tumors or other abnormalities in the brain, while pseudopapilledema is a benign condition. However, differentiating between the two can be challenging when examining fundus photos, which capture images of the back of the eye.

“Identifying the exact diagnosis typically requires longitudinal patient follow-ups and testing, including MRI scans and lumbar punctures,” Dr. Chang said. “We aim to have an answer earlier, and that’s where AI steps in.”

Collaborating with Children’s Hospital Los Angeles, UCLA, Boston Children’s Hospital, and Stanford in a multi-center study, Dr. Chang provides fundus photos to an AI model trained to classify between the two diseases. Currently, the AI model’s accuracy ranges between 70% and 80%, and sensitivity is up to 90%, surpassing human performance.

“AI is essentially programmed to scrutinize minute details in photos to provide an exact diagnosis. I don’t believe AI can replace doctors; it serves more as an additional set of eyes,” Dr. Chang said.

Dr. Chang’s future objectives include expanding the study to involve more institutions, gathering additional data, exploring the integration of other imaging techniques like optical coherence tomography (OCT) to enhance the AI model’s accuracy, and ultimately developing an application that clinicians can utilize to evaluate fundus photos and facilitate appropriate patient triage.

Automating Glaucoma Detection

Dr. Xu’s lab is currently working on a major research initiative to automate the detection and referral of Los Angeles County (LAC) Department of Health Services (DHS) patients with glaucoma using AI and fundus photographs. The primary goal of the initiative is to develop AI algorithms that will transform care and improve access for the underserved, safety net patient populations of LAC DHS.

“Glaucoma is endemic among patients of LAC DHS, where it commonly leads to permanent, severe vision loss. The wait for a glaucoma evaluation typically takes more than six months, even when patients have severe disease,” Dr. Xu said. “This is a significant problem because valuable time is lost when these patients could have received sight-saving treatment.”

Dr. Xu has utilized AI in his research since 2017, at first to address the need for novel clinical tools to analyze OCT images and detect patients at high risk for glaucoma. “OCT provides high-quality images that are effective for identifying patients at risk for glaucoma. I ventured into the field of AI to automate the analysis of these images so that OCT can be used more conveniently by physicians.” Dr. Xu elaborated that analysis of some types of OCT images requires time and expertise that physicians often lack. His lab aims to develop tools that simplify this process for physicians to improve the quality and efficiency of glaucoma care.

Dr. Xu recently secured a two-year grant from the Southern California Clinical and Translation Science Institute (SC CTSI) and LAC DHS as well as generous support from the Carlson Family Foundation to support his ongoing research. The funding allows his lab to develop preliminary AI algorithms for implementation in LAC DHS glaucoma screening clinics. Success in this endeavor will enable Dr. Xu and his team to demonstrate that their bold ideas are feasible and seek additional funding from the federal government.

“It is crucial to emphasize the ambitious nature of this initiative in LA County. There are no other municipal health systems attempting a project of this scale,” Dr. Xu said. “Current challenges lie not in developing AI algorithms, but in implementing them. It requires merging laboratory research with real-world clinical practice. That is why this proposal is so exciting: we are leveraging data from tens of thousands of patients and directly giving back to the same patient communities by developing AI algorithms to serve them.”

Ethical Considerations

As the ethical considerations of AI have become an increasingly hot topic in 2023, both Dr. Chang and Dr. Xu were open about addressing the ethical concerns arising from the incorporation of AI into healthcare.

Dr. Xu emphasized a primary concern surrounding AI’s proficiency to generalize beyond the population from which its training data is obtained. “At present, our AI algorithm is developed using data from Los Angeles County Department of Health Services, where the majority of patients are Latino. While our algorithm may work effectively for Latino individuals, its performance may degrade when applied to individuals of other races and ethnicities. There should always be concern regarding potential biases that are latent due to unique attributes of data sources used to develop AI algorithms.”

Dr. Chang shared similar sentiments, underscoring the importance of AI’s ability to generalize across diverse populations. “That is why we have established partnerships with other medical sites nationwide and aspire to collaborate with even more to ensure a more diverse sample.”

Driven by these considerations and with the goal of establishing a large sample size at the forefront of their missions, both Dr. Chang and Dr. Xu are committed to discovering how AI can be harnessed for the greater good and the benefit of patients and healthcare professionals alike.